Anand Sandy Bridge performance preview is up

[Nemesis 1]

Enchantments, somebody's been playing too much RPG's. 🙂

Are you suggesting delay on 16nm? Can you not do that if you are?


GT1 and GT2, interesting...
http://www.anandtech.com/show/3885/sandy-bridge-graphics-update

Actually I am not suggesting it at all . Sandy Bridge 32. 1 year as 32nm. Ivy bridge 2 years Haswell 2 years and still on 22nm . Isn't that what intels roadmaps say . As there not complete . Didn't intel say 3 generations of bridges. haswell done @22nm. Thats as I understood it.

 

[IntelUser2000]

The 22nm roadmap seems changed now. They look like they are not only on their way to deliver Haswell at 2013, but even Rockwell at 2014.

 

[Nemesis 1]

It would be nice if intel doesn't have to use 3 bridges and haswell is ontime. But those xtors are getting small. crosses fingers.

 

[ilkhan]

The 22nm roadmap seems changed now. They look like they are not only on their way to deliver Haswell at 2013, but even Rockwell at 2014.

When did we get haswell's shrink name?

 

[bryanW1995]

The 22nm roadmap seems changed now. They look like they are not only on their way to deliver Haswell at 2013, but even Rockwell at 2014.

that's just the tick/tock. I wasn't aware that they were planning two years for ivy birdge and haswell.

 

[Vesku]

Looks like the Anand review sample had the highest tiered integrated graphics. Still going to be a great marketplace for cheap laptops next year.

 

[ydnas7]

Looks like the Anand review sample had the highest tiered integrated graphics. Still going to be a great marketplace for cheap laptops next year.

it may be 12EU units but the drivers would be very early, the turbo turned off and the games would not have been structured for Sandybridge graphics. Not optimizing for the transcode unit, the AVX, the L3 cache and the fast CPU cores is a bit like ignoring the Cell SPEs in a PS3, yes it will still work, but its like giving a solder a gun with a grenade launcher and the sending him out with bullets but no grenades...

i would expect significant performance difference between games that are profiled/optimized for sandybridge and games that are not. so basically all current games won't be particularly good.

 

[ydnas7]

Been a while since anything got me interested, nehalem did it, westmere was expected to but basically fell flat to me. Sandy Bidge, however...yum.

Since I love charts, I did some work with the ones from the article. Prices are from yesterday on newegg, like 10 minutes after the article went live. I suspect the i7-950 is supposed to be the i7-930, significantly cheaper and sitting in about the same spot as the rest of that box.

i like the chart, and i think you basically nailed it except for the 2600K price which is more likely to be around the $295 mark (Core i7-950 after price drop)

 

[PlasmaBomb]

That would be worth it, if the 44% wasnt in winrar and video encoding. For the things i'd use it for, id be looking at the low end of that % figure.

This would require a new overpriced intel mobo, $~325 CPU, and at least 8GB of DDR3...

And i don't even get a hex core at the price because intel decided to use up their real estate on the graphics chip that will be disabled by any end-user that plays video games. Oh wait, i think it can play the original unreal tournament at low settings.

Actually your performance benefit would range from 35.4% (SB with no IPC improvement over Nehalem) to 69.3% (SB +25% IPC) with video encoding coming in at 75.3% faster or more...

So what do you do? Sure it's not a great improvement for browsing, but choosing a more likely +10% IPC you are looking at ~49% better performance at 4.4 GHz than your current 3.82 GHz rig...

Granted there are questions that will be answered before you have to make your decision (what the real IPC improvement is and how well SB will clock), then you can make your decisions based on that.

Hadn't realised that you were running a 65nm Core2...

 

[IntelUser2000]

When did we get haswell's shrink name?

That was actually a while ago. Check SemiAccurate.

Not optimizing for the transcode unit, the AVX, the L3 cache and the fast CPU cores is a bit like ignoring the Cell SPEs in a PS3.

@ydnas7 Erm, may I ask how a transcode unit would help the graphics in games?

Actually your performance benefit would range from 35.4% (SB with no IPC improvement over Nehalem) to 69.3% (SB +25% IPC) with video encoding coming in at 75.3% faster or more...

Actually quoting it for Acanthus, but here it goes.

Conroe-->Penryn = ~7%
Penryn-->Lynnfield = 10%
Lynnfield-->Sandy Bridge = ~20%

That's 50% boost along with the counting the clock speed increase, at single threaded apps. Nehalem brings no better benefit than Penryn did in single threaded apps, and with amount of cores being the same, the multi-threaded benefits will be likely small. Even if it gains a bit with lower memory latency, it could be countered by being GPU-bound.

