Intel vs. AMD

I have a bit of a puzzle perhaps someone has an answer for. Last night I decided to see what I could do about my underclocked memory. It's been running at 728MHz with the settings I've been using. I decided to set the multiplier to 2.50, which brings the speed up to 910MHz. It wouldn't run and the MB would scream at me.

I went back into the setup and set the timings at 5-5-5-15. This is what I get in Sandra.



That's a 1000 or so improvement! Here's the real kicker! My other benchmarks stay right about the same as before. All except 3DMark 03. Here's the 3DMark 03 I had recently posted. It was run at 800x600! I couldn't find the post of it at 1024x768, but it was significantly slower.



Now, here it is at 1024x768. Its significantly faster than my previous post.



Both my Math and MultiMedia remained the same, the only thing that's improved is the memory bandwidth, at the expense of Latency. The computer seems much quicker. Even my DVDRB/CCE times have improved by about 2 minutes. Why doesn't the improvement reflect in my other Sandra benchmarks?

Sincerely,
theone

 

Russ,

maybe I'm having a brain freeze but what do you mean by setting the multiplier to 2.50? are those your benchys with the memory at 910mhz and timings of 5-5-5-15? I didn't understand if you got it to run at 910mhz by loosening up the timings to 5-5-5-15. did you try uping the memory voltage at 4-4-4-12? I'm just trying to understand what you did more clearly.

 

Mort81,

In the setup of the GA-965P-DS3 you can set the memory speed by fixed multipliers. For instance 2.00x364=728MHz is where it was at a CPU speed of 3.275GHz. The next setting up that's available is 2.50x364=910MHz. The 364 being the base FSB which is quad pumped to give an actual fsb of 1456MHz.

On the P5N-E, you could un-link the memory and set it independently so that it doesn't change when you change the fsb, sort of. You may have to add a few MHz to the base frequency to arive at the actual desired memory speed! On the GA, it's linked all the time so that it changes with the multiplier set. You can't un-link it, so it's change is constant depending on the multiplier you choose. They only offer you 2.00, 2.50, 3.00, 3.33, 3.50, and 4.00! Since it can't be un-linked you are at the mercy of the multiplier ratio times the base fsb! The Sandra MB posted is at 5-5-5-15. The first 3DMark is at 4-4-4-12 at 728MHz at 800x600. The second 3DMark is at 5-5-5-15 at 910MHz

I haven't tried upping the memory voltage yet as I was looking for some feedback first. It's already set at 2.1v and I have no idea as to what's safe! I'm open to all and any suggestions.

Sincerely,
theone

 

oh ok so instead of giving you different choices for the dram frequency they give you different multipliers which basically would be the same thing. the fsb:dram ratio on my P5W DH Deluxe is not unlinked but they do give you different frequencys to choose from and they change as the fsb changes. I'm actually changing the fsb:dram ratio on my mobo, similiar to the P4P800SE that we both had (I think you had one).

you're basically doing the same thing only it's letting you choose between multipliers as opposed to just giving you frequency choices.

I got ya.

 

Mort81,
That's exactly why I think the 650i chipset may wind up being better than the 965P. If they can ever get it all to work properly, because you can set the fsb and memory speeds independent of each other. There are still minor tweaks needed to arrive at the desired memory frequency, but it's just a few MHz change on the base fsb. I think I had to up the base fsb 5MHz to get Gina's memory to 800MHz.

Yes I did have a P4P800 SE. It didn't have any problems with memory speed mainly because of the much higher starting frequencies of the CPUs we were overclocking at the time. I only raised my CPU speed 700MHz with my 3.0 Prescot, where here with the 4300 I've raised it 1400, and the CPU multiplier was higher to (15x) with the Prescot! You only had to raise the base frequency 46MHz to 246 to hit 3.7GHz. On my 4300 you have to raise the base fsb 164MHz to 364 hit 3.275GHz and that doesn't even bring the memory up to stock speed! I've tried all the CPU multipliers and the only ones that work with my 4300 are x6 and x9. Even the C1E Thermal Management uses the x6 when not under load. I use it because it idles at 2.2GHz but changes instantly under load to 3.275 with no difference in performance or benchmarks.

