Wednesday, July 9th 2008

Intel Bloomfield 2.66 GHz: First Comprehensive Evaluation

ChipHell carried out the first comprehensive evaluation of the Intel Bloomfield 2.66 GHz processor, a derivative of the eagerly anticipated Nehalem architecture, which already has fan-sites mushrooming all over the internet.

The most prominant benchmarks used by enthusiasts and overclockers, 3DMark Vantage (CPU Tests), Super Pi 1M, Cinebench and SANDRA were run on this processor.

In the 3DMark Vantage test, the processor secured a CPU score of 16294. It crunched Super Pi 1M in 15.475 seconds. With the Cinebench, it secured 3048 with a single thread, the multi-threaded bench belted out 12627 CB-CPU hinting at the processor's high multi-core efficiency. And finally, Bloomfield takes SANDRA out on a date. You have to look at the red dot compared to a QX9770 yourself.

I'm appetised and looking forward to a great processor architecture and so could you.

Source: ChipHell
Add your own comment

59 Comments on Intel Bloomfield 2.66 GHz: First Comprehensive Evaluation

#1
holy_
I want this one :)
Sure it'll overclock well like Wolfsdale
Posted on Reply
#2
ShadowFold
Those mean nothing. Get the games rolling!
Posted on Reply
#3
phanbuey
superpi is a good, simple, crappy benchmark for games... faster superpi=better game performance (usually). But yeah, some game benchies would be great.

the way processors are going these days, it may be better to save up money for a kickass gpu, and just get the budget processor with this architecture and overclock.
Posted on Reply
#4
Megasty
Must get...when is it out...plz let it be out now...plz...plz :D

The Super Pi for the stock processor is great and all & the vantage score kills all comers too. Intel better not charge 2 arms & a leg for this too since its basically eating through the QXs even at stock.
Posted on Reply
#5
EastCoasthandle
This is were having HT capable applications will show the strength of the Bloomfield. Or else it's the P4 all over again. In the sense that the P4 had the potential to perform better if developers support HT.

The only thing I find important with these CPUs isn't the CPU itself but, the plans Intel have to get developers to use HT. The only game that I know of that uses HT is Quake 4. I am sure there are others like Lego but they are hardly CPU demanding.
Posted on Reply
#6
Morgoth
by: phanbuey

the way processors are going these days, it may be better to save up money for a kickass gpu, and just get the budget processor with this architecture and overclock.
This is not a budget processor price ofthis cpu starts at 350 us dollar
Posted on Reply
#7
Darkrealms
It looks like a great work horse CPU. But I'll second (or third, etc) the game benches. I'd definately take it in my work computer!
Posted on Reply
#8
chron
by: ShadowFold
Those mean nothing. Get the games rolling!
i agree. that, and some extreme overclocking
Posted on Reply
#9
niko084
by: ShadowFold
Those mean nothing. Get the games rolling!
They go hand in hang to a point.
Because its faster there and has the same instruction set you will deffinetely be faster, how much is questionable, especially being most games wont use quad cores.
Posted on Reply
#10
cdawall
where the hell are my stars
well thats nice intel has managed to exceed 100% scaling accross 4 cores look at the cinebench test 4.14x processor speed up.....
Posted on Reply
#11
BarbaricSoul
by: Morgoth
This is not a budget processor price ofthis cpu starts at 350 us dollar
So I should be able to upgrade for about $750 then? That's figuring $350 for the cpu, $200 for a good mid-line motherboard like I have now, and $200 for ddr3(hopefully even less by then). Not bad considering the performance gain showed by the benches.
Posted on Reply
#12
Tatty_One
Senior Moderator
by: phanbuey
superpi is a good, simple, crappy benchmark for games... faster superpi=better game performance (usually). But yeah, some game benchies would be great.

the way processors are going these days, it may be better to save up money for a kickass gpu, and just get the budget processor with this architecture and overclock.
Not entirely though, the size of the L2 cache for example isnt too important in gaming but makes quite a difference in superPI, for example an E6850 that I had at 4gig was almost 1.5 secs slower than my E8200 at 4gig :eek:
Posted on Reply
#13
cdawall
where the hell are my stars
by: Tatty_One
Not entirely though, the size of the L2 cache for example isnt too important in gaming but makes quite a difference in superPI, for example an E6850 that I had at 4gig was almost 1.5 secs slower than my E8200 at 4gig :eek:
same example my phenom @2.8ghz with a big ol' L3 cache is as fast as my 5000BE @3.35ghz
Posted on Reply
#14
3xploit
Is this the extreme version or the mainstream version?
Posted on Reply
#15
EastCoasthandle
Lets not forget that Cinebench R10 takes full advantage of HT. When you compare the results to current quad core CPUs the difference can be contributed to the use of HT.
Posted on Reply
#16
DrPepper
The Doctor is in the house
I find the 1 volt very appealing :D
Posted on Reply
#17
lemonadesoda
Those results are AWFUL.

It is ONLY 20% faster than my Q6600@2.66 on DDR1 and AGP.

Intel should be embarassed!
Posted on Reply
#18
EastCoasthandle
by: lemonadesoda
Those results are AWFUL.

It is ONLY 20% faster than my Q6600@2.66 on DDR1 and AGP.

Intel should be embarassed!
Ah, the Q6600 ES @ 2.40 @ 1.176 volts.
This ES CPU is clocked at 2.67GHz @ 1.075 volts, a .101 volt difference. We will know soon enough what the retail version will be like. But when you now plug in that 20% difference (using less power) then we see that Intel is on the right path.
Posted on Reply
#19
cdawall
where the hell are my stars
by: lemonadesoda
Those results are AWFUL.

