The Sandy Bridge Preview
by Anand Lal Shimpi on August 27, 2010 2:38 PM ESTA New Architecture
This is a first. Usually when we go into these performance previews we’re aware of the architecture we’re reviewing, all we’re missing are the intimate details of how well it performs. This was the case for Conroe, Nehalem and Lynnfield (we sat Westmere out until final hardware was ready). Sandy Bridge, is a different story entirely.
Here’s what we do know.
Sandy Bridge is a 32nm CPU with an on-die GPU. While Clarkdale/Arrandale have a 45nm GPU on package, Sandy Bridge moves the GPU transistors on die. Not only is the GPU on die but it shares the L3 cache of the CPU.
There are two different GPU configurations, referred to internally as 1 core or 2 cores. A single GPU core in this case refers to 6 EUs, Intel’s graphics processor equivalent (NVIDIA would call them CUDA cores). Sandy Bridge will be offered in configurations with 6 or 12 EUs.
While the numbers may not sound like much, the Sandy Bridge GPU is significantly redesigned compared to what’s out currently. Intel already announced a ~2x performance improvement compared to Clarkdale/Arrandale, and I can say that after testing Sandy Bridge Intel has been able to achieve at least that.
Both the CPU and GPU on SB will be able to turbo independently of one another. If you’re playing a game that uses more GPU than CPU, the CPU may run at stock speed (or lower) and the GPU can use the additional thermal headroom to clock up. The same applies in reverse if you’re running something computationally intensive.
On the CPU side little is known about the execution pipeline. Sandy Bridge enables support for AVX instructions, just like Bulldozer. The CPU will also have dedicated hardware video transcoding hardware to fend off advances by GPUs in the transcoding space.
Caches remain mostly unchanged. The L1 cache is still 64KB (32KB instruction + 32KB data) and the L2 is still a low latency 256KB. I measured both as still 4 and 10 cycles respectively. The L3 cache has changed however.
Only the Core i7 2600 has an 8MB L3 cache, the 2400, 2500 and 2600 have a 6MB L3 and the 2100 has a 3MB L3. The L3 size should matter more with Sandy Bridge due to the fact that it’s shared by the GPU in those cases where the integrated graphics is active. I am a bit puzzled why Intel strayed from the steadfast 2MB L3 per core Nehalem’s lead architect wanted to commit to. I guess I’ll find out more from him at IDF :)
The other change appears to either be L3 cache latency or prefetcher aggressiveness, or both. Although most third party tools don’t accurately measure L3 latency they can usually give you a rough idea of latency changes between similar architectures. In this case I turned to cachemem which reported Sandy Bridge’s L3 latency as 26 cycles, down from ~35 in Lynnfield (Lynnfield’s actual L3 latency is 42 clocks).
As I mentioned before, I’m not sure whether this is the result of a lower latency L3 cache or more aggressive prefetchers, or both. I had limited time with the system and was unfortunately unable to do much more.
And that’s about it. I can fit everything I know about Sandy Bridge onto a single page and even then it’s not telling us much. We’ll certainly find out more at IDF next month. What I will say is this: Sandy Bridge is not a minor update. As you’ll soon see, the performance improvements the CPU will offer across the board will make most anyone want to upgrade.
Facebook
Twitter
RSS
200 Comments
View All Comments
gruffi - Friday, August 27, 2010 - link
Why not comparing with a HD 5570? That is what Llano is supposed to have, Redwood-class IGP. An HD 5450 is quite pointless. It just reflects competition for Ontario. But Sandy Bridge is not Ontario's competition.And what about image quality or GPGPU support? Pure FPS numbers are only half of the truth.
wiak - Friday, August 27, 2010 - link
dont think so, its said that AMD's Fusion built-in GPU will have 400 SPUs (HD 5670 level-graphics), thats a far cry from HD 5450's 80 SPUs ;)so if you wanna game you still have to use something from a real graphics manufacture like AMD when it comes to GPUs bult into CPUs, as a added bonus you also have updated drivers and a decade old DirectX 9 compatibility, so you old games work without any big problems
icrf - Friday, August 27, 2010 - link
I am impressed that you have a functioning sample at least four months before it's available, run it through enough paces for a review like this, and they let you release the numbers. I mean, are they trying to suppress holiday sales?When do you think you'll have a Bulldozer sample from AMD to run a similar preview? Barring a surprise from AMD, at this point, it looks like I'll be building an i7 2600 early next year. The similar spec chip from today is an i7-975 Extreme, which is the fastest quad core in the bench, and Sandy Bridge runs 13-14% faster in the only benchmark I care about (x264). I guess even that might change significantly if it can take advantage of this "alleged on-die video transcode engine." I'd not heard of that before.
