Dual Sandy Bridge-EP Cinebench Benchies!

[Arachnotronic]

From JCornell, a well known Intel leaker.

 

[Idontcare]

2.2GHz?

 

[grkM3]

95 watt for an 8 core on 32nm 🙂

 

[sm625]

So he goes from 2.2GHz to 3.0GHz and yet the score only goes up by 6%? Is the one at 3.0GHz the same exact chip? Does that mean that at 2.2GHz, turbo has room to run, but at 3.0GHz, turbo isnt used?


I dont see how this is good since the 12C opteron outperforms it on a per core+HT basis. ie, a 12 thread opteron scaled up to 32 threads would score 21.2 at 2.6GHz. Thats assuming no performance boost or loss with bulldozer. Now if dual interlagos is 85% as efficient, and any more than 15% faster, then it will beat this no problem.

 

[Arkadrel]

This is about Sandy Bridge E 2.2ghz x 8cores/16threads, in dual motherboards.

So we re looking at what? 16cores/32threads x 2.2ghz.

SandyBrigdeE with 32 threads (16cores+HT) x 2.2ghz = scores 20.29 in Cinebench 11.5

32 threads.... Wow...

 

[grkM3]

There was a 10 core es for sale on ebay.is untel going to go 10 core on 32nm or wait for 22?

A dual 10 core with 40 threads would be a power house

 

[BlueBlazer]

From JCornell, a well known Intel leaker.

Seems to good to be true IMHO. Is it me or does it show a single 8C/16T 2.2GHz Sandy Bridge EP (taking the score and divide by 2) almost as fast as, or faster (compensating for performance scaling loss in a 2 socket system) than 2.9GHz 8C/16T Xeon (in the same chart)? 😛

 

[Arachnotronic]

JCornell is a pretty reputable leaker. If it were anyone but him, I'd be skeptical. But I'm confident this is the real deal.

 

[Edrick]

JCornell is a pretty reputable leaker. If it were anyone but him, I'd be skeptical. But I'm confident this is the real deal.

I remember he first posted screnshots of this 8 core LGA2011 chip at the end of 2010. Then they were taken down a day later.

 

[Edrick]

I dont see how this is good since the 12C opteron outperforms it on a per core+HT basis. ie, a 12 thread opteron scaled up to 32 threads would score 21.2 at 2.6GHz. Thats assuming no performance boost or loss with bulldozer. Now if dual interlagos is 85% as efficient, and any more than 15% faster, then it will beat this no problem.

So you are comparing a 32 core Opteron setup @ 2.6Ghz, to a 16 core + HT SB-E setup @ 2.2Ghz, and using that as an arguement to claim Opteron outperforms it? It is common knowledge that HT only supplies a 10-30% performance increase, depending on application. So I find your comparison and conclusions completely off target.

 

[Idontcare]

JCornell is a pretty reputable leaker. If it were anyone but him, I'd be skeptical. But I'm confident this is the real deal.

He's legit, and sanctioned. I remember when he leaked a bunch of stuff about clarkdale and Francois Peidonel went ballistic claiming it was a violation of NDA and yaddi yadda...then someone "internal" helped him understand the backstory and suddenly Francois was mum about JC's consistent, timely, dependable "leaks".

He's legit, but there is nothing "unsanctioned" about what he leaks, when he leaks it, and how he leaks it.

That said, I'm still going back to the "2.2GHz" factor. That is a yawner.

Even 3.2GHz would be a yawner for a chip that is taking nearly a full extra year to debut after the 3.4GHz 4C 2600K. I understand that it is 8C, but come on this is supposed to be the extreme enthusiast headliner CPU, bust out that 130W, overclock it to have 250W TDP (GPU coolers can take the heat, surely HSF for CPU's can as well) and show us something that is actually worth $1k.

I expect 2.2GHz yawners from AMD, they have an excuse considering they are relying on a foundry process and have 1/6 the resources to throw at any given project. What is Intel's excuse?