It is ONLY 20% faster than my Q6600@2.66 on DDR1 and AGP.

Intel should be embarassed!
it did 22% better in single threading than my phenom @2.8ghz and 18% better than my 5000BE @3.2ghz but multithreading...that chip did 41% better than my phenom @2.8ghz :twitch: thats almost two more damn cores.


now this made sense to me maybe not you but i took that score and cut it in half to give a DC version a comparo with my 5000BE @3.2ghz its a lot less of a beat you into the floor difference of 27%

Posted on Reply
#20
farlex85
Lookin good. Looks like these won't be too much faster core for core than penryn maybe, the real advantage will likely be the qpi and the ht. Once all that gets rolling good we'll see I guess, pretty nice preliminary scores though.
Posted on Reply
#21
lemonadesoda
Let me explain why I think the results are VERY VERY disappointing.

1./ Bloomfield = 4 cores and "new" HT = 8 threads
2./ Quick path
3./ DDR3
4./ Bigger cache
5./ Some architectural tweaks other than HT

Yet it is only able to score 4.14x multiplier on CB10 compared to 3.55 on the Q6600. So that means a gain of about 15% due to HT'ing 8 threads over 4 cores. That's not too exciting, because that 15% gain for 4 "extra" hyperthreads is RUBBISH compared to, e.g., a 355% gain for 4 cores instead of 1.

15% vs. 355%? Find me a dual socket Core 2 Quad for the 355%, thanks.

But if 15% gain is coming from HT, and the performance over my Q6600 AGP DDR1 is 20%, the net is 5% for all the other features.

WTF? 5% from:

1./ ...
2./ Quick path
3./ DDR3
4./ Bigger cache
5./ Some architecte tweaks other than HT

If that's really true, those 4 additional "developments" can only win 5%, then Intel has failed with the new architecture and AMD is in with a chance.


*CAVEAT, all commentary based on CB10 and superpi results shown.
Posted on Reply
#22
Rash-Un-Al
Modest

Interesting…

A Wolfdale/Yorkfield must be operating at approximately 3.05 GHz to match that 1M SuperPi time and at approximately 2.80 GHz to match that single-thread Cinebench 10 score (according to tests run personally).

In other words, (again, as I personally tested, moments ago) a Wolfdale/Yorkield @ 2.66 GHz achieves a 1M SuperPi time of approximately 18.274 sec. (versus 15.475 sec.) and a single-thread Cinebench 10 score of approximately 2936 (versus 3048). This suggests Bloomfield is 15.3% and 3.8% faster in SuperPi and single-thread Cinebench 10, respectively.

That doesn't say much. However, it does look like Bloomfield's strength will be how well it scales in multithreaded scenarios. But, in terms of single-thread performance, the increase is evolutionary, and very similar to the gains experienced from K8 to K10 (and certainly not comparable to the 40% or so clock-for-clock average gains seen from Presler to Conroe).
Posted on Reply
#23
farlex85
by: lemonadesoda
Let me explain why I think the results are VERY VERY disappointing.

1./ Bloomfield = 4 cores and "new" HT = 8 threads
2./ Quick path
3./ DDR3
4./ Bigger cache
5./ Some architectural tweaks other than HT

Yet it is only able to score 4.14x multiplier on CB10 compared to 3.55 on the Q6600. So that means a gain of about 15% due to HT'ing 8 threads over 4 cores. That's not too exciting, because that 15% gain for 4 "extra" hyperthreads is RUBBISH compared to, e.g., a 355% gain for 4 cores instead of 1.

15% vs. 355%? Find me a dual socket Core 2 Quad for the 355%, thanks.

But if 15% gain is coming from HT, and the performance over my Q6600 AGP DDR1 is 20%, the net is 5% for all the other features.

WTF? 5% from:

1./ ...
2./ Quick path
3./ DDR3
4./ Bigger cache
5./ Some architecte tweaks other than HT

If that's really true, those 4 additional "developments" can only win 5%, then Intel has failed with the new architecture and AMD is in with a chance.


*CAVEAT, all commentary based on CB10 and superpi results shown.
Both the benches your focusing on are single-threaded. Look at the sandra, there's your feature improvements, and huge ones at that.
Posted on Reply
#24
Morgoth
by: lemonadesoda
Let me explain why I think the results are VERY VERY disappointing.

1./ Bloomfield = 4 cores and "new" HT = 8 threads
2./ Quick path
3./ DDR3
4./ Bigger cache
5./ Some architectural tweaks other than HT

Yet it is only able to score 4.14x multiplier on CB10 compared to 3.55 on the Q6600. So that means a gain of about 15% due to HT'ing 8 threads over 4 cores. That's not too exciting, because that 15% gain for 4 "extra" hyperthreads is RUBBISH compared to, e.g., a 355% gain for 4 cores instead of 1.

15% vs. 355%? Find me a dual socket Core 2 Quad for the 355%, thanks.

But if 15% gain is coming from HT, and the performance over my Q6600 AGP DDR1 is 20%, the net is 5% for all the other features.

WTF? 5% from:

1./ ...
2./ Quick path
3./ DDR3
4./ Bigger cache
5./ Some architecte tweaks other than HT

If that's really true, those 4 additional "developments" can only win 5%, then Intel has failed with the new architecture and AMD is in with a chance.


*CAVEAT, all commentary based on CB10 and superpi results shown.
idiot quikpath and ht and multi cores doest play a role in Super PI
Posted on Reply
#25
tkpenalty
Guys, remember that intel removed ALL limits on overclocking (the FSB wall, etc)... :D
Posted on Reply
Add your own comment