More Sandy Bridge performance numbers (+Rumoured Pricing)

[mAJORD]

Smarter turbo modes seems to be the theme for both companies moving fwd..

Soon the IPC of the Power Control units will become more important than that of the CPU itself

[JumpingJack]

from what ive seen and heard of that 20% is a best case scenario, NOT average...

You were convinced the best case was only 2-3% ...

[RPGWiZaRD]

Ahhh, it's driving me nuts, I wanna know how these OCs. Exciting times ahead with both SB and BD released.

[PatRaceTin]

interesting

[Hornet331]

You were convinced the best case was only 2-3% ...

Pssst don't tell him you are ruining his negative mood.

[Iconyu]

You were convinced the best case was only 2-3%...

Is there really a worse or best case for it? When you look back at the C2 it's out of order, could do more than four instructions a clock, had crazy branch prediction, and did 128bit SSE3 instructions in a single cycle.

I think back to when the C2 launched and the only thing that looked even remotely bad was the loss of micro-op fusion in 64bit mode (which Intel got running in i7). Other than that it looked devestating and it still does. Intel have stopped making light promises since 2006.

Between sandy bridge and bulldozer I think 2011 is going to be epic.

[JumpingJack]

Is there really a worse or best case for it? When you look back at the C2 it's out of order, could do more than four instructions a clock, had crazy branch prediction, and did 128bit SSE3 instructions in a single cycle.

...

Between sandy bridge and bulldozer I think 2011 is going to be epic.

I was just razzing Saaya a little... I honestly don't care either way. My personal guess is you will see varying results, from 0% improvements to some workloads showing as high as 20 or 25%, with an average somewhere between 10-15%. Saaya has been adamant that at best SB will only boost performance 2-3%.

On your closing thought, I absolutely concur. 2011 will be a fun year to watch the CPU race.

[Drwho?]

Compared to Bulldozer there already is a nice collection of benchmark numbers for Sandy Bridge. For example those posted by Coolaler, a few BOINC benchmark results and a video with a mobile Sandy Bridge running Cinema 4D. The video analysis done in the Planet3DNow forums resulted in a deciphered score of 19641, confirmed by the measured run time (44 s). This means, the tested mobile Sandy Bridge processor was as fast as a Core i7-975 Extreme. Another comparison could be done by using a recently published Geekbench result of a 1.6 GHz Sandy Bridge CPU. So I compared it to a Core i7 also running at 1.6 GHz and made following table with overall results and a diagram showing the differences in detail.



So the average performance increase with those CPUs at the same base clock, but with different Turbo Boost implementations, is about 20%.

...see the rest in dresdenboy's blog: http://citavia.blog.de/2010/08/05/mo...mbers-9128712/

edit: rumoured pricing on some dual/quads

One more non tuned part ...

[terrace215]

One more non tuned part ...

Yes, the memory results did look suspicious.

Questions (if you can say):

1. Will you guys be making some specific performance claims at IDF in September for these first Sandy Bridge parts?

2. What about for the high-end desktop / server variants?

[Drwho?]

Without giving any range of numbers, without giving away anything ...
(I am speaking theory here, not about SandyB itself)
Benchmarks will scale based on the Amdahl's law ( http://en.wikipedia.org/wiki/Amdahl's_law ).
If some sub systems were doubled, and you use a benchmark that uses 99% of the time this subsystem, you ll see a scaling of 2X.
now, there are some specs out there for SandyB that already release publically, not very detailled, but it is there ... you know that it has AVX.
For example, if you look at linpack, it is mostly L1 cache sensitive, and Floating based ... you add on the top of this that AVX is 256 bits, and SSE4.1 is 128 ... You get some indication of the improvement ... I see a lot of speculation here, but I don't see a lot of people trying to do the math properly.
Please try to raise the debate a little, I know there are smart people reading here, I just don't see them posting the math they are asking me questions about by email ...

I am not allow to give away any performance information, just understand that for SandyB, doing the math right will help you more than using a early prototype, those are not tuned properly.

[Mumak]

Doing the math doesn't help much unless they have more information about what improvements are build it - I don't think that all of them were already publicly disclosed yet.. Moreover, the way Turbo works on SNB is also different and it's hard to make any expectations unless that improvement (+few constants τ) has been disclosed too..