 

[Edrick]

I expect 2.2GHz yawners from AMD, they have an excuse considering they are relying on a foundry process and have 1/6 the resources to throw at any given project. What is Intel's excuse?

I recall the first ES samples of SB LGA1155 were around 2.0Ghz. Yet they were released much higher. So I could really care less what an ES sample is set as. I fully expect them to be released at a very reasonable speed considering the price.

 

[sm625]

So you are comparing a 32 core Opteron setup @ 2.6Ghz, to a 16 core + HT SB-E setup @ 2.2Ghz, and using that as an arguement to claim Opteron outperforms it? It is common knowledge that HT only supplies a 10-30% performance increase, depending on application. So I find your comparison and conclusions completely off target.

Yeah but the two options are likely to cost the very same amount, so why not? Core counting is off target. Performance per dollar is what is on target.

 

[Edrick]

Yeah but the two options are likely to cost the very same amount, so why not? Core counting is off target. Performance per dollar is what is on target

If you want to talk about cost and not consider cores, then fine.

A dual socket Intel system = 16C and 32T with HT. Which gives a score of 20.29 running @ 2.2Ghz. We know those speeds will be faster at release.

A dual socket AMD Opteron system = 24C and 24T. Which gives a score of 15.90 running @ 2.6Ghz.

As soon as you start talking about more cores than that, we are talking about quad socket system which increase price significantly. So I still do not find any logic in your arguement.

 

[BlueBlazer]

JCornell is a pretty reputable leaker. If it were anyone but him, I'd be skeptical. But I'm confident this is the real deal.

Yeah, I know, just found the source here at Coolaler forums. I've also compared with these Cinebench R11.5 results. 😉

That said, I'm still going back to the "2.2GHz" factor. That is a yawner.

Even 3.2GHz would be a yawner for a chip that is taking nearly a full extra year to debut after the 3.4GHz 4C 2600K. I understand that it is 8C, but come on this is supposed to be the extreme enthusiast headliner CPU, bust out that 130W, overclock it to have 250W TDP (GPU coolers can take the heat, surely HSF for CPU's can as well) and show us something that is actually worth $1k.

Considering there are twice the number of cores (8 cores in Sandy Bridge EP vs 4 cores in Sandy Bridge), its very likely the lower clock speed it to keep it within a TDP limit. That said, there's a possibility of higher clocked models (some with a few cores disabled?). 🙂

As soon as you start talking about more cores than that, we are talking about quad socket system which increase price significantly. So I still do not find any logic in your arguement.

There's also a performance penalty going from single socket to dual socket, as well as going to quad socket system. Generally quad socket systems do not scale as well as dual socket systems. Its one of the reasons that majority of HPC machines and supercomputers use dual socket blades. 🙂

 

[Arachnotronic]

Also, about the 2.2GHz yawner, here's a listing of all the upcoming Xeon E5 models:

http://www.cpu-world.com/news_2011/2011082601_Intel_to_launch_Xeon_E5-2600_series_in_Q4_2011.html

 

[lol123]

8 cores at 3.1 GHz with 20 MB L3 cache should be pretty fast. I wonder what the turbo frequencies will be like.

 

[Idontcare]

Also, about the 2.2GHz yawner, here's a listing of all the upcoming Xeon E5 models:

http://www.cpu-world.com/news_2011/2011082601_Intel_to_launch_Xeon_E5-2600_series_in_Q4_2011.html

Xeon E5-2687W

3.1GHz, 8C/16T, 20MB L3$ <- now that's more like it.

If my 3.4GHz 4C/8T 8MB L3$ chip cost a mere $315, I expect stuff to double up in the specs if I'm expected to triple up on the purchase price.

Personal opinion of course, sometimes folks are willing to pay 2X for a 10% perf bump too.

 

[ViRGE]

8 cores at 3.1 GHz with 20 MB L3 cache should be pretty fast. I wonder what the turbo frequencies will be like.

It has to be pretty good, particularly when using 4/2/1 cores. That's a lot of TDP that can otherwise be used to goose those remaining cores.

