Video Card Thread (Mostly Gecube x1950xt)

Rich,

I think if you are planning on doing the Asrock route would be the only way you'd want to buy a Gecube, as pairing it up with a C2D is the only way you'll be able to unleash it's full potential (unless of course you had an AMD 64 4000+ or the like, which of course you don't). If you do that then you'll want the new CPU asap. If you're not going that way then just get a X1900 pro agp or a Geforce 7950 agp (as that's even less power hungry card), keep your old psu, your old components will be much more closely matched and then save all that wonga for your big-guns pc next time round. That (the latter) would be my choice, you save money and hassle. You just have to live with a bit less detail in game untill you splash out on the new 2900 crossfire quad core mhz muncher!

Back to Dirt, some of you mentioned it a few posts ago, that was pretty much my main reason for upgrading my cpu/ram. On my p4 rig fps would be in the low 10s, 13-14 fps with low detail levels and really unplayable.
Now I can enjoy 30-40fps at all medium settings 1280x1024 with 4x AA. BUT, RallyCross (mainly) and also some RallyRaid type races are still a nightmare. These are the types with a dense circuit and 7-9 opponents and boy does my pc struggle. And I think Codemasters is to blame here, as discussed with Sam earlier on this thread. Written for the 360 primarily (which has a 5 core cpu?) I think what it must try to do is load 1-2 driver models and their associated physics etc onto each core, which means my C2D laps up the rally style and 1 vs 1 races but it really craps out when you have 8 chunky car models to whizz round a track. Apparantly the PS3 version is more playable than the 360 (which also brought complaints about framerate), presumably due to the extra power of the 8 or 9 cores the PS3 chip has. Still Codemasters should have done a better job of optimising this game, and porting it to the pc. Take Rockstar, they spend 6 months on every pc release of GTA to make sure it actually runs better on a pc than any console. I've been annoyed with Codemasters ever since CMR05 when the high end Geforce 7000 series cards had an in-game glitch. They tried to fob everyone off with a 'no-sun' patch that turned off the sun effect but didn't sort the glitch.
Basically with this new game, if you want to play it seemlessly, in high, in all race modes I suspect you'll need a quad core and a 8800 / 2900.
Sam/Abuzar, if either of you get this game I'd be interested in your findings with your much higher cpu clock speeds.

 

I just got my gecube x1950xt and the girl on the front cover is bangin' lol

 

Abuzar: Depends how much upload you have. Just leave it at half and keep your

torrents running overnight, that gets them done nice and quick.
personal message will be on its way soon.
As for dirt, I don't know how good that is on average, but getting a frame rate

that mediocre at that resolution worries me, considering how powerful your PC is.

if it's any consolation about the RAM cooling, I need to get some Thermal epoxy

for my RAMsinks because they keep falling off. With half of them missing I can

overclock just as far and it still be stable. I think GPU RAM cooling's

over-rated!


Harvrdguy: With the power supply, I honestly couldn't say. If you leave them

stock, Core 2 Duos draw a lot less than P4s, perhaps around 7A on the 12V, maybe a

touch more. If, say the X1950Pro draws 8A (that sounds about right, but again, it

might be a bit more, I don't know the actual wattage) then you're looking at 15A

for the CPU and GPU as a bare minimum. Add one hard drive and CD-ROM and that's

about 18A. I don't know whether much else draws 12V in your system, but that seems

like it's cutting it a bit fine to me. if I were you I'd get a new PSU, whatever

you do, and if necessary, save up. otherwise, you risk spending a hell of a lot more than $500.

I'll be straight up with you on the PCIe debate. PCIe 4x is not great. HOWEVER, test showed that running an X1900XT-X (pretty much identical to my card) at PCIe 4x, only dropped the performance by about 20%. From what I can see, that may not actually be enough to drop the performance level to be equal. If the X1950XT PCIe is as fast or faster than my card, even at 4x, it may still beat the AGP model.




Faugyou: LOL! You mean Ruby? She's ace, huh? Leagues above the various gremlins and demons that adorn nvidia cards!

 

I think we need a more standardized test. I think my 1 GB of ram might be affecting my framerates. I'm talking about the first level in the "campaign" mode of Dirt. What level are you guys talking about? I'll install the demo again and try the first level on the official(not omega) drivers and see what happens.

Does having fraps running effect performance?

 

Abuzar, do you have the full game or the demo?

 

I only have the demo. I don't really like racing games, unless they are on a console, because then I can do multiplayer.

 

I'll grab DIRT and see how I fare.
As for fraps, it doesn't affect performance, provided you're not recording video. If all you're using is the frame rate counter, it doesn't affect the game.

 

At abuzar, I cool the ram on the card with these: http://www.newegg.com/Product/Product.aspx?Item=N82E16835116019

They're just regular bronze heatsinks that work pretty well. Some people complain that they fall off, but I've had none of that yet and they've been on their since the first day i got the tidewater, which was a while back. You probably could do better with something like this: http://www.newegg.com/Product/Product.aspx?Item=N82E16835119061

Although they're smaller, the have more ridges which probably helps the air pass through them more. My Antec 900 has a turbine fan sitting in the window which cools the gecube, so the heatsinks stay pretty cool.

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

Ok, now for the 3dmark06 overclocking thing. I tried it last night, at 720/855 and got worse than what I got with stock speeds! I think I got like 3800 or something like that, so PLEASE DONT ADD THAT TO THE GRAPH! Thats not my final score. I think I overclocked the gpu a little to much to the point where it underperforms... Will try again tonight...

