Anandtech benches Nehalem

[eleeter]

So let me get this straight. Anand had a spreadsheet error, which skewed the results significantly and made Nehalem looks much better relatively in the graph?

I find it hard to trust the entire set of tests now, that is very very sloppy. What other tests results are wrong? And does Anand have the Nahelem rig permanently in his possession? He was able to rerun some tests right away.

used data from the wrong column (first run data vs. average run data) for Cinebench

Why would the first run be so vastly different?

[villa1n]

Damn, why do I always choose to the worst times to build PCs?
Right when I finished my Dual-Core Athlon rig Core 2 came out.

I was dead set on getting a Q9450, and this has to rear its ugly ass.
I don't think I can wait to Q4. Plus, I'm willing to bet that a Q9450 @ 3.8 can beat this at stock, which is good enough for me...
I wonder if the IMC will be as crippling as it was to A64s OCing.

I'm thinking that a Nehalem + Mobo will cost around $1500 - $1700.

if you have a 9450, your in good shape, as you can see, the single thread performance is almost identical to penryn. So the real advantage will be in heavily multithreaded apps, and it remains to be seen as of yet, how far the OC on nehalem can be pushed, but i m sure its not as far as fsb allows.

[fiveprime]

Yes, most of the programs use 2 or more cores, no really a reason to worry much about it. I mean benchmark programs like SupePI will soon use more than a core but that will be a feature and not a removal of it.

Metroid.

I thought the majority of current programs still use single threads.

[JumpingJack]

I thought the majority of current programs still use single threads.

Actually -- this is not quite true.... it is harder to find bench software that is 'common' that is not multithreaded.

Compression/decompression - winrar multithreaded.
Audio encoding - ripping .. iTunes, multithreaded ... LameMP3 multithreaded.
Video encoding - almost all multithreaded, sony vegas multithreaded, Pinnacle Studio multithreaded
Video editing - paint.net (open) multithreaded, any adobe software multithreaded
Games: Any recently released game, UT3 multithreaded, Crysis multithreaded, Lost Planet multithreaded, World in Conflect multithreaded.

Though not all SW has become multithreaded... superPi is still single threaded

[fiveprime]

True, I kinda forgot about the rendering/encoding/editing apps, as I never work with them. I guess well start seeing alot more multithreaded games now.

[NapalmV5]

until page 6.. video encoding

not bad


nehalem + raid @ video encoding is gonna be

[Dainas]

Actually it may work out the other way around.

If Nvidia does not get the ability to use SLI with these chips then how many of you will go for ATI GPUs so you can max out your setups ?

With the per GPU performance of ATIs gpus currently (and including the HD4000s) things will have to change. As it looks now two GTX 280 could blow away any quad ATI HD4000 setup.

Nehalem with its giant socket looks to me like it will not be top to bottom for quiet a while, giving AMD plenty of time to act.. Margins have just gone up where AMD were already being pummeled.

[Metroid]

In this http://www.anandtech.com/cpuchipsets...px?i=3216&p=11
review Penryn QX9450 scores 3297 points for Cinebench 10 1-CPU. This is exactly faster than Nehalem's 3015 points. :d
Multithread scaling are better than C2D's.

Interesting.

So let me get this straight. Anand had a spreadsheet error, which skewed the results significantly and made Nehalem looks much better relatively in the graph?

I find it hard to trust the entire set of tests now, that is very very sloppy. What other tests results are wrong? And does Anand have the Nahelem rig permanently in his possession? He was able to rerun some tests right away.


Why would the first run be so vastly different?

Limited hardware with many issues.

I thought the majority of current programs still use single threads.

Less used programs or I must say old programs yes, most of the most new used programs use 2 or more cores.

The problem is many old programs have not been updated to support multithreading. Probably they are dead.

Metroid.

[informal]

Do note that on the one single-threaded test Anand ran, a Nehalem alpha-silicon part running on a crippled mobo outperformed an identically clocked Penryn by 26%

I hope you checked by now the- now edited- scores in AT's article.Basically,by "some weird mistake" Penryn was shown in a lot worse light,with very low numbers(we know for a fact that Cine10 and Pov ray are skewed).All other tests can't be reproduced by Penryn users,but you can see the pattern here...
I don't blame Anand,mistakes can happen.But with this in mind,in single threaded scenarios it is basically 1.05x Penryn.In MT tho,it is scaling a lot better now that back end of the chip is widened and it features IMC and SMT(giving ~1.33x-1.5 Penryn in some MT scenarios,which will help a lot in server space)

[therightway]

It's interesting how in Cinebench Nehalem gets a 4.18x speedup maxing out the cores. SMT gives better than 100% scaling! Of course, it's 1 thread to 8 thread, but great use of IPC

[Buckeye]

until page 6.. video encoding

not bad


nehalem + raid @ video encoding is gonna be

Make that a SSD Raid

hmmm... I can think of one I would love to test it out on

[Battle_Rattle]

Seems like the thread has calmed down a bit... before posting... go and actually read the entire thing!