So I stand for 50%. 🙂

 

[IntelUser2000]

There was an article I read where it said Intel was considering 6-core parts on LGA1155 Sandy Bridge, but that would be hard to find. Graphics + 6 cores should end up similar die size to a 8 core no graphics part.

 

[ModestGamer]

I predict that if BD is competitive intel will magically remember how to unlock every cpu that they sell. If bulldozer is late/slow ala barcelona, we will have 3 overclocking choices:

1. amd's older crap
2. 1366/1156 intel
3. spend shitload on 2011

I really hate this move. All enthusiast consumers should, because if intel felt that BD was going to be competitive they never would have gone down this route in the first place.

Intel is putting out some nicely spaced quality parts. I also honestly belive do to massive arrogance on there part that if they think AMD and Bulldozer won't meet or beat SB etc performance,it would be a potential FUBAR of epic proportions.

2011 is going to be a great year for the PC.

That siad SB stays on my upgrade list until BD comes out and then I have choices to make.

 

[Idontcare]

Intel is putting out some nicely spaced quality parts. I also honestly belive do to massive arrogance on there part that if they think AMD and Bulldozer won't meet or beat SB etc performance,it would be a potential FUBAR of epic proportions.

2011 is going to be a great year for the PC.

That siad SB stays on my upgrade list until BD comes out and then I have choices to make.

Its not really arrogance at play here, its history. Many of us lived through the "native quadcore" hype that preceded the debut of Phenom.

We lived through the months and months of AMD well-wishers bashing Intel's MCM'ed kenstfield parts in the time that spanned from the release of Kentsfield up until Phenom was finally released.

Its not arrogance, its wariness. We are wary of claims of future products upstaging the existing proven architecture leader. Why shouldn't we be wary?

 

[ModestGamer]

Its not really arrogance at play here, its history. Many of us lived through the "native quadcore" hype that preceded the debut of Phenom.

We lived through the months and months of AMD well-wishers bashing Intel's MCM'ed kenstfield parts in the time that spanned from the release of Kentsfield up until Phenom was finally released.

Its not arrogance, its wariness. We are wary of claims of future products upstaging the existing proven architecture leader. Why shouldn't we be wary?

I really really doubt AMD would spend this much time on a sidegrade. While intel is ticking/tocking AMD looked at the issues it faced and decided to revist the whole design. It implies something that is becoming abundently clear. CPU performance is not going to improve until the architectures are completely revisited. AMD is going the long hard play here. Will first gen BD kick intels ass ??? Who knows. Will it be competitive. I think so. will the generation right after it be a big upgrade once they get alot of feedback from the wild. I'd bet the farm on it.

Only becuase they are visiting the core problems that are partly OS driven and some are cpu driven.

I think AMD has a good shot at bringing us a new design by the looks of things that could really knock our socks off.

or it'll be a dud.

I am awaiting performance data to condem either.

Intel is again offering us a nice upgrade path even if the socket thing does piss me off a bit.

 

[ydnas7]

@IntelUser2000 - Erm, may I ask how a transcode unit would help the graphics in games?

i would venture that a transcode unit would assist the GPU side with some fairly have heavy S3 texture compression. a game with a focus on that could benefit greatly. i expect the transcode unit is integrated enough with GPU to assist with texturing.

its even conceivable that the transcode unit could perform some post-processing like on the X-box 360, although thats less likely. It depends on how flexible/accessible the transcode unit is.

 

[Nemesis 1]

I really really doubt AMD would spend this much time on a sidegrade. While intel is ticking/tocking AMD looked at the issues it faced and decided to revist the whole design. It implies something that is becoming abundently clear. CPU performance is not going to improve until the architectures are completely revisited. AMD is going the long hard play here. Will first gen BD kick intels ass ??? Who knows. Will it be competitive. I think so. will the generation right after it be a big upgrade once they get alot of feedback from the wild. I'd bet the farm on it.

Only becuase they are visiting the core problems that are partly OS driven and some are cpu driven.

I think AMD has a good shot at bringing us a new design by the looks of things that could really knock our socks off.

or it'll be a dud.

I am awaiting performance data to condem either.

Intel is again offering us a nice upgrade path even if the socket thing does piss me off a bit.

Nice post but I would very much like to point out that all the work AMD is doing on BD being moving innovation to me sounds like a crock.

