Core i7 940 Review Shows SMT and Tri-Channel Memory Let-down

**Blacky** · 10-13-2008, 04:07 AM

Core i7 940 Review Shows SMT and Tri-Channel Memory Let-down
As the computer enthusiast community gears up for Nehalem November, with reports suggesting a series of product launches for both Intel's Core i7 processors and compatible motherboards, Industry observer PC Online.cn have already published an in-depth review of the Core i7 940 2.93 GHz processor. The processor is based on the Bloomfield core, and essentially the Nehalem architecture that has been making news for over an year now. PC Online went right to the heart of the matter, evaluating the 192-bit wide (tri-channel) memory interface, and the advantage of HyperThreading on four physical cores. In the tests, the 2.93 GHz Bloomfield chip was pitted against a Core 2 Extreme QX9770 operating at both its reference speed of 3.20 GHz, and underclocked to 2.93 GHz, so a clock to clock comparison could be brought about.

The evaluation found that the performance increments tri-channel offers over dual-channel memory, in real world applications and games, are just about insignificant. Super Pi Mod 1.4 shows only a fractional lead for tri-channel over dual-channel, and the trend continued with Everest Memory Benchmark. On the brighter side, the integrated memory controller does offer improvements over the previous generation setup, with the northbridge handling memory. Even in games such as Call of Duty 4 and Crysis, tri-channel memory did not shine.

As for the other architectural change, simultaneous multi-threading, that makes its comeback on the desktop scene with the Bloomfield processors offering as many as eight available logical processors for the operating system to talk to, it appears to be a mixed bag, in terms of performance. The architecture did provide massive boosts in WinRAR and Cinebench tests Across tests, enabling SMT brought in performance increments of roughly 10~20% with general benchmarks that included Cinebench, WinRAR, TMPGEnc, and Fritz Chess. With 3DMark Vantage, SMT provided a very significant boost to the scores, with about 25% increments. It didn't do the same, to current generation games such as Call of Duty 4, World in Conflict and Company of Heroes. What's more, the games didn't seem to benefit from Bloomfield in the first place. The QX9770 underclocked at 2.93 GHz, outperformed i7 940, both with and without SMT, in some games.

For more information visit Techpowerup

**Hoboclese** · 10-13-2008, 04:14 AM

I'm not really that good with this stuff but is this because of bloomfield's large L3 cache and smaller L2?

**G.Foyle** · 10-13-2008, 04:37 AM

L3 latency is usually slightly lower on Bloomfield than L2 latency on Yorkfield. Also Bloomfields L3 is a bit smaller than Yorkfields L2. I guess Nehalems cache system will start to shine in 3+ threaded apps, where shared L3 will definitely outperform two L2s connected by FSB and with doubled data.

**Eastcoasthandle** · 10-13-2008, 04:39 AM

P4 Deja vu: The search for the threaded apps.

Here we go again...

**~~bigslappy~~** · 10-13-2008, 04:46 AM

sub'd

**Shintai** · 10-13-2008, 04:54 AM

Still same same old pconline review. Nothing new from the last 17 times its been posted from the same site.

**duploxxx** · 10-13-2008, 04:55 AM

oh now this will become a nice flaming thread... All i have to say about and said in the past for the nehalem threads:

When SMT is added value you will see huge advantage see the winrar/cinebench/tpgenc. Most improvements can be seen in benching and multithreading will improve. (finally 1 reason to switch to intel part if deneb fails to launch @ higher speed and oc then current phenom)

for the rest clock/clock it will be more or less the same to york depending on the apps, some wins some losses. Might get better for nehalem if ddr3 mem settings improves.
nice if you have to buy a new system, no worthy upgrade from york, for sure not with the rather expensive mobo + ddr3.

Originally Posted by Shintai

Still same same old pconline review. Nothing new from the last 17 times its been posted from the same site.

and for sure you as one of the main blue bunnies around here has several comments regarding this review......

**Eastcoasthandle** · 10-13-2008, 04:59 AM

It's odd that multithreading is considered "downlevel".
source

More on how DX11 will effect multithreading here. It will be interesting to see what performance is seen from DX11 specific apps.

**Jacky** · 10-13-2008, 04:59 AM

OP, a mixed bag is not a let-down. It is only a let-down if people believed the substantial hype as to SMT and desktop apps. Quite to the contrary, w/o SMT Nehalem would have very weak IPC overall.

Originally Posted by Eastcoasthandle

P4 Deja vu: The search for the threaded apps.

Here we go again...

You're wrong and you know it I guess? I suppose you phrased your opinion a little provocative on purpose? A. there are lots of threaded apps, but quite a lot only go up to 2 threads B. there are also some apps scaling with n-threads, just not enough on the desktop.

I'm just wondering if I'm the only one reading the TMPGEnc data right? Or am I the only one reading it wrong?

(1) http://diy.pconline.com.cn/cpu/revie...426985_12.html
(2) http://diy.pconline.com.cn/cpu/revie...1438115_8.html
If the 2.93ghz Q9770 produces a higher number than the 3.2ghz Q9770 (1,2) it means that "lower is better" in this test (the graph description is Chinese so we have to guess); thus Nehalem is the worst with SMT on, but when you disable it, then Nehalem leads the pack. (2)
Whereas the article by techpowerup and some other copycat article made it seem as if Nehalem was faster w/ SMT on?

