The Official PC building thread - 4th Edition

We have plenty that are 16G to 64G, most though are around 16GB at work. And that's not Flash memory either, but many of the original netbooks were like Sam said.

 

I'm sure they're better now, but in all honesty since the advent of the ipad/galaxy tab etc, I just don't see netbooks any more apart from the old ones people still have lying around, like the original Eee PC and NC10.

 

Bought one of the 3000rpm Gentle Typhoon fans today to try and get the temps down a bit on my CPU when it encodes away (as well as cleaning out the heatsink, that no doubt helped!) - Am down from 71ºC to 53ºC which is a bit more bearable!
Interestingly, you can hardly hear it at full speed over the array of 1900rpm fans in the side panel due to the lack of vibration from the CPU fan mount, it's really quite impressive. Sure, not quiet, but it's very tolerable. If I had known, I might have tried the 4200 version, although I'm not sure how much more of a difference that would make.

 

Later in the game you get some luxury cars, as when the prince gives you his SUV. It has a great big JEEP written right on the front of the car, and boy does it drive smooth, lol. I wonder what Jeep paid for that product placement.

Very true - which is why boat travel is essential. Not the same hassle - quicker, more fun, and more scenic.

I have been away for about 3 weeks in doing a lot of testing - and I have a Heaven crossfire score to post for Jeff, and I need to do a full catch up here, but let me just post a problem I have encountered:


BF:BC2 IS INCOMPATIBLE WITH 7950 PER LATEST CATALYST DRIVERS AS OF THIS DATE

Guys, I have encountered a major incompatibility between the 7950, and BF:BC2. (By the way, have you seen the cheap prices on these cards now - only $300 - wow!)

Every other game, no problem, and the card tears through the games at 2560x1600, everything completely maxed out, with just one card, at fps of 30 or more, except in maybe a total of 5 mins of one of the crysis games, when I noticed per the on screen display (msi afterburner display, but I use trixx for the graphic card settings) that the fps had dropped to about 24 and I could feel just a slight bit of lag. The strange thing was that the huge fps disparity in Crysis between the beginning of the game and later sections, never appeared.

If the snow was supposed to be slow - it wasn't - all the 1700+ shader cores tore through that snow. I maxed AA and everything, and the scenery was beautiful. I finished Crysis, Crysis 2, and Warhead. With one card.

I am running the core 2 quad 9450 at 3.343 ghz, fsb 418, per Sam's notes to me of about a year ago on some basic settings to use. I never got to fsb 450 = 3.6 ghz - too much heat. I have been running perfmon.exe, the built-in performance monitor. My cpu usage on some of those games was around 70% on average, and that's also what the OSD told me.

When I got to Sleeping Dogs, I maxed it out with DX11 and the several gigabyte high texture pack. But my fps was only 22, with my 7950 card at 975 core clock, and 1350 memory clock, 1087 vddc. When I added the second card in crossfire, fps jumped to around 40, very nice scaling. But to be honest, I really couldn't tell the difference that much. When I encountered the BF:BC2 incompatibility problem, I actually thought I had ruined the cards by attempting a little higher clock at 1025 per a review that I read. Right after that I tried running BF:BC2 and all hell broke loose, and I was in the dumps for a couple days, until after a lot of testing, I proved that the cards are okay, but it's just that one game. By then I had taken the crossfire apart, removed one card, and tested them each individually again.

(I'm in the new spedo case, total 11 fans, not counting hsf or psu fan, as opposed to the 3 fans in the antec sonata case - 6 intake: 5 intake 140mm, including one on the floor, 1 intake 120mm in a custom kama bay which I designed top front - I'll post later about that, 4 exhaust: 2 exhaust 140mm, 1 exhaust 120mm behind the motherboard, and one ceiling exhaust 200 mm, intentionally slowed from 700 rpm to probably about 500 rpm to maintain positive case pressure, as the new demciflex filters significantly drop intake cfm probably at least 50% - I'll post on that later also, and one internal 120mm fan pointed at a slight angle up toward the cpu cooler. No fan controller - case wind noise probably is around 25-30 db, I can whisper and hear myself easily - nowhere near the noise of the kazes - but definitely not silent.)

The card that tore through all the crysis games, develops major artifacts and hangs in BF:BC2 within 10 minutes. That's the first HIS IceQ that I bought. The first sign of problems is when the OSD reports 0 temp, and 0 load. As long as it is correctly reporting temp and load, it is perfect. I can end the game, and go back to windows - everything is fine. As soon as I lose the OSD report of temp and load, any retreat to windows hangs and forces me to reboot.

When I run the second HIS IceQ card, I can get about two hours of game play out of the card, before it hangs. The same loss of sensors occurs within 10 minutes, and the same crash on attempting to leave the game occurs, but if I continue to play, I don't get any artifacts.

So obviously, that would be the card I should use in slot 3, and the original in slot 6. I can disable crossfire for that game, and count on about two hours of gameplay. I actually enjoy that game as much or more than BF3. By the way, the 7950 also maxes BF3, at ultra textures, with one card, 2560x1600, around 32 fps. It seems perfectly smooth at all times. Yes, crossfire pushes that fps up near 50 if not more, but again, I can't tell the difference, lol. The problem with BF3 is that nobody is playing it - in comparison to BF:BC2. So the maps are mostly empty.

Anyway - Jeff - do you know anything I should do to solve this problem with BF:BC2. I ask Jeff because he was a big BF:BC2 fan - and I in particular love the Vietnam maps, which almost nobody plays anymore. But if anybody knows anything to help, Sam, Russ, ddp, movies, deadrum, blaze, Kevin, let me know please.

In going through the newegg reviews on this card, and the 7970, I have encounted many comments to the effect that some people have gone back to earlier catalyst revisions, as they found that earlier games that they had played, which were perfect, no longer played right with the newest catalyst.

OTHER STUFF

Thanks anyway Russ, but - doggone, serves me right for being gone so long - "video has been removed."

Everybody is talking about the moisture from the cans - but isn't it non-conductive? I just had the same experience.

Scary! A good wakeup call for all of us - backup backup backup.

That reminds me of last Spring's research on Raid and disk drive failures, after Miles' hard drive failure on his server hosting his irreplaceable child photos (fortunately only the operating system drive failed, but it scared the cr*p out of him, so I created a system for him with three sets of Raid 1 mirrors.) I did a lot of reading about bad sectors and such - and got intimately familiar with sector failures, including the two new 1TB enterprise drives that I ruined with the faulty power-over-esata card, with insubstantial 5V power, which I was luckily able to RMA back to newegg.

I believe that the full format forces the drive to read and write every single sector, and map bad sectors to spare sector locations, as Steve mentioned. Additionally, after a full format one should be able to read the SMART data, with speedfan or with some of the hard drive testing utilities, and get some info from that, and maybe spot some red flags, like pending sector count, etc. As I remember from the wiki article on SMART (which I actually edited regarding one faulty paragraph - lol) according to a google study, a drive with any sector failures at all, is twice as likely to fail relatively soon as compared to a drive with no sector failures.

I too have never performed a full format, but now I'm thinking that I should change that and adopt Kevin and Jeff's habit.

What the hell? I thought you said it was a bad power supply. Ever since your problem I have been watching the heat around my 30" dell, and I placed a special fan to blow on it during the 100 degree days we recently went through. The upper edges can get quite hot, and I keep telling myself - "remember Sam's power supply failure" lol.

