Nehalem-EP......BLOOMFIELD

[informal]

Is it just me or the BOINC results are a little disappointing,being just on par with C2Q@65nm clock for clock?Or there is more than meets the eye?

[JCornell]

Poor Dave...hahaha

That's not I want too

But, for real, some benches are not really much more differences if compares with Quad Kensfield / Yorkfield ... eg : 3DMark06 CPU Score, or ZD Win32 Floating Point, or CPUMark ...

Afterall, Just for references ...

Thats quite normal for one new product to receive criticism on the first time before final release

[Donnie27]

Hell, all I want is a dual socket gainstown board with 2 quad core Nehalems in it running at 4000mhz and I don't care whose chipset is in it as long as it works!

I heard a rumor about a new HP + AMD for ServerWorks and for Intel LOL! This board looks like a Skulltrail type board for the folks like you. Since these cost more than some of the regular boards sold for servers, they have nothing to loose. Just a rumor though.

If this SW board doesn't see the light of day, you'll not have to worry about what Chipset, there'll only be one. IMHO, we're more than likely to see one by AMD than nVidia. Expensive low volume aren't treated the same as high volume cheaper stuff.

[JCornell]

Is it just me or the BOINC results are a little disappointing,being just on par with C2Q@65nm clock for clock?Or there is more than meets the eye?

Maybe If I post the SS of 3DMark06 CPU Score, or ZD Win32 Floating Point Score, or CPUMark Score now, Could be more disappointing from any of you all

But some benches are really showed the improvement of Nehalem compete with Kentsfield / Yorkfield

[Shintai]

Alot is also affected by the singlechannel memory and maybe bogus board. JC is your board also unable to run above 1 channel?

[informal]

JC,you can post anything you like,after all this is just a ES with a not quite finished platform,so maybe things will improve a bit after it is polished.Post the mark06 and ZD Win32 FP Score, if you can of course .
Thanks for all the input you provided,it's probably the first preview of the Nehalem system on the whole web

[JCornell]

Alot is also affected by the singlechannel memory and maybe bogus board. JC is your board also unable to run above 1 channel?

It's able to run 3 channel, Just simply update the BIOS

You think every X58 mobo & every brands that showed at Computex are ready ? hehe ... Think again ...

[Donnie27]

Maybe If I post the SS of 3DMark06 CPU Score, or ZD Win32 Floating Point Score, or CPUMark Score now, Could be more disappointing from any of you all

But some benches are really showed the improvement of Nehalem compete with Kentsfield / Yorkfield

How about some real world apps instead of so much synthetic stuff? Then how can we compensate for you only having one channel of RAM when these will ship with 3 Channels?

Example, Run Yorkfield or Kentsfield with one stick of RAM for comparison sake? Anyway, thank you very much!

[informal]

Donnie 1 stick with Nehalem is a lot better than having 1 stick of memory with Kenstfield since C2Q needs all it can take from mem. bandwidth perspective while Nehalem has IMC.Don't forget this is DDR3 memory,a lot of bandwidth on disposal.I'd guess we'll for sure see some improvement with 2/3 channels of memory but this design choice was mostly made for servers and a future 4P beasts which will need it very much.Single socket/desktop users would probably see some benefits but not nearly as much as servers(look at AMD's NUMA for ex.).
I'd bet more on a things like early bios/board than mem. configurations (but the last will surely help it some)>

[Shintai]

It's able to run 3 channel, Just simply update the BIOS

You think every X58 mobo & every brands that showed at Computex are ready ? hehe ... Think again ...

Hehe, nah. I know Anandtech also had troubles. One working board for a short time and the rest of teh rest on a faulty one

So whats hold you back from the 3 channel?

[JCornell]

How about some real world apps instead of so much synthetic stuff? Then how can we compensate for you only having one channel of RAM when these will ship with 3 Channels?

Example, Run Yorkfield or Kentsfield with one stick of RAM for comparison sake? Anyway, thank you very much!

Actually, I would like to compare Yorkfield vs Bloomfield vs Gainestown with compressed DVD data ...