[duploxxx]

Without giving any range of numbers, without giving away anything ...
(I am speaking theory here, not about SandyB itself)
Benchmarks will scale based on the Amdahl's law ( http://en.wikipedia.org/wiki/Amdahl's_law ).
If some sub systems were doubled, and you use a benchmark that uses 99% of the time this subsystem, you ll see a scaling of 2X.
now, there are some specs out there for SandyB that already release publically, not very detailled, but it is there ... you know that it has AVX.
For example, if you look at linpack, it is mostly L1 cache sensitive, and Floating based ... you add on the top of this that AVX is 256 bits, and SSE4.1 is 128 ... You get some indication of the improvement ... I see a lot of speculation here, but I don't see a lot of people trying to do the math properly.
Please try to raise the debate a little, I know there are smart people reading here, I just don't see them posting the math they are asking me questions about by email ...

I am not allow to give away any performance information, just understand that for SandyB, doing the math right will help you more than using a early prototype, those are not tuned properly.

looks bad if you have to cover yourself only with general AVX + SSE4.1 chitchat...
where are the magic performance enhancements and 7-8GHZ runs

[xVeinx]

looks bad if you have to cover yourself only with general AVX + SSE4.1 chitchat...
where are the magic performance enhancements and 7-8GHZ runs

Was anyone here promised 7-8Ghz runs? AVX and SSe4.1 will be useful when programs are coded to take advantage of them (not any different than using open64 can help boost opteron performance in some HPC applications). Turbo (and some IPC improvement, likely) helps single-threaded apps, and the architecture is tuned to help multi-threaded (or at least multitasking) performance and scaling. I'm not sure how an all-around performance boost makes "them" look bad. Hardware can move as fast as it wants, but at a certain point, the software has to be coded/compiled to make use of it.

[-Sweeper_]

One more non tuned part ...

nice results for a non tuned part...

[Etihtsarom]

My interest here lies only on SB's lowerend in a HTPC role, can any one comment on it? Think I'll just wait till IVY for the desktop part, hopefully 2011 will hit 4.5 - 5 Ghz on air .

[Movieman]

Without giving any range of numbers, without giving away anything ...
(I am speaking theory here, not about SandyB itself)
Benchmarks will scale based on the Amdahl's law ( http://en.wikipedia.org/wiki/Amdahl's_law ).
If some sub systems were doubled, and you use a benchmark that uses 99% of the time this subsystem, you ll see a scaling of 2X.
now, there are some specs out there for SandyB that already release publically, not very detailled, but it is there ... you know that it has AVX.
For example, if you look at linpack, it is mostly L1 cache sensitive, and Floating based ... you add on the top of this that AVX is 256 bits, and SSE4.1 is 128 ... You get some indication of the improvement ... I see a lot of speculation here, but I don't see a lot of people trying to do the math properly.
Please try to raise the debate a little, I know there are smart people reading here, I just don't see them posting the math they are asking me questions about by email ...

I am not allow to give away any performance information, just understand that for SandyB, doing the math right will help you more than using a early prototype, those are not tuned properly.

Looks in to see what teases Francois is giving us today!

Not tuned huh? What's it need? New points and plugs?

[informal]

Not tuned huh? What's it need? New points and plugs?

New points and plugs,you always crack me up

Anyhow ,Francois thanks for the input in this thread it's appreciated.We all know you can't say much but these mobile parts look very interesting.

[Movieman]

New points and plugs,you always crack me up

Anyhow ,Francois thanks for the input in this thread it's appreciated.We all know you can't say much but these mobile parts look very interesting.

I try,someone has to be the comic relief in these threads.
They get too damned serious and besides, tune up? Plueeeeeese.
How about calling it what it is, the first ones got sent out with a crippled bios..

[-Sweeper_]

[Nicksterr]

These chips won't be overclockable? Intel locking buses? I seem to be finding mixed evidence in my searches. You can buy a dual-core i5-680 right now and get ludicrous clock speeds and performance in all games with a cheap 1156 mobo and 4gigs of ddr3. Guess its time for the numbers waiting game.

[Aberration]

looks bad if you have to cover yourself only with general AVX + SSE4.1 chitchat...

Hardly. He is just covering his ass. What is said here does have ways of reaching people in Intel who make decisions about your employment.

Honestly I bet his is oozing with anticipation to talk about it.

[Movieman]

Hardly. He is just covering his ass. What is said here does have ways of reaching people in Intel who make decisions about your employment.

Honestly I bet his is oozing with anticipation to talk about it.

So lets take him out and get him rip-roaring sheetfaced drunk..
OH, Wait..Have you seen him?
It might take a gallon of JD to get it done.
Guy is NOT small..

[mstp2009]

Just one gallon?

Pft. It's worth it. I volunteer to cover the bar tab.

[tool_462]

Just one gallon?

Pft. It's worth it. I volunteer to cover the bar tab.

I'll split it with you.

NDA = Not Drunk Agreement right? So, if you are drunk and talk, you are still covered?

[Drwho?]

Honestly I bet his is oozing with anticipation to talk about it.

you bet ! ! !