 

[skipsneeky2]

:biggrin: While everyone is focused on 8 and 16 core monstor's that will make any 980x/990x owner sweat,i'm hoping they will release a 65watt quad core.

But if we are getting 8 cores even at 2.2ghzs on 95w that's damn impressive when we needed 130w for 6 cores on x58.

 

[dac7nco]

i'm hoping they will release a 65watt quad core.

eh?

http://www.newegg.com/Product/Produc...82E16819115075

 

[exar333]

:biggrin: While everyone is focused on 8 and 16 core monstor's that will make any 980x/990x owner sweat,i'm hoping they will release a 65watt quad core.

But if we are getting 8 cores even at 2.2ghzs on 95w that's damn impressive when we needed 130w for 6 cores on x58.

If you check out the lineup for SB-E Xeons, they already have announced a 6-core @ 60w and an 8-core @ 70w.

http://forums.anandtech.com/showthread.php?t=2187992

 

[BlueBlazer]

Xeon E5-2687W

3.1GHz, 8C/16T, 20MB L3$ <- now that's more like it.

At 150W TDP though.... 😀

If my 3.4GHz 4C/8T 8MB L3$ chip cost a mere $315, I expect stuff to double up in the specs if I'm expected to triple up on the purchase price.

Personal opinion of course, sometimes folks are willing to pay 2X for a 10% perf bump too.

Who knows, maybe 8 cores in SBE will scale better in performance than SB and Gulftown given that quad channel memory controller. Just have to wait for some good leaks of official benches. 😉

 

[Idontcare]

At 150W TDP though.... 😀

Who knows, maybe 8 cores in SBE will scale better in performance than SB and Gulftown given that quad channel memory controller. Just have to wait for some good leaks of official benches. 😉

I'm one of those fringe-enthusiasts who looks over at the GPU industry where they routinely cool 250W GPUs and nowadays sell stock products that go as high as 350W TDP's and I say to myself "why not CPU's too?".

Its not to say I am going to buy a 350W TDP CPU, I don't own a 250W GPU either.

But the technology exists to cool 350W TDP products, and it is not ridiculously expensive either as evidenced by the price points of those products, so there is no technological reason to not offer 300W TDP stock processors at whatever rate the market will support IMO.

My B3 stepping QX6700 (electron guzzlers) at 4GHz used around 270-300W by my measurements, it was not the end of the world. The performance though was out of this world at the time (2006).

I would hope the performance of Intel's top of the line, or AMD's, some 5yrs later is going to be out of this world. And so what if it takes a higher TDP tier to achieve that?

I want AMD to launch a 250W TDP SKU for their halo FX line, at whatever clocks that gets them (5.5GHz to 6GHz by my estimates). AMD fanboys will squeal with delight, anti-AMD fanboys will scream with despair, and the market will have one more option available to choose from for their upgrade purchases.

 

[dbcooper1]

I'm one of those fringe-enthusiasts who looks over at the GPU industry where they routinely cool 250W GPUs and nowadays sell stock products that go as high as 350W TDP's and I say to myself "why not CPU's too?".

Its not to say I am going to buy a 350W TDP CPU, I don't own a 250W GPU either.

But the technology exists to cool 350W TDP products, and it is not ridiculously expensive either as evidenced by the price points of those products, so there is no technological reason to not offer 300W TDP stock processors at whatever rate the market will support IMO.

My B3 stepping QX6700 (electron guzzlers) at 4GHz used around 270-300W by my measurements, it was not the end of the world. The performance though was out of this world at the time (2006).

I would hope the performance of Intel's top of the line, or AMD's, some 5yrs later is going to be out of this world. And so what if it takes a higher TDP tier to achieve that?

I want AMD to launch a 250W TDP SKU for their halo FX line, at whatever clocks that gets them (5.5GHz to 6GHz by my estimates). AMD fanboys will squeal with delight, anti-AMD fanboys will scream with despair, and the market will have one more option available to choose from for their upgrade purchases.

Wow, what are the CPU sockets and the battery cables supplying that kind of current to them going to look like? We're going to need big CPU/socket pins/pads or lots of little ones.