Intel hasn't released 48 core cloud chips to universities and others . Intel has't delivered larrabbee boaeds to developers or Have they . I don't mind cheering for AMD . But when ya make it sound loke intel has stopped innovating The record shows just the reverse of what your saying . If it all about cores / Intels 48 cloud chips sound great even better sounding than 16 core 8 module BD. Haven't got a clue about performance don't care either . BUT holy shit 48 cores all working together on a mesh design . or 32 core larrabbee with ring bus sounds pretty good. Both these are available to developers that intel chooses. So don't worry about intel falling behind the giant AMD

As for the sockets Buy AMD than . Bob has one job left to do for me . That happens to to be about M/B . Both AMD and Intel will be spouting new sockets . IF the pricies aren't in line our group will be sueing . Volumn sales = cheaper production cost and were ready to prove our case we been working on it for over 4 years now. At considerable cost I might add.

 

[ModestGamer]

Nice post but I would very much like to point out that all the work AMD is doing on BD being moving innovation to me sounds like a crock.

Intel hasn't released 48 core cloud chips to universities and others . Intel has't delivered larrabbee boaeds to developers or Have they . I don't mind cheering for AMD . But when ya make it sound loke intel has stopped innovating The record shows just the reverse of what your saying . If it all about cores / Intels 48 cloud chips sound great even better sounding than 16 core 8 module BD. Haven't got a clue about performance don't care either . BUT holy shit 48 cores all working together on a mesh design . or 32 core larrabbee with ring bus sounds pretty good. Both these are available to developers that intel chooses. So don't worry about intel falling behind the giant AMD

As for the sockets Buy AMD than . Bob has one job left to do for me . That happens to to be about M/B . Both AMD and Intel will be spouting new sockets . IF the pricies aren't in line our group will be sueing . Volumn sales = cheaper production cost and were ready to prove our case we been working on it for over 4 years now. At considerable cost I might add.

even AMD is dropping a new socket. The first one since like 2007 or something along those lines. Intel isn't really doing alot of innovating. they are doing a termendous job in refining. AMD is going to attempt to innovate.

Which will reign supreme ? I dunno. I think if they don't screw up bulldzoer its gonna kill intel in performance. Does that work out to market share ??? Not always.

as for cpu and core utlitization. Thats a OS issue and remains one.

 

[Idontcare]

-Intel isn't really doing alot of innovating. they are doing a termendous job in refining.

I have to agree here. They stated every tock was going to be a "new" architecture...but the reality is we don't really get a tick-tock...we just get a tick-tick-tick-tick-tick progression with a node transition and some mild ISA extensions thrown in here and there.

This is the very reason I am typing this on a Q6600 desktop (that I use for a lot of computationally bound apps)...nearly 4yrs old but to "upgrade" to at best 2x the performance I'd have to drop a shizerload of cash.

From Intel we are getting drip-drop-drip-drop evolution instead of tick-tock revolution.

 

[khon]

I have to agree here. They stated every tock was going to be a "new" architecture...but the reality is we don't really get a tick-tock...we just get a tick-tick-tick-tick-tick progression with a node transition and some mild ISA extensions thrown in here and there.

This is the very reason I am typing this on a Q6600 desktop (that I use for a lot of computationally bound apps)...nearly 4yrs old but to "upgrade" to at best 2x the performance I'd have to drop a shizerload of cash.

From Intel we are getting drip-drop-drip-drop evolution instead of tick-tock revolution.

Are you really going to argue that SB isn't a real tock ? Graphics on die seems like a fairly significant step to me. And Haswell sounds like it could be a significant change as well, with possible gpcpu components.

 

[ModestGamer]

Are you really going to argue that SB isn't a real tock ? Graphics on die seems like a fairly significant step to me. And Haswell sounds like it could be a significant change as well, with possible gpcpu components.

LOL AMD's doing it to and they have a much better graphics core then intel by miles. the CPU core however is nothing to write home about. Moe tightening of the ship.

The real reason AMD changed directions is going to be obvious by the next intel generation. when intel can't really do anything but process shrink with the current architecture.

 

[Nemesis 1]

even AMD is dropping a new socket. The first one since like 2007 or something along those lines. Intel isn't really doing alot of innovating. they are doing a termendous job in refining. AMD is going to attempt to innovate.

Which will reign supreme ? I dunno. I think if they don't screw up bulldzoer its gonna kill intel in performance. Does that work out to market share ??? Not always.

as for cpu and core utlitization. Thats a OS issue and remains one.

I don't see it that way at all, AMD is trying to do more threads with Real cores in the same manner they approached in the long term Real men have Fabs. AS IDC already said it about PH I Hype.It was in the same range performance wise as this hype. SB has on average with out any turdo or ht IPC wise 20+ faster than nehalem at same clock without AVX. AS I already stated Intel took a design concept with both Cloud special and larrabbee special and put them into test systems . They basicly said here is 48 core cloud on mesh and 32 larrabee on ringbus . Lets set what ya can do with all these cores in the mean time will continue to refine the projects. I like this kind of innovation were the fab says here lets see what ya can do with working silly. <- INNOVATION AT WORK NOT SLIDES not TALK real working silicon and intel is out of the gate first . When we can use all these cores Intel will be ready.