Rich

 

Somehow missed this last post! Time to get reading...

Your performance issues sound like the infamous "bad memory" bug with dual graphics, where once a particular object in the game has been rendered, it faults dual graphics and you end up using a single GPU's worth of performance (often worse) until either an alt-tab, or closing/reopening the game entirely. It's an issue that affected Flashpointragon Rising, and also affected Left 4 Dead 2 for the first month or so after its release until it was patched. Your symptoms are almost the same.
If you saw my post from a little while back, you'd see Sleeping Dogs runs really badly on any system, it's just too demanding.

The 3TB drive's data is slowly being restored, I've got about 1TB worth of the 2.4TB lost so far, (but also 1.5TB of additional data the 3TB was going to partly contain in the process!) Quite impressively, I managed to pull in excess of 2.5TB through my VDSL in about 10 days.

The fault with the monitor was indeed the power supply, but specifically one schottky rectifier on the DC side of the power board. Googling the symptom eventually led me to a forum thread on badcaps.net with a plethora (I gauge at least 15 people) who all have suffered the same failure with the 3008, and every one of them fixed the problem with the same repair - a replacement of the Vishay V30120S rectifier diode with a higher-rated (reverse voltage) equivalent from ST Microelectronics. Choosing the higher-rated of the alternatives, the STPS60170CT and replacing it, so far so good as of a fortnight.
I confess it wasn't me that actually did the soldering as I was always quite weak at soldering due to having unsteady hands, but one of my fellow LAN-goers at multiplay took to it in our hotel room and resurrected it - just as well, as it was quite a fiddly job - a bit like stripping an engine in that you have to take loads of functional components off before you can reach the problematic part!

The issue with the 3TB drive has seen a few more backup drives be purchased, and they have now been added to the Icybox IB-545SSK that was producing the extra capacity in my server before it was rehomed to the 4U case, and has now been fitted into the fourth machine (Phoenix), which now serves as a network backup device, alongside its primary role of offering a hardware dedicated XP box.

Hard disk inventory now sits thus (Using pastebin as it's a bit long to put here):

http://pastebin.com/tx05H5g4

 

I'm happy your Dell is happy again I would probably repair my Dell as well. Rather difficult to throw away a $500 piece of equipment, let alone 1200 - $1500!

I've toyed with soldering a bit. Soldering an expensive piece of equipment would have me rather nervous LOL!

 

If there had been a new version that offered substantially improved specs/functionality, I wouldn't have lamented the loss so much, but given that the best out there is the U3011, which is just an LED-backlit version that has imo slightly worse connectivity, there was no reason to switch. I'm hoping (although very sceptical) that Intel's predictions for an increase in display PPI/resolution finally comes to fruition towards the end of next year. Still, given how unsuitable Apple's retina displays are for windows use, I wonder if companies will bother. Fingers are still crossed!

 

That isn't quite it Sam, because as I mentioned but perhaps didn't completely clarify - I took the crossfire apart and test only one card at a time.

The original card that I thought was "bomb proof" loses sensor reading on temp and load within 10 minutes, and develops artifacts and crashes soon after that - in most cases RIGHT after that.

The second HIS IceQ card, tested by itself, the only card in the case, in slot 3, loses the sensor readings also at the 10 minute juncture. But it will keep on going for up to two hours. It never develops artifacts - it just eventually hangs and a restart is required.

So it isn't your crossfire problem. What do you think it is? My guess is - driver incompatibility. Perhaps I could roll back drivers. Jeff - this game is one of your past favorites (I like it better than BF3) so do you have any ideas for me?

----------------

SLEEPING DOGS IS FULLY PLAYABLE AT MAX SETTINGS, 2560X1600, WITH ONE 9750 OC'D TO 975

Yes, I noticed your Sleeping Dogs chart - I read it after I posted, while going through three pages to catch up on the forum. You showed average of 21, and I showed 22 - which I think was my minimum. The difference there is the stock clocks of the 7950 are 800, but I am running it at the easily-achievable and stable overclock of 975 which Blaze said puts it at the 7970 performance level.

Kevin, as much as you enjoyed GTA4, you might like this game - very similar. Don't let the karate frustrate you. Pay attention to some of the walk-throughs, and I'll give you pointers any time. (I was initially so frustrated with the karate I was on the verge of erasing the game from my hard drive.)

I have continued to play the game after dismantling crossfire and putting the 2nd HIS card in the box for now while continuing to test the first card - the one that comes apart after 10 mins in BF:BC2. I mostly get fps in the high 20s out of it, and I haven't noticed any lagging. Part of that possibly is due to the fact that it is a 3rd person shooter, and I think, like GTA4, being in 3rd person automatically makes it a little harder for me to feel a lag.

So for me, the game is completely playable with one card, at 975 core clock, 1350 memory clock - which is just a slight memory overclock from stock memory clock of 1250.

But with two cards, it definitely gets in the 40s, maybe even 50, so even for the most sensitive of you, I believe the game is entirely playable with two 9750s in crossfire. Sam, now that your 30" display is working - by the way huge congratulations to you on that - I know you are thinking about waiting for the 8000 family and that might be a good thing to do. (But those 7950s are really dropping in price.)

The thing I found hard about the game, was not any lag issue, but it was the karate. And since I seemed to do better in the earlier parts of the game, I didn't mind playing it with just one card. I reasoned, with broken logic, that "I was handling the karate better earlier, and now in crossfire I am getting killed every single time on the Stickup and Delivery mission, let's see how I do with just one card."

But it had nothing to do with one card or two - it had to do with technique, and health. I looked at a walk-through that suggested I go and visit some of the health shrines, like the one next to Wei Shen's apartment. For each 5 you visit, you gain 10% health. I found a walkthrough that claimed to have all 50 shrines in the 4 boroughs of Hong Kong. I enlarged each map, and systematically visited every shrine on every map, over a total of about 10 hours of play. I visited some gorgeous scenery - the level of detail in the game, with the high texture pack, is just awesome.

I found a bunch of suitcase, at roughly $5k each, and ended up with near $200,000 in cash!

I also found about 5 statues to return to my former karate master, which gave me 5 lessons - it looks at first like you are only getting one lesson, but then when you leave the dojo, your available missions again show "RETURN THE STATUE" until you actually return all of them.

With a lot of trepidation, I returned to that mission, Stickup and Delivery, and after two attempts, I beat the mission.

I think they put that mission in there to kill you.

I am serious. They allow you to keep restarting it, or to cancel the mission.

I read on google guys talking about how hard the karate was. Your heart starts pounding - which is a sign you are low on health - and you can stand off facing the remaining bad guys - you can wait for an hour if you like, and your damn racing heart is still racing. That made me mad - that is not real world. But I realized it is just an audible health indicator, so while I was low on health, my heart would always be pounding. With a pounding heart you are close to death.

When I was out visiting all the little health shrines, I found that I didn't have to always fight those bad guys guarding some of the special money suitcases I wanted to open. I just used my scooter or motorcycle, or car, and assassinated them that way, lol. Yeah, the scooter - just drive it up the steps and run over the guys. And they don't really even bother you as you back up slowly and clumsily, in order to run them down again. Hahaha. One time after doing that with a car, a cop car came by but he didn't seem to be bothered by all the bodies lying around. A few of the suitcases had a gun. You can pull out a gun, but you can't put it back. But if you get into a car or jump on a motorcycle, the gun is away when you get out of the vehicle.