Btw, is it true that over clocking can make your card work worse with higher clocks? I think I went over the limit, although it didnt freeze or anything like that.

Also, ATi Tool will only let me go to 715 if I'm lucky on the GPU clock. Why is this? I can get above that with ATi Tray Tools, which is what I used to get the 720 speed. I can get to 740 I believe, and then the screen freezes. But, if I play a game a 736, it will freeze also, so I guess I need to find the max stable limit for the gpu.

 

ATI probably uses a more stressful test render, so it may cause the card to crash earlier. That's good, as it means you're not totally stable at anything above what you can get with ATItool. I've not heard of the score dropping that much with an overclock, unless there was an overheat problem, which sounds unlikely.

 

Ray, I won't change the graph!

Sam, so you're saying a 20% performance hit with 4x pci-e. That's good information. But you lost me on the pci-e version of the gecube that would be faster than the agp version. I believe the agp gecube puts out about 10000 fill rate - I'm not aware of a pci-e card around $200 that is much faster than that - maybe 10400 from the charts I have seen. By the way, you said there was a fill rate test on 3dmark6 - when you run it Sam, what fill rate do you get?

Mike, I hear you on the cpu bound. So maybe I'll be a little less lazy and go for option 1 and get the gecube and the new psu. Then like Faugyou the little ruby chick will be bangin, lol.

That should give me 4400 3dmark6 like Ray and Travis. Then, like Travis, when I want to start getting into the 5500 range, like Abuzar and you, (I started to bitch about the pci-e 4x speed rob then I realized I'd still be on the 8x agp!) I can pop the Asrock and core 2 duo. Yeah one way or another it will be nice to come off the bottom of the graph (and play MOA-Airborne.) -rich

 

OK, running fill test now. Results soon.

UPDATE: Okay, results:

Single texture fill rate: 5651.6 Megatexels
Multi texture fill rate: 10,662.8 Megatexels

Tests run at 1280x1024 standard settings in 3dmark06.

 

Sam, thanks, that's what I thought - 10,600 is roughly where I thought you'd be on fill rate. Hey can we get one of you gecube guys to run the 3dmark6 fill rate test like Sam just did - Mike, Ray or Travis? (I'm thinking the gecube is also around 10,000.) - rich

 

Hey Ray and Abuzar,

I have those thermaltake copper vram sinks as on Ray's first link. What is the best way to attach them with so they don't fall off like Sam is experiencing? I have AS5, and I also have Thermattach 412 - thermal two-sided tape. I didn't think Arctic Silver 5 would form a bond that would hold something, but I read something a while ago that indicates that it does harden up. Is that true? Or is the thermal tape a stronger way to go?

And finally, what about the tape that comes already attached to the little sinks? Should I just use that, or should I scrape that off and do AS5, or thermattach? What do you think, Sam, about those 3 options? Anybody else?

 

Just use the tape that's already on it. If you want better conductivity of heat, then use thermal epoxy, thermal grease wont cut it as it will fall off before it gets a chance to harden.

 

Also update my score. It went up a LITTLE bit when I overclocked my CPU to 3.52 Ghz.

Main Test Results
3DMark Score 5486 3DMarks
SM 2.0 Score 1999 Marks
SM 3.0 Score 2180 Marks
CPU Score 3068 Marks

EDIT: If any of you use DDR2 RAM, this is one of the best stuff out there. You can't beat it for 60 bucks.
http://www.newegg.com/Product/Product.aspx?Item=N82E16820146565

 

Ok, at the default clocks for the gecube, I got 4410.207 on the fill rate single textures. On the multiple textures, I got 103.22.

This is kinda weird, but usually when I run the "show 3d view" in ATi Tool, the fps will fps around 580, but now I get 147fps. What the hell is going on? I think I will try to reinstall the drivers, but I dont think that's whats causing it...

Any ideas?

EDIT: Nevermind, fixed it by uninstalling and reinstalling atitool.

 

Ok Abuzar, your score is updated and I added your Ram - page 35.

 

So this is how it stands.

MaccerM Mike
Kardson Travis
Abuzar1 Abuzar
Sammorris Sam
Waymon3x6 Ray
Harvrdguy Rich

Sam................... 6755 ........................ E4300-3ghz 2gig.................... x1900xt 512mb
Mike................... 5982......(5600)...........E6320 2.11ghz 2gig................ gecube
Abuzar............... 5486 ......................... E6400-3.52ghz..1gig............... x1900gt 256mb
Travis................ 4499 ..........................P4 3.0 2gigDDR400 .................gecube
Ray................... 4422.......................... P4 3.2 oc'd to 3.75.................. gecube
Rich ................. 2260 ......................... P4 3.2 1gigDDR400 ................ x850xtpe 256mb

 

Everything you want is in my sig... I'm about to run 3dmark06 soon (tonight) so stay tuned.

 

Ray, thanks. That's a gecube fill rate of 10,300 - close to Sam's rate of 10,600. That's what I have been thinking - the gecube is right up there with Sam's card, and 25% more powerful in raw fill than my x850xtpe, besides all the extra shader stuff.

Hey Abuzar, I'm thinking your $85 GT delivers around a 6900 fill, which is outstanding considering you didn't pay anything for it. Do you wanna run the fill test?

Travis and Mike, you guys feel like confirming Ray's results with your gecubes?