• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

** SandyBridge-E Benchmarking **

rjk

rjk

rjk

Caporegime
Joined
8 Aug 2007
Posts
25,380
Hi Guys

With the launch of SB-E today, here is the bits and bobs I have done on it.
we haven't had much time thrashing it and with the cost of the CPUs I was being gentle. Didn't do much messing with the BCLK, just wanted to see the max frequency.

This was just using the Stock Intel cooler, so temps aren't ideal.
All testing done with 32GB of ram :D


Anyway, enjoy...

Stock temps on LinX
stocklinxtemps.jpg


4.8GHz Temps on LinX
48ghzlinxtemps.jpg


Stock Cinebench
CinebenchStock.jpg


4.8GHz Cinebench
48ghzcinebench.jpg


Stock Vantage [with a single GTX 580]
3dmarkvantagestock.jpg


4.8GHz Vantage [with a single GTX 580]
3dmarkvantage48GHz.jpg


5GHz Cinebench
5ghzcinebench.jpg
 
Do you have any 3930K's to test?

Not many people are going to be interested in the 3960K regardless of how well it overclocks simply because of the price.
 
The 12 thread Core i7-3960X is only 17% faster than a 8 thread Core i7 2600K in productivity applications:

http://media.bestofmicro.com/0/G/314944/original/compare 2600k.png

In the best case scenario with multi-threaded applications the Core i7-3960X is around 35% faster.

It is 28% faster than a Core i5 2500K and 32% faster than a FX8150 too:

http://media.bestofmicro.com/0/F/314943/original/compare 2500k.png

http://media.bestofmicro.com/0/H/314945/original/compare fx-8150.png

All these CPUs are £170 to £250. The Core i7-3960X is around £800 IIRC and has more expensive motherboards too.

On top of this the Core i7-3960X is only 12% faster than a Core i7 990X:

http://media.bestofmicro.com/0/E/314942/original/compare 990x.png

The Core i7 990X is also a 32NM CPU.
 
Last edited:
Well those numbers aren't quite right, according to the results up there, Cinebench is basically MASSIVELY faster than a 2500k, you're talking 50-55% faster, a lot of that will be better scaling due to bandwidth I would think.

Interestingly I did point out to people, and was ignored as usual because people seem to love slating Bulldozer, that when you get a certain amount of cores, you need increasingly large amounts of cache to keep things fed, AND more specifically in server duties(both bulldozer and sb-e are such chips) this requirement is bigger.

people keep talking about 8-12 threads.... these chips are NOT built for 8-12 threads, but built for DOZENS if not HUNDREDS of threads that are competing for resources. 8mb l3 is wasted in gaming, but when 100 people all want to access different webpages on a website and a chip is running through different threads and constantly switching, its insanely important.

But look how huge SB-e is, Bulldozer, the FIRST real high end chip and the first chip designed for 32nm from AMD(llano is not at all, its been ported multiple times and a very old architecture baked into something quite brilliant).

SB-E, 2.27billion transistors for 50% more cores the transistor count has either just over or under doubled(950ish million for SB is actually the "design" transistor count, the real life count goes up by anything up to 15-20%, the 2.27billion could be the real or the on paper number) is huge, thats largely due to the cache.

Yet Bulldozer is 315mm2, compared to SB-E's 435mm2, while Bulldozer is 2billion transistors vs SB-E's 2.27billion.

Power wise aswell, Bulldozer uses less power, and is clocked higher. People who called Bulldozer a power dog, are stupid, people who called bulldozer huge, are stupid. This is your real comparison, performance on Bulldozer, sucks, 2billion transistors in only 315mm2, and 125W, is freaking awesome.......... performance per transistor or per W might be horrible, but its size and actual total power usage are anything but.
 
Shayper said:
Cat seems to just be posting slightly wrong negative information, in this and the other thread.

I have no idea why.
Click to expand...

???

No I am not. Maybe you should ask Toms Hardware then instead of trying to blame me for the results they have got.

CAT-THE-FIFTH said:
The 12 thread Core i7-3960X is only 17% faster than a 8 thread Core i7 2600K in productivity applications:

http://i559.photobucket.com/albums/ss33/CAT-THE-FIFTH/compare2600k.png

In the best case scenario with multi-threaded applications the Core i7-3960X is around 35% faster.
Click to expand...

compare2600k.png


Toms Hardware used actual applications instead of using synthetic benchmarks. Many of the applications are the latest versions too like Photoshop CS5.5(not CS4 or CS5) and 3ds Max 2012(not 2010 or 2011) which means they will have optimal support for SB.
 
