In Theory: How Does Lynnfield's On-Die PCI Express Affect Gaming?

[cangelini]

Thank you Erik--that was exactly the purpose here
For those who want to see CPUs compared to CPUs, I did two more stories reviewing the actual Core i5/Core i7 platforms and getting as many gaming benchmarks run in CrossFire/SLI as possible!

 

[FSC]

[citation][nom]CookYouAll[/nom]Sad, AMD losing again, again & again..[/citation]

Unfortunately so true... I hope AMD could come up with something so good that they could kick Intel right off the throne like they did during the P4-era.

 

[surda]

why did u chose a core i7 975 and clocked it down to 2.8 insted of getting core i7 920?

and u do know that on die PCI controller is actualy on the Uncore of the cpu, and the Uncore of i5 is 2.13, the i7 975 is 2.66 as for the i7 920 is 2.13

so when comparing both core i7 and core i5 "as you hve mention'd clock per clock" then you should of also picked the same uncore speed to get the perfect results for having a "PCI Controller on the CPU insted on the motherboard"

hope this helps.

 

[Alkapwn]

[citation][nom]erikstarcher[/nom]Most(not all) of you people here seem to forget that this test was not a cpu to cpu comparison, so stop worrying about which cpu was used. It was not about cpu speed. It was about 2x8 or 2x16 pci-e performance on different platforms. Don't read the article looking cpu to cpu, look at it from one to two video cards. Then see if that difference is the same percentage from platform to platform. That is what was trying to be shone here.[/citation]
I disagree completely. I believe the intent was to see how an on-die PCIe controller in the processor would affect gaming. In fact, the title of the article is "How Does Lynnfield's On-Die PCI Express Affect Gaming?" The written review did not answer the question to my, and many other's, satisfaction.

The i5's integrated controller means ultra low latency to the GPU(s), but there are other factors which will impact performance. 2x8 vs 2x16 controllers is one. Uncore frequency is another. The i7 Extreme chips run the uncore faster (memory controller, IO ops, etc.), has 2x16 lanes available for multi-GPU configs, etc... so will this make up for the i5's onboard PCIe supposed GPU performance increase?

If you want a true measure of the effects of integrated PCIe on the processor, you would ideally compare chips running the same speeds in core and uncore. Otherwise, it is similar to comparing a 2008 Ford Mustang to a 2009 Ford Mustang... oh, but the 2009 model only has a V6 and burns ethanol. See the point?

 

[jsc]

Another small complaint: The Q9550 also should have been using DDR3 memory in this comparison.

 

[ritsu]

I liked the idea for the comparison. Would have preferred to have also seen single GPU cards used, like 4890. Wouldn't have minded seeing the AMD platform dropped to make room.

 

[unlimitedro]

indeed the article should have made the testing of the on die pcie more clear.

 

[avatar_raq]

Great article. I read it a month ago and read it again today with one question in mind: Will an OCed i5-750 be enough for 2 5870s in my next build. Lets put it in 2 questions :
1- Will overclocking both the core i5 and core i7 (1366) to ~3.8Ghz diminish the difference between 1156 and 1366 in multiple GPUs setup?
2- Will we see the same pattern if 2 5870s are used? Since the performance of single 5870 is close to that of 4870x2, Will 2 GPUs fair better than 4 on P55?
N.B Chris, I know the evergreen GPUs weren't out when this article was written, but I really need a followup! Thanks and Great work.

 

[AZRAELCRUZ]

this reviews are sick in a way of course the nehalem has the edge over the deneb...its the size of the caches 6 vs. 8 of the intel cpu...