I just need time

[JCornell]

Hehe, nah. I know Anandtech also had troubles. One working board for a short time and the rest of teh rest on a faulty one

So whats hold you back from the 3 channel?

Nothin holding me ...

Just made me remembering last year, the first Wolfdale ES need up to 1minutes to complete SuperPI 1M

[Seraphiel]

Far from your expectation...

Aren't the score actually pretty good?

Kentsfield @ 3600 MHz (4 CPU)
3558 Whet / CPU
8289 Dhry / CPU

Bloomfield @ 2933 MHz (8 CPU)
2786 Whet / CPU
7095 Dhry / CPU

Now this is where I might get it all wrong:

(score) x (number of CPU)

Kentsfield:
3558 x 4 Whet = 14232
8289 x 4 Dhry = 33156

Bloomfield:
2786 x 8 Whet = 22288
7095 x 8 Dhry = 56760

Looks to me like Bloomfield / Nehalem is much stronger, or did I get it all wrong? If I have gotten the above wrong, then that WCG benchmark sure is misleading or confusing.

[informal]

BOINC shows total scores(per CPU package),at least that's what i understand.Someone please confirm/deny this.
Now if this is the case,scale down the Kentsfield score to 2.93Ghz and you end up @ pretty much the same score with Nehalem,clock/clock.

I've looked at tomshardware.tw preview of Nehalem scores,and especially at Crysis CPU test.Managed to find a database of C2Q scores for the CPU test 2 at techarp:
http://www.techarp.com/showarticle.a...tno=499&pgno=3

Nehalem @2.93 scores:

Crysis: 1280x1024 noAA
CPU TEST2 18.29

while QX9650 gets around 18.7fps @1280x1024 noAA according to techarp database.Wolfdale e8400 even scores better than Qx9650with ~18.9fps in CPU test2.
Anyone knows what are the scores for QX9650 in CPU test 1 in Crysis bench tool @1280x1024?

[Seraphiel]

BOINC shows total scores(per CPU package),at least that's what i understand.Someone please confirm/deny this.
Now if this is the case,scale down the Kentsfield score to 2.93Ghz and you end up @ pretty much the same score with Nehalem,clock/clock.

I can confirm that the results in the benchmark are per core, not the total CPU or package performance. I get the same with 2 or 4 cores (within 100 pts). So the answer must be clear.

Also, Bloomfield have 4 "real" cores and 4 "HT" ones, or whatever they call it today. So the score must be an avarage of the stronger and weaker cores.

So I think, that if the HT cores were disabled, and only the real ones are left, then the score would actually rise quite a bit, but the number of cores would only be 4.

Just speculating, but I think Dave will actually be happy about the results

while QX9650 gets around 18.7fps @1280x1024 noAA according to techarp database.Wolfdale e8400 even scores better than Qx9650with ~18.9fps in CPU test2.
Anyone knows what are the scores for QX9650 in CPU test 1 in Crysis bench tool @1280x1024?

EDIT: The CPU tests in Crysis aren't very good ones.

From the very little infos about how the CPU tests were done, and with what settings beyond resolution and no AA, then I could have my Nehalem killer setup (all of these benchmarks and infos are true, but not telling the complete picture):

1280 x 1024 noAA
CPU TEST1 162.70 avg fps
CPU TEST2 94.29 avg fps

Point is, there is no sense in comparing the CPU tests with the limited info given by Tom's.

[informal]

Hmm i think that JC was disappointed with BOINC since the scores where lower than Movieman expected.
Anyone else care to post their thoughts on BOINC benchmark results?

[Donnie27]

Actually, I would like to compare Yorkfield vs Bloomfield vs Gainestown with compressed DVD data ...

I just need time

Cool Please take your time and get some rest! Folks tend to screw up more when they're tired.

[NeedMoMegaHurtZ]

JC - Thanks for sharing all this good preview info!

