Bloomfield 2.93 GHz Performance Data Disclosed

[btarunr]

Benchmarks of Nehalem derivatives are on a roll. We had seen the Bloomfield 2.66 GHz scores and thought it was great. A couple of days ago, Tom's Hardware showed off their newest toys to the world in which was a 2.93 GHz Bloomfield we covered here. Interestingly, they had then stated that Non-Disclosure Agreement (NDA) prevented them from releasing any benchmark data, though following ChipHell's publication, they thought they would disclose theirs as well.

They carried out their tests on the Foxconn Renaissance X58 motherboard with dual-channel Crucial Ballistix 2x 1GB DDR3 1600 MHz, ATI Radeon HD 4850, Windows Vista SP1 and hotfix_vista32-64_dd_ccc_hd4800series_64906 patch. A 750GB Seagate SATA II hard drive was used.

In 3DMark 06, it secured a CPU score of 5183. In PCMark 05, a CPU score of 9583 with a memory score of 9010 was noted. In 3DMark Vantage, the CPU score was 17966 (CPU Test1: 2515.1 Plans/S, Test2: 23.08 Steps/S). 2.93 GHz Bloomfield had a Mere 11% performance advantage over a QX6800 (that clocks at 2.93 GHz). It is also said that this 2.93 GHz chip is 23% faster than a Phenom X4 9950.



View at TechPowerUp Main Site

 

[OzzmanFloyd120]

I don't think it's going to be as great as everyone expects.

 

[btarunr]

Many thanks to HTC for sending this in.

 

[farlex85]

I don't think it's going to be as great as everyone expects.

Core for core speed no, it's no pd to core 2 (seems anyway). It's lookin like intel has put most of their time into multi-threading expertise, w/ HT and QPI, which they seem to be doing quite well at these early stages (seems anyway).

 

[a111087]

kinda expected a bit more...

 

[OzzmanFloyd120]

Wait until tomorrow when the fanboys see it.

 

[Kei]

Good numbers but what's the point of including the Phenom (not you I mean) in that comparison...the processor comes clocked roughly 12.5% higher than the Phenom does anyway? It is quite impressive however if it can achieve that additional 11.5% advantage with the Phenom at the same speed...in all tests that is. I wonder where the 23% came from, this processor for sure will be good though I won't buy one I think the 2.66Ghz model excites me far more for some reason.

When are these slated to hit retail distribution and what kinda price tag will they (speculatively? sorry for spelling) come with?

K

 

[pentastar111]

Meh...I'm still going Phenom with next build...

 

[Rash-Un-Al]

Good numbers but what's the point of including the Phenom (not you I mean) in that comparison...the processor comes clocked roughly 12.5% higher than the Phenom does anyway?

I'm with you on that… although, Intel has little choice but to pit their best against AMD's best… and AMD's best happens to be the Phenom X4 9950.

It is quite impressive however if it can achieve that additional 11.5% advantage with the Phenom at the same speed...in all tests that is.

Considering the X4 9950's operating speed is 2.6 GHz (and not 2.66 GHz), the IPC advantage Bloomfield (at the same speed) has over it is slightly smaller – 10.9%.

When one takes into considering AMD's Shanghai K10.5 45nm processors will boast significantl IPC increases, are said to be able to launch at 2.8 to 3.0 GHz (with faster variants to follow), are expected to operate below current-generation TDP, with 6MB of L3 cache (versus the current 2MB), and a future native octal-core (Sandtiger), Intel simply must release its new architecture, just to stay ahead. Otherwise, current Wolfdale/Yorkfield processors would either be par or not enough for Intel to remain on top.

 

[mk_ln]

11%?...thats odd as i just performed the test with my Q6700 @ 3.3GHz and i got 11750 on the performance setting (3DMark score was P5981). wouldn't it be something along the lines of 50% increase just based on the numbers?

 

[Rash-Un-Al]

11%?...thats odd as i just performed the test with my Q6700 @ 3.3GHz and i got 11750 on the performance setting (3DMark score was P5981). wouldn't it be something along the lines of 50% increase just based on the numbers?