**Eastcoasthandle** · 10-13-2008, 05:05 AM

Originally Posted by Jacky

OP, a mixed bag is not a let-down. It is only a let-down if people believed the substantial hype as to SMT and desktop apps. Quite to the contrary, w/o SMT Nehalem would have very weak IPC overall.

You're wrong and you know it I guess? I suppose you phrased your opinion a little provocative on purpose? A. there are lots of threaded apps, but quite a lot only go up to 2 threads B. there are also some apps scaling with n-threads, just not enough on the desktop.

I'm just wondering if I'm the only one reading the TMPGEnc data right? Or am I the only one reading it wrong?

(1) http://diy.pconline.com.cn/cpu/revie...426985_12.html
(2) http://diy.pconline.com.cn/cpu/revie...1438115_8.html
If the 2.93ghz Q9770 produces a higher number than the 3.2ghz Q9770 (1,2) it means that "lower is better" in this test (the graph description is Chinese so we have to guess); thus Nehalem is the worst with SMT on, but when you disable it, then Nehalem leads the pack. (2)
Whereas the article by techpowerup and some other copycat article made it seem as if Nehalem was faster w/ SMT on?

No, it's not me being wrong vs posting an opinion on a mix bag. A similar mix bag we saw from the P4 back in the day.

**Warboy** · 10-13-2008, 05:31 AM

well, Then if this is reliable then the old Cinebench-R10 benches we've seen are not R10 but R11. Lmao....

**~~Bellisimo~~** · 10-13-2008, 05:34 AM

looks a bit like the results i expected, but well, 1 month to go

**tiro_uspsss** · 10-13-2008, 05:46 AM

& yet again 3D Mark Vantage shows its useless-ness

**Slovnaft** · 10-13-2008, 05:53 AM

ohe noes nehalem floppzingg1>?

idk if i'd trust these benches. i'm not sure why. i'm waiting to draw conclusions though.

**Shintai** · 10-13-2008, 06:02 AM

The first 2 slides posted is loading times if people should be in doubt. Its obvious that techpowerup dont know what it is.

Even in games such as Call of Duty 4 and Crysis, tri-channel memory did not shine.

It says time to enter screen in rough direct translation. And the little sign 秒 is seconds.

Pretty sad when people copy/paste some either halfasses translated chinese through babelfish/google language or simple just guesses.

Fang Chou Pi!

**Slovnaft** · 10-13-2008, 06:21 AM

Originally Posted by Shintai

The first 2 slides posted is loading times if people should be in doubt. Its obvious that techpowerup dont know what it is.

It says time to enter screen in rough direct translation. And the little sign 秒 is seconds.

Pretty sad when people copy/paste some either halfasses translated chinese through babelfish/google language or simple just guesses.

Fang Chou Pi!

And that would be much more affected by hdd speed.

**Shintai** · 10-13-2008, 06:29 AM

Originally Posted by Slovnaft

And that would be much more affected by hdd speed.

Hence trichannel dont offer much. Perhaps if they used an SSD...

Its just overall a very bad thing to measure "trichannel" with.

**NOFX** · 10-13-2008, 06:36 AM

Core i7 CPU -not for a gamers?

**Hornet331** · 10-13-2008, 06:38 AM

lol why does this preview pops up every once a week as news?

how old is this, 1 month?

Sad to see that some news sites dont even look at the date when the original was published...

**Metroid** · 10-13-2008, 06:54 AM

Originally Posted by Hornet331

lol why does this preview pops up every once a week as news?

how old is this, 1 month?

Sad to see that some news sites dont even look at the date when the original was published...

That is the problem, the published date could be 09/10/2008 USA (September) or 09/10/2008 Europe (October). I think this was published 4 days ago. We also had this posted in Nehalem thread, nothing enormous to have its own thread.

I suppose we have to wait for more reviews before we say our conclusions, too little.

Metroid.

**[XC] Synthetickiller** · 10-13-2008, 06:55 AM

News too me, but I've been busy.

This makes me wanna buy a new board and push my q6600 harder. If nehalem turns out to be more of a dud than a stud, well, socket 775 will live on a lot longer for me than I anticipated.

**Shintai** · 10-13-2008, 06:56 AM

Originally Posted by Metroid

That is the problem, the published date could be 09/10/2008 USA (September) or 09/10/2008 Europe (October). I think this was published 4 days ago. We also had this posted in Nehalem thread, nothing enormous to have its own thread.

I suppose we have to wait for more reviews before we say our conclusions, too little.

Metroid.

Its over a month old...september 10th.

**informal** · 10-13-2008, 06:59 AM

In original pconline review there was no tri-channel or SMT comparison IIRC,so this is worthy to be a new thread.

**shiznit93** · 10-13-2008, 07:03 AM

It boils down to this. If a game has a working set large enough to constantly thrash the 256KB L2 on Nehalem, it will most likely run faster on Penryn.