If you have to, and there is no vehicle close by, you can drop your weapon. But then you lose it of course. One time with it out, I killed some guys, but the police are all over you - I had to kill one of my fellow officers of the law to get away. Then I dropped the gun and ran off, and the police alarm disappeared. So pulling a gun is in a different category than pulling a tire iron, or running over somebody which doesn't trigger the big police alarm.

I was burned out on the karate, and pleased to discover running people down. Haha. By the way, you only get a few save points, so I was constantly saving over two separate save points, according to the time-stamp on each, saving over the earlier one. That way I wouldn't lose all the loot I had picked up or the statues, if I got killed. Plus I didn't want to pay a big hospital bill, I preferred just to go back to a save point, which starts you out in your apartment, then set the goal on the map, and speed back to where I was.

Anyway, finding all those shrines was mostly about getting ready for that one mission.

However, regarding vehicle assassination, on that mission that kept killing me, Stickup and Delivery, there is no scooter nearby. Also, carrying a tire iron doesn't help - the tire iron magically disappeared from my body when I was faced with the 6 guys to fight.

But this is what helped:

HOW TO BEAT THE KARATE AND SURVIVE THE DIFFICULT STICKUP AND DELIVERY MISSION
As I say, I went around and found those 45 shrines. That gave me a health boost of 90% more health - but you have to eat food to get that health. I also found 5 statues, which got me 5 karate lessons, which I believe unlocked karate moves that the character started to do somewhat automatically. I also did some police and triad missions, and I chose to increase damage or increase resistance to blows, any time I had a choice about what payoff I could take from my police point or triad point.

In regard to health - there are 3 health indicators. On the map, there is a bowl with flame - that is the apothecary shop where you can buy tea. There is also a little booth when you leave your apartment. I accidentally bought the tea, thinking it was the power drink. The tea gives you increased resistance to damage!!!!!!!!!!!!!!

Then there is the power drink. That is on the map - it looks like a beer mug. That gives you increased melee damage - increases the damage you do when you punch somebody. It is the same as the power drink in your refrigerator in the apartment. So I took one from my apartment, then on the Stickup mission, I pulled the van over and drank another one before I turned the corner to the stickup place, because the effect slowly wears off.

Finally the third health indicator, is to just eat some food, which gives you overall more health. The little stalls that sell pork on a stick, or noodles, are not on the map. The only food place I know of for sure, is the ice cream next to the Bam Bam club, which is on the map near the apartment. Ice cream does not quite give you as much health as pork or noodles, but it is better than nothing.

So when you drink the tea, drink the power drink, and eat any kind of food, you have increased damage, increased resistance to blows, and more general health, and you will see all these indicators on the right side of your screen, as 3 closed fists.

So when I finally beat the mission, my heart was not racing.

There is one more tip. As you practice in the dojo - the W key of course moves you forward, but when you add the Space key, you run forward. A lot of the moves have you running - so that is W plus holding space. Another critical keyboard key, involves grappling. The default is F. That key allows you to grab a guy around his neck, and with w plus space you can run with the guy. If you hit E, the use key, you can then flip him over. Running with the guy and flipping him are very helpful, because if you are attacked by a crowd, as they start to punch you, you are automatically moving away from them and you don't really have to counter their offensive move. I ended up putting grappling down on the C key, and then I moved it to the X key. So when I fight now, I use a lot of W and space bar with my thumb, and X key with my pointer finger. On the mouse I am aiming toward the guy I am running into, I hit left mouse to do a punch, and if somebody turns red, I might have time to hit the right mouse key to counter the punch before I take damage. Very rarely do I finish a grappling move with the E, flip the guy, because in the dojo you also learn a power flip, by running and hitting the grapple key just before you reach the guy, which causes you to power flip him and slam him on the ground. So it's just mostly W and thumb on space, and finger on X. Lots of left mouse, and some right mouse thrown in, but he is pretty much now finishing the guys off that he flips, because the dojo lessons unlocked some finishing moves.

Even after all this - I got into trouble at the Bam Bam club, on the Bam Bam mission. I was fighting on the second floor, and accidentally jumped down into a thick group on the floor below. I didn't mean to do that. I got killed. I redid the fight, stayed on top, finished them off, then came down. Sometimes you can move away and get a new fresh run at the guys, but I find that the new finger positions are helping a lot, and as I say, the constant running, meaning the W and space held down all the time, is giving me distance between the others who would like to attack.

I try to eat and drink and drink, before any mission that I think will have karate. Now I got to the gun missions - but in the Warehouse fight, I had to do a lot of karate first, before picking a gun off a guy. Now in guns, I am back in my element. Just go for a head shot, and it's a one-shot kill. Line up your circle, then go back into cover before they shoot you. Get your circle lined up. Then when they come out of cover, go out with your aiming circle, and try to frame it exactly around their head for the one-shot kill. That is ever so much better than karate, lol.

The car races are pretty cool. It is wild to think that anybody could have a car race through Hong Kong, but some of the races actually have opposing traffic!!! They are fun and intense, and not too easy to beat, but not as bad as the karate. I have saved some screen shots, but I see they are not in the local folder. If anybody has any interest in the game, I'll do another post with some screen shots. Hong Kong is beautiful, and diverse. Boating around the harbor and out around the islands is awesome!

Rich

 

Really wish I could help you out with the BC2 glitch Rich, but have never encountered the like. AMD always had shoddy driver-side support for the game to begin with. Crossfire scaling fluctuated wildly between driver releases for me. Interestingly the Vietnam maps always scaled perfectly while the vanilla maps did not. Might have to re-install it.

 

Reinstall the game? Hmmm, well thanks for the comments, Jeff.

Last night BC2 was pretty good for 3 hours straight - the Vietnam maps had one good solid group of players on a rotating server with reasonable 65 ping.

Here is one of my favorite maps - Phu Bai valley. It is a beautiful map, and despite the tanks and helicopters, there are plenty of places to take cover, below the ridge of the Hill position whose flag we are trying to capture, in the huts of the fishing village, or even in the water sometimes, or in the tank-proof bunkers of the A first camp right next to Marine spawn. It is one of the best maps that anybody has ever designed.



I had one particular battle that was kind of fun. A GI had just killed me in the Fishing Village - which you are looking at the edge of in this picture. He didn't go for flag capture, so I respawned in that village and went looking for him. I caught him hiding behind a small shed, and fired my PPSh and tossed a grenade. His character yelled out "Take cover, grenade." That's when you know the enemy is around - there is a lot of built-in talking on those maps. I fired some more but he was well hidden, and then I switched to my rocket, and launched. The hut was destroyed and the screen said "50 points killing enemy." Haha. I went over to his lifeless body which disappeared 10 seconds later. That was fun - I rarely pull out my rocket launcher unless I am battling a tank.

Another fun time was when I switched briefly to sniper rifle - a guy in a tower at the GI spawn was harassing me and finally got me in Camp A. I spawned in fishing village and crept over toward the tower with the M40 one-shot powerful gun. Just his head was exposed. Bang. Then in the fishing village, there was a GI way hunkered down trying to take the flag, just a bit of his head exposed. Bang. No more flag capture, lol. The sniper also comes with mortar attack by using special binoculars, but I am out of my comfort zone with that gun. I prefer assault class with M16, or engineer class with PPSh submachine gun.