I would also like to point out BD looks OK for servers. Lots of threads.

But I have to say If people would have talked like this 6 years in a enthusiast forum about performance on the desktop Threads being more important than IPC when 6 cores and 12 threads on Desktop is amuzing . Its dajavou of AMD64 big hype and just recently is it coming into its own . I would guess that multi threading = or greater 64 bit in complexity.

BD 4 module 8 core 8 threads will be up against SB 8 core 16 thread than in 2011 1st qt Low end IVY will likely bitch slap amds top of the line on desktop. Just like lowend SB is going to bitch slap present AMD highend desktop. 4 cores 8 threads is more than I need , But others can use more threads I am sure. This place and others are more about servers now than performance. You are going to love IDF. Going to be the best ever,

 

[Idontcare]

Are you really going to argue that SB isn't a real tock ? Graphics on die seems like a fairly significant step to me.

has what to do with the architecture? if they upgrade the sram does that count too?

And Haswell sounds like it could be a significant change as well, with possible gpcpu components.

I've said it before as well, Haswell is the first architecture since Core that has me intrigued.

But yes, from my perspective it has been just tick-tick-tick since the tock of 2006.

Was PIII really a tock to the PII? No. The P4? Oh hells yes. Core? hells yes. nehalem?

yawn, add an IMC and HT..when they added HT to P4 they didn't call it a new architecture, and moving an existing circuit like the memory controller from the NB to the CPU isn't exactly innovation either in my book (sorry AMD)...so much so that they actually took it back off the die for 32nm clarkdale.

Haswell has potential. Shoving stuff that already existed elsewhere in the system onto a SOC isn't what I'd call a tock. Whether the memory controller and IGP is on-die or off-die the x86 architecture is still doing 1+1=2 the same old way.

Don't get me wrong, performance has improved, every tick is expected to do that, but the tocks are expected to do a little more even still, otherwise why bother calling/labeling them as a tock?

Did nehalem change the pipeline over penryn in as much of a way as conroe did over prescott? What about Sandy?

 

[khon]

I think you're asking for too much. Intel has been making x86 processors for over 30 years now, so you're really not going to have many drastic changes at this point.

In that context I do think that graphics on die is a fairly large step, and they also added 10&#37; improvement per clock.

 

[ModestGamer]

I don't see it that way at all, AMD is trying to do more threads with Real cores in the same manner they approached in the long term Real men have Fabs. AS IDC already said it about PH I Hype.It was in the same range performance wise as this hype. SB has on average with out any turdo or ht IPC wise 20+ faster than nehalem at same clock without AVX. AS I already stated Intel took a design concept with both Cloud special and larrabbee special and put them into test systems . They basicly said here is 48 core cloud on mesh and 32 larrabee on ringbus . Lets set what ya can do with all these cores in the mean time will continue to refine the projects. I like this kind of innovation were the fab says here lets see what ya can do with working silly. <- INNOVATION AT WORK NOT SLIDES not TALK real working silicon and intel is out of the gate first . When we can use all these cores Intel will be ready.

I would also like to point out BD looks OK for servers. Lots of threads.

But I have to say If people would have talked like this 6 years in a enthusiast forum about performance on the desktop Threads being more important than IPC when 6 cores and 12 threads on Desktop is amuzing . Its dajavou of AMD64 big hype and just recently is it coming into its own . I would guess that multi threading = or greater 64 bit in complexity.

BD 4 module 8 core 8 threads will be up against SB 8 core 16 thread than in 2011 1st qt Low end IVY will likely bitch slap amds top of the line on desktop. Just like lowend SB is going to bitch slap present AMD highend desktop. 4 cores 8 threads is more than I need , But others can use more threads I am sure. This place and others are more about servers now than performance. You are going to love IDF. Going to be the best ever,

if you say so. I'll wait for the testing.it's also a shame we are stuck with poor performing os's that can't capitolize on the treading.I still see amd coming out swinging but you can think what you want.

 

[TuxDave]

nehalem?

yawn, add an IMC and HT..when they added HT to P4 they didn't call it a new architecture, and moving an existing circuit like the memory controller from the NB to the CPU isn't exactly innovation either in my book (sorry AMD)...so much so that they actually took it back off the die for 32nm clarkdale.

There's only so much you can get done in 2 years. It's a little saddening to see people sum up Core i7 as HT + IMC and that's it.