Unless there's something I'm missing in my understanding of how the settings work, I've run out of answers. I've even considered a 4400 as it has a 10x multiplier and will easily hit 3.6GHz, but has the same fsb limitation as the 4300 mainly because of it's lower 800 MHz fsb. I've seen claims of running them at a 1600 fsb (400x4), but I've yet to see stable benchmarks to back up the claims!

Best Regards,
Russ

 

yeah, I know where you're coming from. I got one of the good E6600's that will run at 400mhz x 9 buss speed. works out perfect for my memory since it is running at 800mhz (2x400) with the fsb:dram ratio at 1:1. too bad you can't use a cpu multiplier of 8. then you could set the fsb at 400mhz giving you a clock speed of 3.2ghz and you could run your memory at 800mhz at 1:1 fsb:dram ratio.

I know when I was playing with OCing my rig, up to a little over 3.2ghz the memory performed better OC'ed but as I went up from there, untill I reached 3.6ghz (400fsb which is where my memory runs at it's rated speed at 1:1), it performed better a little underclocked. even at 900mhz I was still able to maintain timings of 4-4-4-12.

I determined on my rig that the closest I could keep my memory to 800mhz whether it be OC'ed or UC'ed it performed better. example: at 3.24ghz (360fsb) it performed better at 720mhz than it did at 900mhz.

 

Mort81,
Here's both 3DMarks, both at 800x600. the first one is the memory set at 728MHz at 4-4-4-12.



The only change for this one is raising the memory speed to 910MHz and changing the settings to 5-5-5-15!



That's a significant difference! I'm just trying to understand what's making such a huge difference at the same resolution and the same CPU speed!

BTW! What is a safe limit on the memory voltage?

Best Regards,
Russ

 

Don't be bombing my single core preferences lol.

I loved the 4000+. And to my knowledge you all used to be saying how AMD's dual cores and single cores were faster than Intel's at that point. I can agree that in most intensive applications that the D940 will demolish a 4000+. Only thanks to Dual Core. But you have to look at the inescapable speed gap in single core apps which, in this world, are still far more common.

At that point it's like comparing a 640 to the 4000+. Then the win is far in favor of the AMD component. And the A64 X2 was the clear winner over the Pentium D because of the superior speed of it's single core counterparts.

Don't you guys get me wrong. I LOVE the Pentium D line. They were, and still are, wicked fast and very powerful. I think they would have seen more mainstream popularity if Intel had held off a year or so before releasing Core 2 Duo. In fact, I would have been just as happy buying a Pentium D vs the Core 2 Duo. Not as blazing fast but as theonejrs has shown us; They can can still haul ass when it counts.

Theone. Your Pentium D only won out over the 4000+ because of the extra core. Especially in arithmetic benches. I would be surprised if it lost. But if you half that you get around what a 640 would get. And that's what you get in general computing(most games, 3D work, etc.)

That's why I always used to(before C2D) say that AMD was better for gaming and Intel was better for video work.

 

theone,

I can't remember what memory you have and the brand and model is not in your sig.

 

Estuansis,
I was just "funnin" with you. You are absolutely right. Sam has it as only being able to beat out a 3000+, but in reality the performance of an A64 3800+ is slightly slower than half of the speed of a D-940, and it's not too far off the mark of a 4000+ either. Like I said I presently own both although the 4000+ is presently for sale. It's a very nice, very fast single core machine in a MB that's better than average. Uses DDR400, has Sata 3.0, 16x PCIe, 1x PCIe, 2 PCI and 2 IDE connections for 4 drives. It will sell pretty quick for about $450-$500! The 4000+ doesn't overclock very well, but most times the higher end chips don't overclock well. It's way better at games than just about any single core Intel!