The only real weapon upgrade on those maps is the upgrade to the Garand rifle that I am often killed by. I don't know when I'll ever get that - I am still in warrant officer class. While perhaps not as good as the Garand, my basic m16 works pretty well, even with iron sights, and is quite powerful, and the Russian PPSh submachine gun that you get for engineer class has surprising distance on short bursts. The Vietnam map packs are still my favorite mp games - right up there with the Left 4 Deads.

I don't know why BF3 is not doing it for me - besides nobody on the maps, the washed out colors don't look that great to me, graphically. And while the iron sights on the m16 and m14 are the same, they somehow block the screen more than on BC2, and there is more muzzle flash - so it ends up being almost impossible to track a moving target which I just don't run into on BC2. On the vanilla BC2 maps I have my FC2000, with either red dot, or 4x scope, so I can't complain at all about levelling up.

Using the first street of the Heaven benchmark as an example, with the 1920x1080 16x anisotropic filtering, normal tesselation and 4xAA settings you suggested, Jeff, here are my new graphic horsepower options, in terms of fps on that first street.

A - Low Power: 43 - crossfire disabled and stock 800 1250 clock
B - Medium Power: 51+ - crossfire disabled and 975 1350 overclock
C - High Power: 83 - crossfire enabled and no overclocking
D - Ultra Power: 94 - crossfire enabled, ULPS disabled, Trixx overclock to 975 on both cards

The ULPS, ultra low power state, must be turned off, says Trixx and ITurbo, to overclock both cards - I found that AMD Catalyst, in crossfire, only overclocked one card, which apparently has to wait for the other, so you get no benefit - no use even overclocking one of them.

So last night on A - Low Power, I was getting 38 fps on BC2, everything maxed, HSOA giving a really nice look to everything, which I didn't have before the 7950 cards. I have been playing Sleeping Dogs on B - Medium Power with minimum 22 fps from time to time, but mostly upper 20s and fluid. I might go to C - High Power on that title which puts me in the upper 40s for fps. Next time I play I'll try that and see if I can tell any difference. I really could not tell any difference the last time I played it with crossfire enabled, but I'll pay closer attention.

If I ever go back and play Metro 2033 and try to max it, graphics-wise, I'll use D - Ultra Power, but so far no current title needs that much power, and to my surprise I sailed through all the crysis titles fully maxed out including max AA, on Medium Power. (By the way, Jeff, I still need to post a Heaven screenshot, but I hit a score of over 2900 at Ultra Power + Boost, meaning both cards actually at a little higher core clock of 1000, memory 1390.)

It's nice to have these graphic options, and to be playing just fine on Low for right now. I plan to pick up Medal of Honor Warfighter as soon as it comes out at the end of next month, and I'll try to get right into the multiplayer so I can level up to some decent weapons so I'm not the only guy out there with iron sights.

Speaking of weapon advantage, I tried out the new Counter Strike: Global Offensive, known as CS: GO, in the casual play. The old favorite maps were back, slightly dressed up. What a great game, and the only way to get a better weapon is to save some money and buy it right then and there. I finally got my mic working on push button with the C key, and the game is as fun and as addicting as ever. The mic adds a nice little element to the game, but keyboard chat is good too - especially team chat planning our next move as terrorists. "Rush up the middle short A." That type of thing.

Last night on BC2 during the 3 hours before the hang, I got a special treat - the server rotated over to that one map with the burned out hill on the Vietnamese spawn area



where fires are still raging, and the Cong need to break out and get over to the other side where it is still lush, which leads down to the lagoon and the boat. There is one tank on the burned part, but they didn't use it - maybe the server removed it. I haven't played that hill for a long time - it was great!

Anyway, I am getting about 38 fps on BF:BC2 - completely smooth and fluid - with my A - Low Power setting, crossfire disabled, and no overclock at all. My most stable BC2 card, HIS card #2, is now in the main slot. As I say, I got three hours out of the game last night before a hang. So that is encouraging - I guess my setup is stable enough - anything more than an hour is gravy.

I'll do a little more googling on the subject. The loss of the temp and load sensor readings is very strange - I never came across that EVER. Sounds like bad components - but only on that one game - no other - puzzling.

Rich

 

Jeff, here are some Unigine Heaven results finally a month late, on crossfire 7950, with the settings you recommended, normal tesselation, 16x anisotropic filtering, 4xAA, and 1920x1080, full-screen of course to enable crossfire.

Here is the score with the totally stable 975, 1350, using vddc of 1087 which I am not sure is any kind of overvolting. For the cards I settled with, the HIS IceQ, stock clocks of 800 come with vddc of 981, however pressing reset on trixx with certain other 7950 cards (I tested 3 additional brands, Power Color, Gigabyte, and XFX) resulted in a reset vddc of 1087 on a few of them. So - doesn't really appear to be an overvolt - and I am not overvolting memory at all. Stock is 800, memory 1250, vddc 981. Here is my super stable 975, 1350, vddc 1087:


And here is with a little more aggressive one that I tried after reading a magazine article on this specific HIS IceQ 7950 card. But for sure 1025 was not stable - it hung about 3 minutes into Heaven. The 1000 however did seem to be stable, but I didn't test it exhaustively. Right after this I tried running BF:BC2 for the first time since I started testing these cards, and I got artifacts, hanging every 10 minutes, and I was very depressed for two days thinking I had ruined both cards.


How do those scores measure up, Jeff?

I almost got to the 3k mark - maybe 1010 with 1400 would have done it (but 1025 was unstable.) I guess I'll call this my Ultra turbo - that's a 4.3% performance increase on a 2.5% increase in core clock, and 3% increase in memory clock. But I already had a huge scare with the BF:BC2 problem, so I'll stay away from this ultra turbo for now, unless you want me to try to crack 3000 for you, lol.

Rich

 

Well Rich you are going for long-term reliability no? That means you'll have to settle on a steady set of clocks and leave them intact. I also don't quite understand why you keep enabling and disabling Crossfire. Unless the game specifically has a problem with it, wouldn't you want it on?

My advice would be to find a set of clocks that are stable, not necessarily the best performance. Any boost you get from OCing is already free performance. You might want to focus on making it work well, not pushing its limits.

My cards OC from 775 core to 900 core pretty easily. I can push to 940/950 in some games, but it's simply too much heat and too much uncertainty for a very small return. So I settle at 900 as an already sizable increase and well within stability on every game. Likewise my memory will go comfortably from 1000 to 1150, but memory clocks don't normally show very large gains, and stability isn't perfect on everything so 1100 has become my middle ground.

I'm not saying to stop trying to find the sweet spot. But don't worry about what others claim to get with the cards, and just take them to what works. Many people claim to get 1000/1200 with cards just like mine, but that's with a significant overvolt, and video cards are NOT very tolerant of overvolting, despite what many would say.

Video cards are not as durable as CPUs. If you want them to last, you need to stop overvolting. Some might disagree with me, but I've seen enough people fry hardware to know.

As far as you scores go, they are well above mine. I have currently reverted everything to stock for now though to lessen the load on my dying PSU. I will be getting a new Seasonic unit soon but until then an Antec NeoPower Blue 650 will be my replacement PSU. I will be testing it to see if the flaky Corsair 620HX was the cause of my stability issues when OCing.

About the time I get my Seasonic PSU I will more than likely be getting an X6 1090T as well so we will be able to make a more fair comparison. 6 cores at 4GHz+ should be a nice boost over 4 at 3.8GHz.