Regarding the BOINC benchmark, it is per core, so for total work done the hyperthreading looks promising

If you assume linear scaling with clock speed (big assumption): 3600/2933 = 1.227x

Nehalem @ 3.6 would be:
Whetstone 3418
Drystone 8715

I think there might a difference in OS as well - 64-bit OS has advantage over 32-bit (I think Movieman's numbers were on 64-bit) - some guys are claiming 5-10% difference but maybe someone can confirm/refute this?

[Donnie27]

I can confirm that the results in the benchmark are per core, not the total CPU or package performance. I get the same with 2 or 4 cores (within 100 pts). So the answer must be clear.

Also, Bloomfield have 4 "real" cores and 4 "HT" ones, or whatever they call it today. So the score must be an avarage of the stronger and weaker cores.

So I think, that if the HT cores were disabled, and only the real ones are left, then the score would actually rise quite a bit, but the number of cores would only be 4.

Just speculating, but I think Dave will actually be happy about the results

True. Virtual Cores will always run slower than Real Cores=P


EDIT: The CPU tests in Crysis aren't very good ones.

From the very little infos about how the CPU tests were done, and with what settings beyond resolution and no AA, then I could have my Nehalem killer setup (all of these benchmarks and infos are true, but not telling the complete picture):
1280 x 1024 noAA
CPU TEST1 162.70 avg fps
CPU TEST2 94.29 avg fps

Point is, there is no sense in comparing the CPU tests with the limited info given by Tom's.

QFT! Folks just need to understand that these are Preliminary only, nothing else.

[JCornell]

JC,you can post anything you like,after all this is just a ES with a not quite finished platform,so maybe things will improve a bit after it is polished.Post the mark06 and ZD Win32 FP Score, if you can of course .
Thanks for all the input you provided,it's probably the first preview of the Nehalem system on the whole web

...

[FlanK3r]

Actually, I would like to compare Yorkfield vs Bloomfield vs Gainestown with compressed DVD data ...

I just need time

yes, im waiitng . THX.

[Hornet331]

...

lol i like how the other program shows 16 logical cores.

[Shintai]

It turns out its a real BOINC monster. We simply forgot SMT. Happy Dave then

[Movieman]

Far from your expectation


FYR
...

First of all Thank you very much for running the benchmark.
Now just so I am sure, this is a single socket quad core correct?
If that is so, then the new Hyperpath(Hyperthreading) is skewing your numbers. If it is possible to shut off that function so that only 4 threads are working you may see some VERY incredible numbers.
Not quite double but maybe 60-75% above what was shown.
If that is true, then you are sitting on essentially the most powerfull machine that exists outside of some supercomputer.
I don't say that lightly, on the old dual single core xeons the difference in benchmarks with HT turned on and off were app 70% greater.
Now add into this that your only on single channel memory on a board that will do tri channel, a pre release board and probably still some bugs and added all up my thought is Dear God Almighty!
Oh, and I forgot to add: You only at 2933mhz..
At the risk of sounding like some schoolkid instead of a jaded 56 year old guy, stop and think what this will do when on a retail fully debugged board with tri channel DDR3-1333 or DDR3-1600 and running in the 3400-3600mhz range.
Just think on that a minute or two, before you call and buy Intel stock!

Aren't the score actually pretty good?

Kentsfield @ 3600 MHz (4 CPU)
3558 Whet / CPU
8289 Dhry / CPU

Bloomfield @ 2933 MHz (8 CPU)
2786 Whet / CPU
7095 Dhry / CPU

Now this is where I might get it all wrong:

(score) x (number of CPU)

Kentsfield:
3558 x 4 Whet = 14232
8289 x 4 Dhry = 33156

Bloomfield:
2786 x 8 Whet = 22288
7095 x 8 Dhry = 56760

Looks to me like Bloomfield / Nehalem is much stronger, or did I get it all wrong? If I have gotten the above wrong, then that WCG benchmark sure is misleading or confusing.

read above, your on the right track..

It turns out its a real BOINC monster. We simply forgot SMT. Happy Dave then

Oh yea, quick, get the nitro pills for my aging heart!

[Seraphiel]

Thanks Dave, and now off to sleep with you, dreaming about retail systems performance for a DP version (DP Beckton especially)