Sign in with
Sign up | Sign in
Your question

Explanation for i7 higher fps using sli/cf?

Tags:
Last response: in CPUs
Share
July 26, 2009 8:09:35 AM

OK, this is a question, and if anyone knows more, speak it here, cause since i7s come out, its been shown using CF/SLI setups, its walks away with fps abilities, even over C2Ds and P2s which may have higher cloks.
I was surfing and found this:
http://www.realworldtech.com/page.cfm?ArticleID=RWT1028...
Now heres what I think. The uops used on C2Ds arent very effective, but the Macros are, and theyre defined as to more singularly based , risc type (uops) vs more complex types macro ops (x86)
But, a funny thing happened in their findings. Prey did poorly under the lessor demanding lower settings on the C2Ds, while when they cranked up the settings, the C2D seemed to excel.
This is very similar to what we see today with i7. Single card setups, i7 sometimes loses to its older weaker brother (C2D ) and P2, but add a card in, and it excels like mad. Im just wondering if the game code becomes complex enough using CF/SLI that it fits into i7s arch in that it reads the code as macro, thus giving it its advantage?
Anyone help with this one? Theories, thoughts and opinions are welcome

PS Maybe using multicard setups somehow triggere this in i7s arch, but C2D is incapable of it, and may have something to do with HT?
July 26, 2009 9:05:38 AM

I am not sure but maybe it is just because the i7s are more powerful, and when you use 2 graphics cards the cpu gets more stressed and that's maybe why the i7s perform better than the c2ds because the c2ds aren't as powerful as the i7s and can't keep on with the graphics cards?? Maybe the c2ds are too weak and causes a bottleneck, I am probably just out bicycling now hehe
m
0
l
July 26, 2009 9:11:12 AM

Intel and M$ have worked hard with their multithreading approach, as to the SW and HW compatibility. If it is read as macros, it would certainly favor i7, and there isnt any reason not to think so. Thing is, given i7s IPC over C2D, it doesnt show up this high ever, unless its SMT, and thats not really IPC anyways
m
0
l
Related resources
a c 317 à CPUs
July 26, 2009 7:02:25 PM

The limiting factor in a game's FPS is sometimes the vga subsystem, and sometimes the cpu. It may not be the same during a benchmark run. As the vga subsystem becomes more powerful, it will be the limiting factor less often, and the stronger cpu will make more of a difference.

As to the i7 vs phenom or core2, the differences in architecture and the raw amount of cpu cache comes into play. If a game is coded with predictable branches in logic, then a cpu with a long pipeline will do best. If a game has tightly packed active code so that much of it can remain within the cpu cache, then it will run better on a cpu with a large enough cache.
Some instructions are implemented more efficiently in one cpu or another, so performance may depend on the preponderance of those instructions being executed. There are many, many other small differences which add up to the totality of performance. That is why we have benchmarks. The trick is to select a benchmark which is identical to what YOUR usage will be.
m
0
l
July 26, 2009 7:21:57 PM

OK, a longer pipeline and more cache dont seem to help i7 or C2D in single card, non gpu limited games. But once you add another card, thats when you see the differences, but usually only on i7.I dont want to lose sight of this , because that is really the main difference here, not the particular benchmark, because if you follow alot of the benchmards, its quite common.
Im trying to understand and to delve into the finer points here, and avoid generalizations, and thats where hopefully I can understand where this widely seen extra performance is coming from, as its really not been explained anywheres Ive seen.
So, if anyone has a link, or further explanation, Id appreciate it.
To be clear, by non limited gpu games, I mean those that dont lose performance when the games maxxed out at a certain resolution, but ocing the cpu brings higher fps, which there are plenty of games out that fit this, usually most of the newer games, games which also show the fps difference in a multi card setup using i7 vs everything else
m
0
l
July 26, 2009 7:27:07 PM

Im along the thought of W7/Vista and i7s arch make the difference together, leaving out the gpu factor, other than having the multicard setup, as being the prime factors of these higher marks, and then to have a further grasp on them as well, hopefully

So, I guess a good question would be, is i7s pipeline longer than C2Ds?
m
0
l
July 26, 2009 9:17:06 PM

interesting you brought this up JDJ. a lot of sites since i7 launch proclaim it king in multi-GPU setups but the real question is why. I dont think it is HT wehn seeing some of the templates used for higher overclocks with HT disabled so there has to be something else. Maybe it is the pipeline or maybe their arch responds well to the IMC interaction or maybe the PCIe lane management of the X58?

m
0
l
a c 317 à CPUs
July 26, 2009 9:29:08 PM

JAYDEEJOHN said:
Im along the thought of W7/Vista and i7s arch make the difference together, leaving out the gpu factor, other than having the multicard setup, as being the prime factors of these higher marks, and then to have a further grasp on them as well, hopefully

So, I guess a good question would be, is i7s pipeline longer than C2Ds?

I suspect that it might be shorter. But, the branch prediction is better, and more work can be done in parallel. The pentium processors had a longer pipeline, and a faster clock speed, but they were slower running code. I think in cpu design, there are lots of trade-offs, and as the designs mature, they work better with the instruction patterns that programs actually use. It is academic to us users, since we can really make no worthwhile decisions based on how a cpu works internally.

m
0
l
!