Best Regards,
Russ

 

Mort81,

That's what happens when you are half asleep! LOL!! I updated my sig to reflect both the brand (G. Skill) of memory, latency (Cas4) and the settings (5-5-5-15)! Runs very nice at 910MHz. Any ideas as to what is a safe memory voltage. I don't want to burn something out!

Best Regards,
Russ

 

you guys make einstein look foolish.. there is no way in hell icould ever do what you guys are doing. hell i dont even understand ahlf the stuff you all are talking about...lmao. im just gonna buy me a system with the c2d in it and just and some more mem and my liteon burners...


WOW you all do amaze me. its a honor to even be talking in the same forum as all of you guys/gals

if i may quote some one...lol
CLOCK ON

 

cincyrob,

Trust me, the honour is all ours!

I have a question for you! Do you like spending way more money than you have to? Do you like having to compromise on what you really want? Just because a lot of us have technical backgrounds doesn't make us any different than you. We have a lot of folks here that have no backgrounds in electronics at all. We have folks from every walk of life, from all over the world that wouldn't know a capacitor from a diode. None of us knew Jack about computers when we first started out. A lot of us didn't have the benefit of a Forum to turn to for advice or help. We all learned and you can too! Just because it sounds so technical and difficult doesn't mean that it really is. Besides, it's much easier today than it was years ago. Much of the brain work is already done for you. All you really have to know to start is the ability to use simple tools and some common sense. It's even hard to really damage something these days because of all the built in fail-safes. Almost always the computer will shut down if you try to do something that will blow it up.

The rewards and sense of accomplishment that go along with building your own, make it all worth while. A big part of those rewards is getting exactly what you want rather than having to settle for what someone else offers. Come here, ask advice and then buy the parts you want and put it together. Even if you just put it together and just turn it on, it will almost always work right out of the box!

The whole trick is to buy good parts to begin with so that you don't have problems with inferior stuff. One of the problems with a lot of barebones kits is companies put them together mostly out of parts they are trying to unload. Most times they come with inferior power supplys, poor quality memory and so-so video. You want that for your next computer?

I've built computers for over 18 years and didn't know a thing about them until I joined AD a few years back. I just assembled them and sold them! Nice safe stock vanilla machines. I didn't know a thing about overclocking until I joined AD. There's lots of help from a good bunch of folks here, and there's no better time to learn. Just the simple fact that you can usually build a better, faster, higher quality machine for yourself than you can buy, unless you go to Alien or some place like that and have one custom built to your specs, for a lot more money than you would spend building your own.

You said that there is no way in Hell that you could ever do what we guys are doing! Well you're wrong! You can and will learn the things you need to know, but only if you want to. My introduction to this forum came right in the middle of a huge discussion between the merits of the AMD vs the shortcomings of the Intels. I had no clue what they were talking about, even though at that point I had built around 200 computers. Just stop selling yourself short. Help will always be available here when requested. Like I said, it's not really that hard. Besides, if you buy something someone else built and decide you want to increase it's performance, you'll come here for it anyway. Why not start with a fresh page with exactly what you want and be happy when it's all finished. With a pre-built or bare-bones system it will never be exactly what you want or do things with the speed you thought you would get. Doing it yourself means you get to decide what you want and how far you want to go with it, without being foiled by someone else's bad decisions! besides, sooner or later you will do it anyway. What will really hurt the most is how much money you will spend before you get aggravated enough to try. I know, I could own the best top of the line super-killer computer for all the money I've spent just learning that lesson! A lot of us here could!

Clock On,
theone

 

well said, we all know things others dont, and help where and when we can.

back to the thread.

@russ

@600/800 res almost all of the work is done by your cpu and as it is is a memory intensive task the extra bandwidth can feed it faster and aids in your 3dmark score. the extra mem bandwidth will also help improve all other memory intesive tasks such as, games, image rendering and video and audio re-encoding.

as for the mem volts, i have read reports of memory melting when its been run above 2.4v for long periods of time. this was also talked about by nvidia a few weeks back when it claimed it was the case on all chipsets not just their own. i must say that i have only seen reports from people who were using the 680i chipsets and they were occasionally running upto 2.6v. i would say for 24/7 2.3v should be safe, not sure i would run mine on much more.