 

Well, I suppose you're right, but my favorite, BFBC2 Vietnam map, or the vanilla map, definitely has a problem with the card in slot #6 - I get only 10 mins. So I have to have crossfire disabled in order to get 1-3 hours out of that game before a hang. That was specifically why I changed the positioning of the two cards after testing determined that, for whatever reason, one of them is especially incompatible, and the other not so much, lol.

Otherwise, you're right, it would just be easier to leave crossfire enabled, at stock clocks, and run like that all the time.

Why not? Well, maybe why not - to save electricity. This is Sam's old subject of going green, and I guess he's right, to a point.

Let's think about it. If the tdp of one card at stock is 160 watts, the line power coming in before PSU efficiency robs some of that power and converts it to heat inside the PSU itself, is more like 200 watts, since 80-85% is a good PSU efficiency, as I recall. At overclock, the card pulls up to 200 watts, meaning line in of maybe 240 watts. Let's say that non-ULPS disabled crossfire (when I disable crossfire that also disables ulps and the card runs at 500mhz, however it draws only 6 amps vddc current in, versus 60-80 amps under load, so only 1/10th to 1/13th the power draw compared to full load) saves me 200 watts. For every 5 hours, that is one kilowatt hour, which costs me 31 cents at my top rate. So I save 6 cents per hour leaving crossfire disabled.

You played one game recently for 400 hours - I forget which one.

A savings of 6 cents an hour, saves $24 over 400 hours of gameplay. I can game that much within two months when I am hard at it - 7 hours a day average - you bet. So a really active gamer could save $150 a year. Again, that would be a reason not to have crossfire enabled, if the machine could handle the game with one card.

That's why I like ULPS. They actually turn off the second card completely, in Ultra Low Power State, when not needed. But I only get that by enabling crossfire, and by not overclocking the cards. That combination is my High Power 83 setting. There is a setting in Sleeping Dogs that allows you to cap the fps rate.

Now that I’m talking about the game, I will finally post a few screenshots that I took with Steam – hit F12 and you have the screen automatically, in my case in full 2560x1600 resolution, which I scaled back to 1000 wide for the full shots here, and then pieces of the shots in full resolution so you get the idea of how much detail there is.

But seeing these, I missed out on many exciting shots, like motoring around the islands, participating in the car races, not to mention all the karate fights, or the practice in the dojo.

Kevin, maybe these shots will get you to finally save up and go for a 30 inch display – I know you want it, lol. You’ll need two 7950s, for totally fluid although I have played at least 25 hours so far with just one at 98% totally fluid. Those cards are becoming dirt cheap and are now down to $300 each – your HAF ensures no heat problems. Do you have a crossfire board? Also in Steam, I see that I probably should have spent an extra $4 for some downloadable content in two bonus packs, which increase health and gain face points, without driving all over to find the health shrines.



That screen was originally captured in full 2560x1600. Here's a piece of it at the original resolution:



I know there is a cap at 60. I am going to check to see if there is a cap at 30. I doubt that I hit 30 very much with one card, but if so, the drivers would actually be able to completely turn off the other card from time to time.

When I went all around the 4 boroughs, worshiping briefly at each health shrine, to increase my health by 90% (45 little health shrines - you light a candle and bow a couple times, all done automatically) I came across many areas of great beauty. Here is one:



And so that you can see the original 2560x1600 resolution, here is a little piece of the upper right corner - notice that I am carrying a gun.



But - now that I think about it - that's probably not a good idea for that game - I would be likely to hit big stalls from time to time as a turned-off card was suddenly activated, lol. And furthermore, I am very unlikely to ever hit 30 with a card at stock clocks, where as Sam reported, the average is about 22.

There is a lot of advertising in Hong Kong.



Here's a piece at original 2560x1600 resolution - from the 2.5 gig high-res texture pack.



Probably better would be to use Trixx, disable ULPS, and then set the 30 fps limit, if the game has that setting. That might allow the drivers to throttle back the second card at times to not try to get to the 40 fps range, for energy savings, since 30 minimum to me is totally fluid.

When you motor around one of the islands, which you have to do to get to one particular health shrine, you can see these incredible rocks jutting out of the South China sea, which we see beyond the magazine rack. I read recently that Hong Kong was formed from a super volcano, that erupted, then collapsed into the water. The rocks jutting out are due to that, I suppose.



Here are the magazines from the above picture, at full res. When you actually go up to them close up in-game, you can read the entire front page, lol.

I think we share the same philosophy. I got a huge scare when I first tried to run BC2 and had to take the crossfire apart - I really thought I had ruined both cards just trying that 1025 overclock, or maybe something a little higher that was in the magazine article. Oh, yeah, it was higher, maybe 1075, and I didn't even get in-game. Out on the 2d desktop, all of a sudden there was a checkerboard pattern, and I had to reboot, lol. So I dropped to 1025, and it worked, but for only 3 mins in Heaven.

I mentioned that I got tired of the karate, and resorted to running over people. This wasn't a health shrine, but a suitcase with $5,000, or maybe $2,000 and a gun. I walked over to this parking lot, broke into a car, and ran over the 3 bad guys guarding the suitcase.



It started to rain as I prepared to drive off.



In original resolution, here's one of the bad guys, and on the right is the suitcase I opened.



Also, I have been taking notes all this time in a special overclocking file, and I read last night when I closed up the notes and finalized, and put the file in my special section of technical files for myself, so I don't forget everything in two years - I read that the 1000 clock failed after 20 minutes, crossfire, in Sleeping Dogs. It hung. Otherwise Sleeping Dogs has been totally stable. So, the 1000 proved not to be stable.

There is a lot of variable weather in Sleeping Dogs. Here's kind of a foggy rainy day. The weird outfit he is in gives me +5% more melee damage - sort of an official street thug outfit. One of the little $2 bonus packs lets you go retro, with red hair, goggles like the green lantern, etc. Might be cool! Plus it comes with extra face points which translates to better fighting skills.



And here's some background detail in the original resolution.




So you are 100% correct, the 975 core - not really much of a boost since for $20 extra IceQ comes clocked at 900 - and I doubt they are cherry picking the chips, but maybe they are. I should have just spent the extra $20 is my new thinking. Anyway, Blaze said the 975 gives me 7970 performance, so that was my target, and in all my crysis and every other thing, including BFBC2, 975 is rock solid stable. Also I don't believe I am overvolting. I am not juicing memory up at all - I started to with iTurbo, which by the way is almost identical to Trixx except it allow memory voltage changes - but now I don't run iTurbo anymore after that big scare, just in case IT was part of the problem.

There are all types of girls wherever you go, in all kinds of outfits, from casual dress to professional dress, to geisha.



And here's a bit of a closeup at original resolution.




So like I said, on some of the 7950 models, the reset button of Trixx went to a 1087 vddc voltage, so if that is what some of the cards think they need for stability, I decided to use that voltage for my HIS cards, rather than their reset which is 981. However, running them at stock clocks, I do let them run at the 981.

The low power, which means one card at stock clocks, runs BC2 just fine at 38 fps, maxed, and I can get up to 3 hours of play time before a hang. In the new Spedo case, I now let the card self-manage on fan rpm, since I can clearly hear the fan especially above 70% tachometer, since all of the other fans, 11 fans plus hsf and psu, are a total steady case noise of maybe 25 or 30 db, very quiet compared to what I had before with the two screaming 3000 rpm kazes. Fixed fan speeds will reduce the 7950 temps by at least 5 degrees, but they never run above mid 60 anyway, so pushing them down to high 50s doesn't really seem necessary.