Good observation... as pointed out in the other (ealier) Bloomflield thread, the increases in performance are nowhere near the same for each application... we are currently speaking in terms of average gains, based upon the article's aforementioned "23% faster" claim. I would imagine this is based upon comparisons accross a suite of different applications and scenarios.

Click here, to view the speed increases in SuperPi and Cinebench (single-thread):
http://forums.techpowerup.com/showthread.php?p=877721#post877721
You'll notice that, against a Wolfdale/Yorkfield, the Bloomfield speed advantage can vary considerably, depending on the application. The increases you've described above display Bloomfield's multithreading power at work, in an optimized-for-multithreading environment (3DMark Vantage), while these performance increases still represent the minority of current compute scenarios (and while not all multithreaded applications will yield the same degree of gains).

 

[lemonadesoda]

11%?...thats odd as i just performed the test with my Q6700 @ 3.3GHz and i got 11750 on the performance setting (3DMark score was P5981). wouldn't it be something along the lines of 50% increase just based on the numbers?

Problem is... other tests dont indicate CPU performance increasing like that. It could be a quirk in 3dmark due to HT. In a more realistic test, e.g. CB10, the HT only added a small % to CPU score.

Also look at 3dmark06. A Q6600 at 2.93Ghz scores 4,680 on DDR1. The bloomfied is getting 5183 according to Toms. That's only a 10% increase on the same clocks. Not a very big gain given new CPU, Quickpath and DDR3!

 

[farlex85]

Problem is... other tests dont indicate CPU performance increasing like that. It could be a quirk in 3dmark due to HT. In a more realistic test, e.g. CB10, the HT only added a small % to CPU score.

Also look at 3dmark06. A Q6600 at 2.93Ghz scores 4,680 on DDR1. The bloomfied is getting 5183 according to Toms. That's only a 10% increase on the same clocks. Not a very big gain given new CPU, Quickpath and DDR3!

What makes cinebench more realistic, I'm curious. And what makes vantage less reliable than 06? You seem to be snatching whatever poor increases you can and then ignoring the positive ones. We're seeing a mix b/c this tech is not supported fully utilized yet, on a software and hardware level.

 

[lemonadesoda]

Not really, look at PC Mark 05 also. 4 out of 5 tests (3Dmark05, CB10, PCmark05, Superpi) do not show significant wins with Bloomfield.

"Snatching?", pot calling kettle black.

It's pretty clear that results so far indicate bloomfield is luke warm. That's all we've got to go on. Don't you think it's wishful thinking to ignore that and just blow the bloomfield trumpet? Let's hope things change, but until they do, GO WITH THE FACTS.

 

[farlex85]

Not really, look at PC Mark 05.

4 out of 5 tests (3Dmark05, CB10, PCmark05, Superpi) do not show significant wins with Bloomfield.

"Snatching?", pot calling kettle black.

lol, not quite, just pointing out the inconsistencies of your argument, although I am inclined to believe this is going to be quite a nice line-up once everything is working. I don't know about cinebench, or superpi (but I know at least that one has been around for a while), but I'll take the results from a more recent benchmark over ones from 3 years ago when dealing w/ new tech. Also, I agree that the increase in core for core speed is not impressive at all, but maybe we're spoiled w/ the core 2. 8 logical cores and triple channel ddr3 (which I don't think has been implemented yet) along w/ qpi will yield significant gains even if core for core the proc worked slower, how could it not? It will take a multi-threaded app to show it though.

 

[lemonadesoda]

I really dont know farlex. I "held out" for bloomfield, but now, I actually think a better powerhorse is xeon quad x 2 on skulltrail, e.g. http://geizhals.at/deutschland/a318034.html
There you've got 8 real cores. If you look at the CB10 results, or media encoding results, for such a monster it blows bloomfield. http://www.3dfluff.com/mash/cinebench/top.php

But I guess bloomfield will be a lot cheaper than the workstation/server solution to über-power.