Last edited:
20-30% faster than a 2600K in the majority of applications is not to be sniffed at, that will only improve as applications become more CPU heavy.

We should also be looking more at 3930K because clearly that is where the value is at (if you can call it value!), I certainly can't justify 200% of the cost for a meagre 3MB cache (3960X).

In the absolute worst case scenario 3930K is going to be no slower than a 2700K, in the absolute best case scenario it will be 50% faster. Bulldozer is similarly held back by applications except absolute best case scenario it only really matches 2700K and absolute worst case gets pummeled by a whole range of lesser CPU's.
 
Last edited:
drunkenmaster said:
Well those numbers aren't quite right, according to the results up there, Cinebench is basically MASSIVELY faster than a 2500k, you're talking 50-55% faster, a lot of that will be better scaling due to bandwidth I would think.

Interestingly I did point out to people, and was ignored as usual because people seem to love slating Bulldozer, that when you get a certain amount of cores, you need increasingly large amounts of cache to keep things fed, AND more specifically in server duties(both bulldozer and sb-e are such chips) this requirement is bigger.

people keep talking about 8-12 threads.... these chips are NOT built for 8-12 threads, but built for DOZENS if not HUNDREDS of threads that are competing for resources. 8mb l3 is wasted in gaming, but when 100 people all want to access different webpages on a website and a chip is running through different threads and constantly switching, its insanely important.

But look how huge SB-e is, Bulldozer, the FIRST real high end chip and the first chip designed for 32nm from AMD(llano is not at all, its been ported multiple times and a very old architecture baked into something quite brilliant).

SB-E, 2.27billion transistors for 50% more cores the transistor count has either just over or under doubled(950ish million for SB is actually the "design" transistor count, the real life count goes up by anything up to 15-20%, the 2.27billion could be the real or the on paper number) is huge, thats largely due to the cache.

Yet Bulldozer is 315mm2, compared to SB-E's 435mm2, while Bulldozer is 2billion transistors vs SB-E's 2.27billion.

Power wise aswell, Bulldozer uses less power, and is clocked higher. People who called Bulldozer a power dog, are stupid, people who called bulldozer huge, are stupid. This is your real comparison, performance on Bulldozer, sucks, 2billion transistors in only 315mm2, and 125W, is freaking awesome.......... performance per transistor or per W might be horrible, but its size and actual total power usage are anything but.
Click to expand...

What are you on about?

Of course this chip will be better than a measly 2500K on Cinebench. Performance scaling is surprisingly good with more threads and per clock using the benchmark.
 
lol at d'master - trying to make up for his faildozzer predictions :D
are you trying to say that this chip is worse than faildozzer? man,what are you smoking?
you trying to compare size or performance of chips? who *****n cares about size? you amd fans talking about size like it is really matter to average Joe - nobody cares?
fact is,that faildozzer using enormous power and still stays behind previous amd cpu's gen (not to mention Sandy),while this chip is just Sandy step up to 6 cores,and this chips works
fact is,that faildozzer failed miserably,but not this chip,that shows price,this chip leaves 8c faildozzer in dust even before start
funny,that is the same ppl who predicted "bright future for amd f'dozzer" now trying to make this chip bad LOL
 
DM you're banging your head against a brick wall (or intel skulls) on this thread :)

I see where you are coming from re power and size (heat). I cannot find enough testing to say whether either BD or SB-E will reduce power requirement overmuch at low volts. Most testing has been at auto volts so far AFAIK.

Anyway it is all a bit too rich for me, my PC cost less than this CPU.

andy.
 
Last edited:
bhavv said:
Sorry to burst your fantasy bubble, but click the links above. SB-E is to a 2600k what a Bulldozer was to a Phenom X6.
Click to expand...

I wouldn't quite say that at all.
In the games, there's no CPU bottleneck with the Intel CPU's, so it wouldn't make a difference what CPU you used, as long as it wasn't a bottleneck.
From a purely gaming stand point, I fully agree, 2011 isn't worth it over 1155.
Power consumption's a let down as well, in comparison to the 1155 CPU's.
 
Well yea, I was so hyped up over this chip, and was just so hugely disappointed :(

Oh well, bring on some huge gulftown price cuts!!!
 
You must log in or register to reply here.
Back
Top Bottom