SLEEPING DOGS
I was going to play some more Sleeping Dogs yesterday, and I was going to enable crossfire, but leave everything at stock clocks, but the night grew too late to play. That's my High Power 83 setting. (The 83 is fps in Heaven on the first street after it settles down from a little above that.) That will give me high 30s fps, like what I get right now in BC2 at Low Power 43. From prior running like that, the game will be totally fluid, and in all cases I am talking about the graphics fully maxed out.

The indoor market that you find early in the game, is alive with activity.



Including part of Wei Shen, here's the upper right part in original resolution.




I get almost 30 fps with my Medium Power 51, which is one card at the 975 overclock, and in only a couple of instances, have I ever noticed a slight lag of any kind, and it was only for a minute or so. I have played that game for 6-8 hours at a time with the Medium Power 51 setting - one card at 975 - in all cases graphics is fully maxed out as I mentioned. At the High Power 83 setting, which is crossfire at stock clocks, I hit mid 30s easily of course. And of course that is totally fluid, since for the most part Medium Power 51 also feels totally fluid 98% of the time.

Enabling crossfire, both cards overclocked, my Ultra Power 93 setting, which means using Trixx to disable ULPS, forcing a reboot, I run Sleeping Dogs with fps in the 40s. I have done that before, and it was hard for me to feel a qualitative difference from the Medium Power 51 setting, in the high 20s. I know the eye, for most people, is fooled at 25.5 fps as I recall. I think the concern in gaming, is what happens when there is a lot of explosions and screen action, which tends to slow things down. In that case, as I mentioned, I have spotted Sleeping Dogs lagging at about 22, or maybe 24 fps - I sensed a certain lag and looked up and noticed that fps number in the OSD. But in general, I get the idea that the 1700 + shader engines can handle most of the extra activity, without a major drop in fps. At least that was my experience on crysis, where I ran through the whole game at Medium Power 51, one card overclocked to 975, and never experienced the high fps drop that was supposed to come with the snow, for example. And everything was maxed out - even though you said I didn't need it, I maxed out the AA at 8xAA I recall, and I think I could tell the difference graphically when I did that.

This girl was interesting, as she was leaning against the railing, as you see, but look closely - she was at least a foot away from it, lol.



Here's a closeup with some of the background people, at original 4 megapixel resolution.




So if High Power 83 gives me mid to high 30s fps, never dropping below 30, always to my sensitivity level totally fluid, both cards running at stock 800 clocks, I would just as soon save the electricity and not push my components by overclocking to 975 for Ultra Power 93.

So I think we have a similar philosophy - don't push the components if the game runs fluidly without doing that.

I'm sorry to hear you believe you might be having a problem with your psu.

For a time I was worried about mine as well, and I wondered if that was my part of my new problem, (before I firmly tested out and discovered it was just the BFBC2 game itself, nothing else.)

The toughpower 750 watts is divided out over 4 circuits, so I have always been a little unsure whether I have equally distributed the juice among the two cards. In my individual card testing, one at a time, I put the two pci-e 6-pin lines that came straight out of the modular connections to the power supply, onto the individually tested card. But now, a little more sure that power was never the issue, I have one of those on each card, and for the other 6-pin connection, for each card, I am using a two-molex to one 6-pin adapter, coming from two add-on molex cable harnesses that I recently purchased from Thermaltake to fully populate the psu harnesses. As I recall, I only have to get about 20 amps (at 12 volts) total, so I think I am fine, and so far so good. The pci-e port itself, I believe, supplies 75 watts, and each card, even with the +20% power of the overclocking, if I am using medium or Ultra power, has a total cap of 200 watts, so that's really only 10 amps I need from the total of two cables to each card.

Your power requirements are probably the same as mine I guess, up to 200 watts per card. So I hope you figure out your psu, since you're running a slightly less powerful unit, but at 650 watts with 400 for the two cards, you should still have plenty for the phenom and then some.

Like Sam you'll probably take advantage of the 8000 family, and by then you'll want at least 2 gigs vram. (But by then, the 7950s, with 3 gigs vram, will probably be $200 a card - they're down to $300 now - so that might be a great deal for you, and Sam, at that time.)

Well, talking about fps cap in Sleeping Dogs was so interesting, I turned on Sleeping Dogs just now, and yes, in addition to a 60 fps cap, there is also a 30 fps cap. Hahaha.

So like I said, I am going to try it right now at the Ultra Power 95 setting (I have been saying 93, but I tested just now - it's 95 - just a little stronger than I remembered) with both cards in crossfire, both cards overclocked to 975, and ULPS disabled. This will allow the drivers to throttle back one card to a very low power state, when I am hitting the 30 fps - very interesting that they have that. As I said, they have those two caps, 60, and 30, and a couple under that, 20 and 15, why I don't know. I already set the 30, so I will let gpu-z graph both cards, and report back.

Regarding gpu-z monitoring, however, I believe that I have noticed that disabling ULPS removes the vddc current in reporting, and it also removes the load reporting, as I recall, but I will have the core clock information, and from that I can infer whether the card was under load. I will be comparing the second card with the first, to see if the second has been allowed to throttle back as I play the game for a couple hours.

Interesting!! More science. Plus potential energy savings! A green planet - Sam are you paying attention, haha.

Rich

 

Have not personally had an issue with any readings in GPU-Z with ULPS turned off but I suppose hardware varies. The cards are a generation apart.

I would also recommend leaving the cards at factory stock voltages. To each his own though, all I can give is my opinion.

As far as performance and clocks go, IMO it's better to err on the side of stability, and not performance. Just be careful pushing those cards ;P

 

Are you sure that's not just VSync? That's exactly how VSync normally behaves...

 

Well, Sam, yes, I think it sort of is virtually the same as vSync. But not just a cap at screen refresh of 60 - also a cap at 30, and even 20, and 15. I am totally amazed at all those caps. Why not 40 as well - probably needed to be a discrete divisor of 60. Divide by 2, that's 30, divide by 3 that's 20, divide by 4 that's 15. The only one that's really useful is the 30 I think.

Jeff, that was great advice - about not pushing the cards. I took your advice just now right after noticing that my temps went through the roof. WTF!!

First I set the game for fps cap of 30.

Then I set Trixx 975 overclocking, and verified that it was working by running Heaven for about 30 seconds. I see that it actually gives me a 95 fps number on the first street in Heaven, after settling down from a little higher, so, as I edited in the post above, I started out with my top power, which is Ultra Power 95.

But the first thing I noticed is that my temps got high, very quickly. I took action on that, FAST! Sam you say I shouldn't blame myself, that it wasn't that one furmark test that killed the Power Color card, but nonetheless, I have been very afraid of heat ever since.

It is a hot muggy 85 degrees in my office/trailer and the first card was on its way to 80 degrees, rapidly! Right away, I alt-tabbed to Trixx, leaving the game in its window - it is very stable about alt-tabbing - and I took the fans off auto and set them to fixed, 100%.