 

Mort81,

I'll bump it up a notch and just lower the cas to 4 and see if it works. I'll do it right now as it won't take long.

UPDATE: It will boot with the cas set to 4, raising the voltage to 2.1 but the MB setup gives a warning. I ran it and ran Prime95, which crashed before you could release the mouse button. I've set it to 5-5-5-13 and it seems to run well.



Unless someone had a little Voodoo out there, I think that 910 seems to be the limit. Why Sandra reports it as 908 I have no idea. When I went to school 364x2.50 was 910! Must be the "New Math"! LOL!!

Best regards and Thanks,
Russ

 

Russ: Your fill rate tests went down the second time, what on earth?
Cincyrob: Pretty much read what theone wrote, you'd never believe it but in the 3 1/2 years I've been a member of this site I've pretty much learnt everything I know now, with the help of a few 'closer' friends too, if you know what I mean. Literally, in Summer 2004, I upgraded my first component (without paying someone else to do it for me) by replacing a graphics card, and I was petrified about it! Now it's barely something I'd think about, I can't wait to do it again. I still like someone else being there to check my work for re-assurance, after all, silly mistakes, even if they don't cost you any money in damage (they usually don't these days), they can waste your time whilst you try and figure out what you did wrong. Cables left unplugged is the usual culprit!

 

Sam,

No idea! I just decided to run it since I already had the other saved. My 1024x768 did the same thing and was about 4000 higher as well. I thought about running it after I saw the 12,000+ for 1024x768 where it was only 8,000+ or so before. At first I thought it was a mistake. But since it doesn't hurt my other benchmarks and runs better, I'll just leave it at 910MHz and 5-5-5-13. I took the memory voltage up to 2.3v but it crashed the instant I started Prim95, so I put it back to cas5 and 2.1v and it runs better than it did at 728MHz and 4-4-4-12. In general, it seems quicker overall and it boots quicker too!

Clock On,
Russ

 

That I found the case with my P5N-E, when I started messing around with the memory it seemed slower. Besides, the stable frequency my PC boots at changes every couple of days, so I will have to reset and change it periodically. This alone means I can't really be bothered with continually changing my OC to get more, because I have to anyway!

 

Sam

You might want to think about returning it for a refund and getting a GA-965P-DS3 like mine instead. Just set it and forget it! It sure would be nice not to have to muck about with it every couple of days. I don't even want to think where I might be right now if Mort81 hadn't said something! I would probably be bald, and that's a lot of hair to lose! up till that point I was going P5N-E. That's what happens when you start thinking! LOL!! Thank you Mort81! Gina's lost the Seagate hard drive and I'm not thoroughly convinced that the underlying problem isn't the MB. She got an NTFS File Structure error just before it died. Now all it does is re-boot at the same exact spot. I've never heard of the hard drive itself causing that error before.

Clock On,
Russ

 

amen. even the dumbest person on earth knows something I don't. it may not be information I consider valuable but point is they still know something I don't.

Russ,

I don't know why your mobo was giving you a warning but obviously something isn't right. I personally wouldn't run it over 2.2v. try to achieve the tightest timings at the lowaest voltage with everything stable and run benchys to compare.

sounds like your bandwidth is much improved with the memory OC'ed so see how tight you can get the timings there with the least amount of voltage with no errors and stable. if 5-5-5-15 is as tight as you can go so be it. it's still better than it was with the memory divider at 2.0 and timings at 4-4-4-12.

luckily I don't need to mess with all the different scenarios since my fsb is at 400 and therefor my memory is running at spec speed, voltage, and timings. I do know what you're going through though because as I was working my way up to 3.6ghz (400fsb), I went through the same thing.