Conclusion
Today’s analysis began by questioning whether Intel's Lynnfield-based Core i5 and Core i7 processors, along with the P55 platform that complements them, is good enough for CrossFire, and the answer is probably “sometimes.”
Test results showed that a PCIe x8 slot provides 4% less performance when using the fastest possible single GPU, and those results did translate into a 4% performance deficit in CrossFire. That 4% loss isn’t horrific, and less powerful cards would likely show less performance difference. However, these results shouldn’t dissuade anyone from using an LGA 1156 processor to host only one dual-GPU card, since the card’s x16 interface limitations perfectly match those of LGA 1366-based processors.
Less surprising was the horrific performance of the x4 slot available on many P55 motherboards through the chipset’s PCIe hub. These lanes offer only PCIe 1.1-class bandwidth regardless of their PCIe 2.0 labeling, dropping 26% behind even the x4 slot available on similarly-priced X58 motherboards. A slot that slow might be useful for adding a graphics card, but only if that card is used for lower-performance applications, such as a dedicated PhysX processor or 2D desktop expansion.
Perhaps the biggest question answered here is whether we should have forgone the P55 solution in our most recent $2,500 PC build and the answer, of course, is yes. A 4% performance difference might not make or break a real-life build, but it’s a huge obstacle in any competition such as our SBM.
So what’s left? We still haven’t seen how an X58 motherboard with x16/x16/x4 slots compares to today’s x16/x8/x8 sample, whether or not a PCIe hub such as the NF200 can help the P55 to cope with additional cards, or how badly a P55 motherboard in x8/x8 mode can hurt a pair of dual-GPU cards, such as the Radeon HD 5970. We don’t have the parts in place to test all those variables, but will continue to consider your suggestions while we assemble components to put at least some of these questions to rest.
Latest CPU News
Latest CPU reviews
- 24/05 – Overclocking Core i7-3770K: Learning To Live With Compromise
- 22/05 – Core i5-3570K, -3550, -3550S, And -3570T: Ivy Bridge Efficiency
- 15/05 – AMD A10-4600M Review: Mobile Trinity Gets Tested
- 14/05 – Core i7-3720QM: Ivy Bridge Makes Its Mark On Mobility
- 23/04 – Intel Core i7-3770K Review: A Small Step Up From Sandy Bridge
Interesting, and it matches the results found elsewhere on the subject.
With the i5 seeming to be a popular mid-upper bracket choice, why not pair it off with similar placed cards like the HD5770/HD5850?
I would be very interested to see how these less potent cards perform with the popular x16/x4 or x8/x8 'boards out there rather than see a HD5970 crippled by an x8 'board-a configuration that seems pointless in the extreme.
So if there is not much difference between x16 and x8,it doesn't really matter if a P55 system does't support 16x/16x eg.only performs 16x/8x-8x/8x or whatever?
i found a similar report here http://www.hardware-revolution.com [...] e-problem/
But why bother about x16/x8 there are true high-end chipsets from AMD with full CF at lower price?
In a previous look at PCIe 2.0, Toms Hardware found variation in the impact of bandwidth reduction according to the application running. See this http://www.tomshardware.com/review [...] 915-9.html
A particularly difficult program was Microsoft's FSX flight simulator (scroll down on the linked page). Well you might say, FSX doesn't benefit from SLI or Crossfire, you'll be using a single graphics card, so why are you bothered?
Well unfortunately a lot of people might be doing things like adding a USB 3.0 adapter card soon. That only requires PCIe 2.0 X1, BUT if you stick it in the other PCIe 16x slot it'll cut the graphics card down to PCIe 8x with P55.
Or you could put the high bandwidth adapter in a PCIe x4 slot. However, in Intel's wisdom they created the P55 chipset with 2.5 GT/s PCIe generation 1.1. Even if you were happy with that (it'll limit a USB 3.0 adapter) I'm still stuck because I want a Micro ATX build. All Micro ATX P55 motherboards have the PCIe x16 slot positioned so that a dual slot graphics cooler will obstruct the PCIe x4 slot. Great.
A useless comparison since the whole point of getting an X58 board over P55 is that it can do 16x/16x whilst a P55 can only ever get as good as 16x/8x so a real test of Crossfire performance would have been a better X58 board that is up to the Intel spec rather than one hobbled to a P55 spec. You might as well be testing two P55 boards.
What a pointless article.
Just when I thought it was safe to come back here... the smacktardian smell of payola wafts through the door as Intel and nVidia give Toms little brown envelopes for stating the blindingly obvious. Yet. AGAIN.
So, from this test we draw the conclusion that P55 is EVIL and ATI are EVILLER!!! No, Toms, we're not as jack-**** retarded as Intel thinks we are, and will not rush out to buy X58-based systems and GTX295s to put on them just to avoid the EVIL of bandwidth bottlenecking. Your test is risible; who in HELL has the money to use three €400 graphics cards in CF these days?! I'll tell you who: someone who wouldn't buy a P55 system even if Hell froze over just for their pleasure. Where the hell are the HD5750/5770/5850 2-card comparisons?! The above article is just utterly pointless - of course three cards that could eat a HD4870X2 or GTX295 are going to overwhelm a P55 if you jam them all on together! That's just common sense!