The top card was already at 78 degrees by then, and I really prefer to keep it under 70. In addition, I opened the ceiling vent, and I also turned my main trailer desk fan on the other side of the trailer to full speed, and I activated an old 12 volt bus driver type fan bolted to the cabinet that oscillates back and forth, near the gaming rig. It is ancient, and very noisy - with a lot of rattling as it oscillates. Those bus driver fans are only $40 - I should go buy a new quiet one from the RV place. One more thing to do!

I like that 12 volt fan, because it is right next to my equipment, designed to cool this particular table, which has always been the main eating table. Being here next to the gaming rig, it blows on the back of the monitor, as well as on the tower, and I have observed that it noticeably keeps the monitor cooler around the upper vents. I don't want to get out my solder gun and have to ask you, Sam, to tell me exactly which part to pick up and how to take the entire monitor apart. Knock on wood that your 30" keeps running smoothly post the repair, and that mine never stumbles, lol.

So with the fans on the cards going 100% and the additional cooling fans cranking, and noisy, temps dropped to 73. But I still didn't like that, so I took your advice, Jeff, and dropped both cards to stock clocks, down from 975 to 800 core clock (and down from 1087 vddc, to 981 vddc) resulting in an immediate drop of temps by 5 degrees cooler! Yay!

Again, I don't think the 1087 is much of an overvolt, as I explained before it was the reset stock voltage for a few of the other 9750 cards that I had bought and tested, and maybe someday I might be able to prove to myself that the 975 core clock is stable at the 981 vddc voltage. But whether or not it is an overvolt, there is no debate that that the extra voltage adds to the heat from the higher frequency. So with stock clocks and lower vddc, as I say, the hottest card dropped immediately by 5 degrees down to 68 degrees.

That is so much better - 68 degrees on the hottest card is fine! Okay I'm good on temps. (The other is about 57 degrees.) I am just walking my character back and forth in the apartment, and now I'll go downstairs.




The guy in the shop, by the way, shows a talking balloon, which means that I can go up and interact with him, and in this case buy tea, which increases my resistance to damage. I passed right by dozens of times before I learned accidentally about tea. If I were going to a fight, I would for sure want to do that - but it's only $30 and I have about $150,000, so I think I'll buy some tea anyway just for good karma.

Here's some detail with a small piece of the above picture at the full original 2560x1600 resolution:




The guy has been waving his arms, trying to get me to buy his tea. So I'll oblige him.

Dropping from the 975 clocks on both cards helped with the heat. Now, as I said, the hottest card is 68 degrees. The trailer fans are making a lot more noise than anything else, and although it is late in the evening, about 11:00 pm, it's still hot, over 80 degrees, so I'll put on my headphones in a minute and get in about 2 hours of gaming.

When I dropped the clocks to stock, from 975 to 800, the load on each card increased when I did that, from about 68% to about 78% on each card.

That was a surprise, the surprise being the same load on each card.

I guess I have done so little crossfire, I had forgotten that the load was balanced like that. From my prior post, you can probably see that I thought they would run the first card full blast, and maybe really sort of ease back totally on the second if it weren't needed.

So now that I see they run both cards at the same load, a light load if possible like now with an fps cap of 30, or full load of 98 or 99% without an fps cap - now I think I finally understand what you were saying to me, Jeff.

The energy savings comes not from disabling crossfire, but from running the cards at a lower load level, if you have an fps cap, such as vertical synch, which caps them at 60 fps. This is a rare game, that offers fps caps at 60, 30, 20, and 15.

You get your energy savings, from the reduced load on each card. You don't have to disable crossfire - (except in my case with BC2, to get that really incompatible card in slot 6 out of the operation so it doesn't hang the game within 10 mins.)

By the way, the gpu-z sensor readings that I lose on the second card, Jeff, are load, and both memory readings, memory dedicated and memory dynamic. I do get vddc current - but no load. However, on my OSD, which is MSI Afterburner, even though I am not using that to drive the cards (I know I am not, because it doesn't correctly show any vddc at all, and there was a slight incompatibility in the drivers, and I purposely have not upgraded) it does show both loads, and as I say, with 975 they were both around 68 percent, and at 800, they are both around 78% load.

Again, as I mentioned, I was not aware that the load was balanced out like that. Well, I guess that makes sense if crossfire is a tiling situation, or a situation of handling every other frame. I guess both cards have to be running at about the same load, doing about the same work. Yeah, that makes sense.

Also I see they are both running at the full core clock, which is now 800. The game is running at 30 fps, sometimes 30.1, sometimes 29.8. It feels totally fluid. Normally right now I would be seeing fps pushing in the mid to high 30s, so that's about right that I would see the cards at 78 percent load with the 30 fps cap.

I'll play for a few hours, after getting something to eat, and maybe I'll have some interesting findings, although for me I already discovered this hugely interesting balanced load concept.

Well, I didn't get to do any actual playing - but I left the game running with Wei Shen in front of the tea stand, and the loads were maintained while I ate, and while people moved in the street below, and while the tea guy waved his arms from time to time and moved around. I came back and I have a lot of gpu-z sensor recording data and here it is in Notepad - well, not all of it, just 52 seconds worth on each card, picking out a section at each of the two different core clocks, 975 and then about 30 minutes later, at 800.









I went through, tediously, and added up the numbers under vddc current highlighted in yellow, hoping they would turn out to be interesting. I don't exactly know what it gives me, sort of a - what's the opposite of a differential in calculus - oh, the integral - so it sort of gives the overall energy usage of the card, I think, with a snapshot every second. The amperages vary widely while the voltage is fairly constant, so looking at amperage and adding each number, I believe, gives us an idea of the flow of energy through the card.

The grand totals, as you can see, of the column shaded in yellow, are:

Running 975 clock, card #1 total energy flow: 4713.1, card #2 total energy flow: 2563.6
Running 800 clock, card #1 total energy flow: 3736.7, card #2 total energy flow: 2441.6

Let's look at those numbers.

In both cases, I noticed even before adding all the numbers in each column, that card #2 was always consuming less current than card #1. Why - I don't know - maybe being first in line card #1 has extra things to take care of. Running at overclock of 975, card #2 has 54% of the energy flow of card #1. Running at stock clocks, card #2 has 65% of the energy flow of card #1. Wow. Jumps up closer, from 54 to 65%. Hmmmm That's interesting!

Here's my guess on that. At overclock, card #1 is closer to providing the entire 30 fps. At overclock of 975, the cards are both running around 68% load. But because card #1 is pretty close to handling the whole thing, card #2 can coast a bit. It shows that it is running the same load - I don't really know where gpu-z gets that load number and what it actually means, maybe it's a timeslice of when the card shows it's busy - but the energy flow to the card, at least looking at vddc current as we are, indicates only 54% of the energy flow of card #1. So it has the same load, but only 54% of the amperage - to me that means in some sense only 54% of the work compared to card #1.

At 800 stock clocks, both cards seemed to be running around 75-78% load, up from the 68% load of the 975 - they are running at slower clocks and so to keep up the 30fps, they have to be under load for a greater percentage of the time, if that's what percentage load means, if it is based on a time-slicing of busy-ness. Okay, that makes sense they would have to be under load for more time, since they are running slower.