[NapalmV5]

Make that a SSD Raid

hmmm... I can think of one I would love to test it out on

oh yeah ma bad.. yes ssd raid

i know.. your 8x ssd raid lol

[BullGod]

Damn. I was just getting ready to buy a new comp and now this...

[Macadamia]

That's actually incredibly good.Why ? Because , IMO , Penryn has a better cache subsystem :

Penryn - 32KB L1 3 cycles , 6MB L2 15 cycles
Nehalem - 32KB L1 4 cycles , 256KB L2 11 cycles , 8MB L3 39 cycles

Nehalem has 4 cycles L1 , that's a lot , but if probably hidden with SMT and other techniques.The small L2 has very good latency , but at only 256KB it's really tiny.The L3 is very large , but also very slow.

Basically , it's far from optimal for single threaded apps ( Core/Penryn are best there ).

To be honest , I expected Nehalem to be slower than Penryn in single threaded apps that aren't BW dependant.This might still be the case , but it looks like Intel did its job.
Hell , even maintaining Penryn single thread performance coupled with K8/K10 scalability ( in fact even better ) makes Nehalem an excellent all around monster.

K10 actually went forward in "all-roundedness" (mostly multi-threaded but single threaded perf. went UP at least), and all it got was bashing by basically the same people rejoicing now. I'm gonna have to call you (and the others) out on this.

K8 scaled like C2Duo BTW. About the same underpinnings. It's just C2Q that started to show signs of fatigue with 4 cores.

[STaRGaZeR]

I don't blame Anand,mistakes can happen.

Blame him now that you can, Q4 is going to be VERY unpleasant for you AMD people

Another 2006

[Donnie27]

for single threaded?
50% is a lot you know

Let's see all test done on Penryn with one stick of RAM then? Now please come and school me here? The IMC is made up of 3 64bit banks to total 192bit or 64 + 64 + 64 if only a single channel is working, even the playing field and remove one of the sticks from the Penryn system 128bit vs 64bit is hardly fair.

No it's not when you think about IMC, lower latency, then more bandwidth, I do mean processor, system and memory.

[Donnie27]

K10 actually went forward in "all-roundedness" (mostly multi-threaded but single threaded perf. went UP at least), and all it got was bashing by basically the same people rejoicing now. I'm gonna have to call you (and the others) out on this.

K8 scaled like C2Duo BTW. About the same underpinnings. It's just C2Q that started to show signs of fatigue with 4 cores.

it will still go up for Nehalem as well Just like last time, some one here will out Intel and you guys can doubt them all over again as well.

[Movieman]

I read the article, I read this thread.
My answer is when I see a Nehalem that one of the guys here will crank up over 3500 and then watch the Red Sea open up for the second time.


Now to logic: Notice that he said he had a board that wasn't running exactly right and he still got these numbers?
What does that tell you that you'll see on a retail board hmm?

[mstp2009]

i was talking about Q1 and Q2, i dont think nehalem is that ready, but it is shaping up quite nicely though
Maybe some chips are in awesome condition and they are trying to send a strong signal they are ready for it, and AMD is not.
altough the initial results look really good, i just hold off til i see some more meaningfull benchmarks

You are dead wrong.

I know an Intel fab engineer (or 2) and they say these could be pushed out in under 60 days if absolutely necessary (i.e. AMD pulled a rabbit out of it's hat).

They work, they are stable, and they run within the same thermal envelope as current 45nm Penryn Quads.

[Movieman]

You are dead wrong.

I know an Intel fab engineer (or 2) and they say these could be pushed out in under 60 days if absolutely necessary (i.e. AMD pulled a rabbit out of it's hat).

They work, they are stable, and they run within the same thermal envelope as current 45nm Penryn Quads.

QFT

[Donnie27]

I read the article, I read this thread.
My answer is when I see a Nehalem that one of the guys here will crank up over 3500 and then watch the Red Sea open up for the second time.


Now to logic: Notice that he said he had a board that wasn't running exactly right and he still got these numbers?
What does that tell you that you'll see on a retail board hmm?

Coo!!

[Donnie27]

You are dead wrong.

I know an Intel fab engineer (or 2) and they say these could be pushed out in under 60 days if absolutely necessary (i.e. AMD pulled a rabbit out of it's hat).

They work, they are stable, and they run within the same thermal envelope as current 45nm Penryn Quads.

QFT!

The Roadmap shows 2.83GHz Nehalem in Q4 08. It is listed in the 320 to 400 dollar range.

[fiveprime]

So wait, Q4 == July?!?

[Donnie27]

So wait, Q4 == July?!?

July = Q3