• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Good article on x58 vs p67 GPU scaling

Q

QuiteSufficient

Sufficiently [H]ard
Joined
Mar 2, 2001
Messages
5,152
They announced plans to release this article a while ago, and I personally was waiting for it http://www.tomshardware.com/reviews/p67-gaming-3-way-sli-three-card-crossfire,2910.html
(In my 10 years here Ive never been told it was incorrect to link another site's reviews, admins please do as you see appropriate)

They compare a 2600k at 4.0 with triple 570s, then triple radeons on the p67 platform with the added nvidia NF200 PCIe controller against an i7 920 at 4.0 with the same card setup.

P67 comes out on top, in some tests by a fair margin, despite the PCIe disadvantage, dual vs triple channel memory, and any performance that comes out of an upped base clock on the 920.

This is what I expected, after seeing their article sometime back on GTX 480s only taking a very minor hit when jumping from 16x to 8x to 4x PCIe.

Obviously the SB architecture is superior, but apparently superior enough to overcome the other limitations. I only wish they had used a 2500k, as it offers such a great gaming bargin. They also mention at the end that they could run the SB chip at much higher speeds, perhaps greatly increasing the performance gap.

This raises a lot of questions as to whether or not ivy bridge will open up any performance for gamers, other than purely through clock speed if moving beyond dual channel memory, and opening up the PCIe lanes grant little real world advantages.
 
I don't quite understand what this gauges as we all know SB is faster clock for clock. Should lower SB clock to the point where it matches the 920 in CPU benchmarks and then compare.

It also only measures FPS
 
Flegg that may be a good test, as running multiple screens in a more stressful situation may also be a good test, I think the conclusion is that for people currently in the market to buy a gaming system that may call for multiple GPUs - SB is the better choice despite the reduced PCIe lanes
 
flegg said:
I don't quite understand what this gauges as we all know SB is faster clock for clock.
Click to expand...
It shows that even with only 16 lanes on die plus a bridge, that the performance of the platform beats the next fastest one in gaming, even when spotting the 920 with a larger overclock.

There was so much whining here about how much gaming would suffer with LGA 1155 that it even hilariously involved people stating they would move to even slower AMD platforms because of this "huge flaw." It was silly before and still is. Of course you need to look at the whole platform, not just processor IPC or get alarmed that OMG 16 > 8. Review after review after review has shown PCIe 2.0 x8/x8 vs PCIe 2.0 x16/x16 in gaming is pretty negligible, over several GPU generations.

Then there's the fear that in this new dawn of PC centric gaming (lol) that future GPUs will somehow suffer because obviously the trend has been to cater to the ultra high end users to the exclusion of all lower cards. Yeah, game makers are making sure they use all available slot bandwidth to stream massive amounts of dynamic content in these innovative games. lol again.
 
Forceman said:
Also shows there isn't much to gain by going with the NF200 boards.
Click to expand...
What they didn't test was a three card configuration on P67 without the NF200. If their previous results with that configuration on P55 and their current results with dual card x16/x4 are anything to go by then I'd expect them to be pretty poor, maybe even worse than a dual card configuration.
 
plugwash said:
What they didn't test was a three card configuration on P67 without the NF200. If their previous results with that configuration on P55 and their current results with dual card x16/x4 are anything to go by then I'd expect them to be pretty poor, maybe even worse than a dual card configuration.
Click to expand...

I didn't read the whole article, but the performance summary page has an entry for P67 x8x8x4, which is separate from the P67 NF200 x16x8x8 entry. So that seems to indicate they did test it without the NF200.
 
sorry my mistake, I didn't spot that the nvidia and ATI testing was on seperate pages and the nvidia stuff was first (afaict nvidia won't let you do SLI with a card behind the southbridge).
 
plugwash said:
What they didn't test was a three card configuration on P67 without the NF200.
Click to expand...
The on-die controller in SB only does x16 or x8/x8. Without a bridge, a 3rd slot would have to use available lanes from the PCH. I couldn't find any boards that do that, but some may exist.

If their previous results with that configuration on P55 and their current results with dual card x16/x4 are anything to go by then I'd expect them to be pretty poor, maybe even worse than a dual card configuration.
Click to expand...
Dual card on P55 didn't have to use x16/x4. x8/x8 SLI/CFX configurations were available and much preferred for one big reason. If you did see x16/x4 benchmarks and were disappointed with the results, it is because of how that x4 slot was added. That second slot used PCIe lanes from the P55 PCH, which are version 1.1, 1/2 the speed of PCIe 2.0. Those 4 lanes were severely choked off and poor performance is expected. That is the main problem with blindly demanding x16 over anything else. It's not the best choice sometimes.

The oddball x16/x4 P55 SLI/CFX configuration can't be directly applied to even the P67 PCH because the PCIe lanes on the newer chipset are version 2.0, greatly easing the bottleneck. PCIe 2.0 x8/x8/x4 still isn't as good as PCIe 2.0 x8/x8/x8, but it's not completely terrible like x16/x4 on P55.

Anyways, if a motherboard maker decided to make an triple x16 slot (x8/x8/x4) P67 board, the x4 slot would be twice as fast as the one on the P55 board. Adjust your expectations accordingly. 2 card SLI/CFX x8/x8 would be pretty much as fast as x16/x16. 3 card SLI/CFX would likely show somewhat degraded performance, but not as bad as the same config on P55.

If someone were buying 3 cards for SLI or CFX, I doubt they're going to get a hypothetical x8/x8/x4 board anyways. An exercise in whining is all that is. :p In real usage, 2 card SLI in x8/x8 mode doesn't really degrade real game play performance in any noticeable way.
 
Gasp! SB has higher IPC and clocks than old stuff? :rolleyes:

Too bad its still too marginal of an upgrade to make it worth it for 920 owners short of triple/quad card users.
 
Last edited:
flegg said:
I don't quite understand what this gauges as we all know SB is faster clock for clock. Should lower SB clock to the point where it matches the 920 in CPU benchmarks and then compare.
Click to expand...


That is what it comes down to.

Even with the test results with the configuration the SB p67 doesn't look like it is worth upgrading from a decent x58 build.
 
Last edited:
You must log in or register to reply here.
Back
Top