By the way, the 75-78% comes from msi afterburner, and I also saw it on the Afterburner OSD earlier. But with afterburner, I also get a full 40 minutes of graph on sensor logging, including load on both gpus. So when I came back after eating, I scanned that 40 minutes, and read the data at each vertical point on all the graphs. The reason I get so much logging time, is due to a neat trick. I unhooked the sensor reading window from the other settings display window, then I took that sensor window, and made it only 1/4" high, but I stretched it across the entire bottom of my 30" screen, just above the task bar, so it is almost a full 2560 pixels wide. When I open msi afterburner, I get the regular window and the 1/4" high fully wide window - it now opens that way. When I want to see the results, I expand the window vertically and everything is there.

(That's really useful, because you only get about 3 minutes of sensor logging in the current versions of gpu-z, whereas in the older versions, you used to be able to do the same trick, pull the window sideways, and log for a much longer time period, maybe 10 or 20 minutes depending on horizontal pixel count of your monitor. Note that the sensor logging of msi afterburner does indeed include load on each gpu. Gpu-z loses the load on the second card - as well as the memory dedicated, and the memory dynamic.)

So, continuing on with my busy-ness theory, since at 800 clock, both cards have to be busy longer, card #1 doesn't come as close to filling the whole bill as when it was cranking at 975, and so card #2 can't coast as much, and must come up to 65% of the energy usage of card #1, versus only 54% when they were both running at 975.

Those conclusions are pretty weak science, and in fact may be baloney - that's my best guess for now until Sam or Stevo or Russ explains what's really going on.

Okay, now what about total energy usage?

If you add up both cards for 975, you get 7276.7. If you add up both cards for 800, you get 6178.3.

Keep in mind, that at the 975 core clock, memory clock of 1350, I was running a vddc of 1087. The stock clock of 800, had 1250 memory clock, and 981 vddc. So the vddc of the 975 is 0.106 higher, or about 10% increase in voltage. So I suppose you could argue that the 975 energy flow numbers should be increased by 10% higher, since energy is voltage times amps, and the voltage was roughly 10% higher.

For 975, if you apply the difference in vddc, which I apologize, in moving some of the columns over, I overwrote the actual vddc column, but the setting vddc as I just said was 10% higher. So let's just use that - in all cases the actual vddc is always a bit lower than the setting you choose, which sets an upper limit - the actual vddc varies but not really too much. So using 10% higher voltage, 7276.7 x 1.1 = 8004.

Versus the 6178.3, that's 30% more energy usage from the cards in 975 overclock, compared to the cards in 800 stock clocks.

Is that possible?

Well, maybe it is possible. For one, remember that my first card was running a lot hotter. Where did that heat come from? Well of course from the energy being consumed by the card - a certain percentage of which always goes to overcoming resistance and results in energy loss through heat. So yeah, more energy flow, more waste heat.

But still, that's a very interesting number. More energy used, possibly 30% more energy used, resulting in a lot more waste heat, and no benefit at all - just energy thrown out the window heating up the card.

Jeff was so right to tell me to throttle down the cards for the 30 fps cap.

(None of this is very scientific, I realize. It's all just rough approximation. I am only looking at one item, vddc current, which I don't thoroughly understand, I confess, and there are other things that use energy on the card, like data flowing to memory, I presume, because I never see almost 200 watts, which would require amps of around 200, the most I see on amps are an occasional 120, but averages a lot less than that. There was one google article, where a guy was trying to correlate graphics card energy usage from the gpu-z data, but I couldn't quite follow it the other day.)

Anyway, it's 3:00 in the am and I still haven't played any Sleeping Dogs. Haha.

Let's presume that there is something to this analysis. In both cases the second card does not work as hard as the first - okay that's great! We aren't really cranking the second crossfire card needlessly, a lot less energy flows through that card than the first, even though as reported by gpu-z, the load is the same.

Also, it appears that energy consumption by overclocking, in this case with the fps cap, is 30% greater than by running the cards at stock clocks, for absolutely no benefit, and just more waste heat, so for sure, absolutely no reason to do that.

Jeff was correct - run at stock clocks.

But what if I didn't have the game set to the 30 fps cap? Well, of course, without the cap, you will see higher fps numbers with the overclock, so your additional energy usage, and heat, will pay off in smoother play, if you are sensitive to the difference between 30 and 40 fps. With no cap of any kind, crossfire, I have seen mid 30s in Sleeping Dogs, stock 800 core clock, versus mid 40s at 975 overclock.

Heaven shows the same thing. That's why I call the overclocked crossfire my Ultra Power 95, versus my High Power 83, because on the Heaven benchmark, on the first street, after settling down from a slightly higher number, I see those fps numbers - about 95 fps when both cards are overclocked to 975, memory 1350, versus 83 fps with both cards at stock 800 core clock, memory 1250.

Jeff and Sam, you guys rock! Great advice, Jeff! Science - sort of. And now, definitely time for bed.

Rich

 

Hey guys, especially Sam and Jeff, you guys are the crossfire kings.

I went over to the gaming rig just now after finally finishing the post, ready to shut it down. I had left the game for several hours, paused. I took the game out of pause, temps at 61, and moved Wei Shen around and noticed frame rates of about 22. "That's strange, what happened to my 30fps cap?" I thought.

I noticed the first card at 98% load, stock clocks, and the second card at 0%, then 6%. Basically no contribution from the second card. I exited the game, then opened Heaven, and I got the regular High Power 83 bench result for stock clocks. I immediately went back into Sleeping Dogs, went downstairs, saw perfect 30 fps, 78% load both cards, like nothing happened. I ran around a bit and closed up the game.

Remember, this second card is the one that quits within 10 minutes in BC2, and I don't mind letting HIS fix it if some component on the card is shaky.

Is what I just described normal and natural? Basically, the second card went to sleep during the inactivity - I actually had the game paused that whole time, unlike before when I did not have the game paused. Maybe that was the difference. So coming out of the pause, the second card was not contributing. But that was immediately solved by exiting and then reopening the game.

Is this expected behavior - nothing really wrong - or should I contact HIS about it?

Rich

 

Geese Rich! You have to be the king of long posts

Yes, I want a 30" display pretty bad. But there are other things that take precedence over it. A high dollar camera, and a high dollar USB microscope to name a few.

This is the building thread remember. You've turned it into a gaming thread

 

Those temps are hilariously low for video cards. My cards average 75ish in-game without issue. Have never owned a card that played games below 70. In more intensive games like Crysis and whathaveyou, they will touch ~78.

BTW almost done with the side panel mod for my case! Pics to come soon! I now have a side fan

What's more important(at least in my opinion) is to find a set of clocks that work for all situations no matter what you're trying to do. That means you need a cushion of stability, cooling no higher voltages, etc. Take them to what they will do solidy on factory voltages and back them off a hair. I'm not necessarily saying to run at stock clocks Rich. But don't be concerned about pushing the limits. Personally if you can get the cards to 975 safely, it would be wise to back down to say 950 for the sake of longevity. My point is not to be scientific or exacting, but to find a nice, happy, touchy-feely middle ground where good hardware doesn't get fried


Just a few things discovered from experiments:

- If you don't disable ULPS, there can be many issues from bad sensor readings to low performance to interference with clock settings to instability. If staying stock, these issues shouldn't occur, but they do occur when OCing. Your issue sounds like this.

- Not a good idea to mix OCing programs or even have multiple ones installed simultaneously. Some settings are affected just by having a registry entry present. Symptoms can manifest as above.

- As long as your heat never tops 80*C, you are WELL within safety. Those temps are insane! I understand each generation is capable of running cooler to some degree, but that is well below average for high-end cards.

 

Dido!