Anand Lal Shimpi - Friday, August 27, 2010 - link
Honestly we're probably several months out from having Bulldozer silicon in a similar state. With the past few generations of Intel CPUs, by around 4 - 6 months before launch we're usually able to get access to them and they perform very well.With AMD the lead time is far shorter. I don't expect us to have access to Bulldozer silicon that's worth benchmarking until Q2 2011 at the earliest. I'm more than happy to be proven wrong though :-P
icrf - Friday, August 27, 2010 - link
I guess I'm mostly surprised that Intel would do it. Conroe made sense. They had to show the world as early as possible that they had something significantly faster than AMD, suppressing sales of that for their own a little later. But now that they own that performance crown, why show previews so many months early? I suppose I could be over-analyzing it and the vast majority of the market couldn't care less so it makes little difference to their bottom line. Bragging rights simply make for good PR.Sad to see Bulldozer so far out. I assume the server chips will ship before the consumer ones, too, so it'll be at least a solid year before it could be in my hands, anyway. Oh well. To be honest, my C2D E6400 still does well enough for me. Maybe I'll just make my upgrade an Intel G3 SSD. If I got both that and SB, I don't know what I'd do with myself.
Thanks, and keep up the good work.
Anand Lal Shimpi - Saturday, August 28, 2010 - link
This preview wasn't Intel sanctioned, I believe Intel will release its own numbers at IDF in a few weeks.Take care,
Anand
icrf - Saturday, August 28, 2010 - link
Oh, I had assumed you got this chip from Intel and they had a typical NDA that said when you could talk about what you found. Where'd it come from, then? One of Intel's motherboard partners with whom you have a friendly relationship?aegisofrime - Saturday, August 28, 2010 - link
I must say, I'm really grateful for this article. I'm in the middle of planning an upgrade and information like this is really valuable to me. (and I guess to a lot of people as well!) I would just like you to know that your articles actually do influence some of our buying choices. So... Thank you! :DNow, all I need is a Bulldozer preview and all the pieces are in place...
vajm1234 - Friday, August 27, 2010 - link
few things clear and few unclear as of nowthis sandy bridge review sample here do not have TURBO enabled. The CPU runs at 3.1GHz all the time, regardless of workload as anand stated
it says "Both the CPU and GPU on SB will be able to turbo independently of one another. If you’re playing a game that uses more GPU than CPU, the CPU may run at stock speed (or lower) and the GPU can use the additional thermal headroom to clock up. The same applies in reverse if you’re running something computationally intensive."
QUESTIONS
Q} will the on die GPU unit work in tandem with the other discrete GPUs out there or it will shut off? if yes will it work when sli or crossfire is enabled :p
Q} whatever the above statement says will it happen if we use discrete graphics from nvidia or ati?
Q} will there be any possibility to disable ONLY GPU and in certain cases ONLY its TURBO FEATURE
Q} any possibility to remain the GPU overclocked the whole time when cpu is IDLE
Q} what about accelerated hd video playback using the on die gpu?
Q} it support VT-x and AVX is it possible for you anand to use specific benchmark for these instructions, same request goes for the AMD
Q} as someone asked will there be a cheap 6 => core processor for mainstream market
Q} again as per the last comment ......When do you think you'll have a Bulldozer sample from AMD to run a similar preview?
this Ques Must be answered
all n all what i think even if there is a 15-19% perf. Jump its not worh the spending when u consider u have to upgrade the entire platform
and moreover limiting Overclocking features damm! a retarded decision i am not in a mood for amd but if the overclocking hits then i will move 10000...% :angry:
regards
DanNeely - Saturday, August 28, 2010 - link
If you're asking about an SLI/CFX pairing with the IGP almost certainly not. The only company to ever attempt something like that has been Lucid with the Hydra chip and the results have been less than impressive. Architecturally I don't know that it'd even be possible for them to try with the on die GPU. The Hydra chip sat between the CPU and the Gfx cards on the PCIe bus and looked like a single card to the OS. There's no way for them to insert themselves into the middle of the connection to the IGP.