'Nehalem' 2.93 GHz Benches Revealed

[tagasur]

heres the deneb info..

http://www.itocp.com/thread-11103-1-1.html

can you read chinese?

 

[tagasur]

..that last amd deneb45nm super-pi benchmark hit 20.515s, and they got it down further to 19.9 secs.

http://www.itocp.com/redirect.php?tid=11147&goto=lastpost#lastpost

not bad at all, when you pay through the nose for a nehalem and 6 few seconds.

http://www.madshrimps.be/vbulletin/f22/nehalem-bloomfield-2-93ghz-does-superpi-1m-13-969-a-45147/

 

[ZootyGray]

THANK YOU tagasur

I will check those links out.

ALSO:
I did find this link - (maybe same info?) - with CPU-Z pix
- they oclokd /AMD /45nm /DENEB /to 3.4 Ghz.
- it seems stock clock was /2.2, /2.3 (2 engineering samples)

http://www.tcmagazine.com/comments.php?id=20790&catid=2

"And now we wait".

 

[ZootyGray]

Re: "can you read chinese?" ummmm - Chinese: well, No.

I can sometimes catch the drift in French. Je ne parle pas Anglais seulement.

Latin sometimes good for understanding root words, like "Phenom"
Semper ubi sub ubi.

 

[tagasur]

..the bloomfield xtreme's projected sale price is $1000 at the minimum, ref this.

..what i also see in the forums is that the nehalem xtreme QPI=3.2GHZ, while the overclocked deneb's HyperTransport=2.4GHz at most.

let's make a chart:
nehalem xtreme : 3(?)channel DDR3-1333, QPI=3.2GHZ, corespeed=2.9GHz(NO OC)
deneb : DDR3(AM3) or DDR2(AM2+), QPI=2.4GHz, corespeed=3.4GHz(OC'd)/2.3GHz(NO OC)

..i think intel really grabs back the crown at least for the next few years, and does AMD have equivalent technology to intel's hyperthreading? nehalem's 8 core designs each have 2 hyperthreads, thats 8 physical cores + 16 virtual cores.

..maybe AMD's back in it's place for at least a while now (2nd), and the upset it did to intel on Athlon vs. Pentium 4 is just a distant memory.

 

[ZootyGray]

@ tagasur

I am still working thru these links - btw the "lastpost" link goes to page 2 of that forum and the CPU-Z results are on "page 1" of the thread.

What I see (so far) is that:
- there's no PI-result for the standard stock clockspeeds at 2200 and 2300.
- all the oclox from STOCK = 2200 up to 3400 are achieved by only cranking the multiplier and voltage! from 11x to 17x! and 1.224V to 1.568V. With Constant FSB=200 AND HT=1800.
- the final results are achieved by tweaking FSB speed + HTspeed to achieve (near) 3500 Ghz (almost)
Actual = fsb 222 x 15.5 = 3441 Ghz (with HT=1998)(1.568V)
and
Actual = fsb 268 x13.0 = 3484 Ghz (with HT=2412) (1.568V)
(the lastpost result). (Those are interesting numbers with the elevated FSB.)

Test setup:
CPU = AMD Phenom codename/Deneb /cores=4 /1.224 V /6meg L3cache
Mobo = MSI = MS7376 /RD790/SB600 chipset, socket AM2+
(AMIBIOS JAN/08)
Ram = DDR2/800 @ 5,5,5,18,24,2T (note-DDR/1066 not used)
Cooling Solution = Unknown? (no melted icebergs were observed!

(more info? - perhaps if you can read Chinese) (can you?)
(there seems to be some celebration re "Am I the first?" (to oclok to this level?)
If they did this on 'air' that would be amazing.

I need to investigate more of these links.

 

[ZootyGray]

hmmm (see Madshrimps link)

Nehalem at 2.9Ghz with 1.1V does 13 seconds in PI.

That's a huge performance gap.

So, it seems that the best outlook for AMD might still be the "bang-for-the-buck" catagory.

All of this is theoretical because none of these products are available yet. These test reports are all 3rd party. But it is interesting info.

I feel better that we seem to have more info. I still think Thg could have presented a better article to illustrate the "potential" of Nehalem. I must admit, it looks pretty powerful at relatively low clox. More will be revealed in coming months; especially with unbiased, thorough testing, accurate information, and precise reporting.

I am also encouraged that AMD seems to be improving Phenom - more info re Deneb is needed also, and the search goes on. The next few months will be interesting.

This thread has been a real experience for me; so far - is there more?

Thanks for the ride so far, people.

As a self-confessed AMD fanboy for ethical reasons, I have largely ignored ntel - but I am learning more about all that from you ntel fanboyz - thanks - keep up the good work.

The search for my next pc continues - this is becoming quite a journey.

 

[tagasur]

..me too, i just started reading up on uP/cpu technologies and i'm learning quite a bit. in my previous post, i may have misunderstood some concepts.

Intel's Quick Path Interconnect=AMD Direct Connect Architecture(cpu channel with north bridge/devices)

Intel's QPI is only available on the Bloomfield version, NOT on lower end nehalem models. See THIS.

Intel's Integrated Memory Controller (IMC?)=AMD HyperTransport (cpu access to DDRAM memory)

So it seems intel just is about to adopt the Hypertransport/Direct Connect features long present on AMD cpus. No more FSB.

With that said, it seems the nehalem architecture performs way better than the K10/K10.5 arch.

The above forum i linked to is very informative..

 

[tagasur]

i wish thg allowed edits on posts..

amd hypertransport=intel quickpath interconnect

amd directconnect architecture is comprised of hypertransport + integrated memory controller

geez, this stuff gets me confused.

 

[ZootyGray]

Yeh. If it read and kinda meditate on it for hours, it can actually make sense. I indulge to the point of forming a personal impression. After that breadcrumbs of info add more detail to the picture.

I suppose it's possible to master the info - I don't think it's needed. They have names and that says it all. That's my take on the good article you posted - and I bookmarked it for future reference - but I am not sitting the Intel catalogue exam

Get into it once or twice and you will recall enough for future ref. It's endless. I am a generalist until I really want detail. I only learn what is interesting - and that is a simple, natural, experiential process. It's the experience that is remembered.

 

[SonDa5]

That 3dMark06 score looks wrong. Based on those numbers the benchmars score should be 15,388 which is a great benchmark for the HD4850. Also it appears that the HD4850 is set at stock core settings. Some tweaking on the HD4850 cores speeds and this score would be close to 17,000 benchmark for 3dMark06.

Great performance and the huge boost for the 3dMark06 benchmark can be credited to the 5000+ this new CPU benchmarked. Power.

 

[Guest]

OK so you have a fast new processor. Quad cores and all that jazz.

And the first thing you do is benchmark it against a complete different dual core processor which just happnes to have the same clock speed and slap up a salivating post about a 29% performance bump.

Huh???

Huh???

Is it just me or is that a complete non sequitor??? Imagine its 2007 and someones gave me a shiny new 2 Ghz C2D Allendale to bench. And the first thing I did was post gushing comparisons with a 2 Ghz P4 Northwood. Twice the no of cores? Tick. Same clock speed? Tick.

I'm not particularly bothered that late more germane comparisons to Phenoms and Quads were posted. I'm just trying to get round the convolated thought processes of someone who things the original comparison is in any way meaningful.

Frankly, I'm struggling. If anyone can help I'd be much obliged...