Sandy Bridge-E Details Revealed

[ElFenix]

Since when was $200-220 for a 2500K not considered affordable? 😕

You realize that when the Phenom II X4s came out most of them cost above $200 also, right?

*ignores the whole athlon2 line*
*ignores the e2140/60 darlings*
*ignores the previous gen i3s*

 

[HendrixFan]

I've been ready to buy this thing for over 5 months now.

Dibs on your 980x.

 

[Ares1214]

I lol'd


I'll be pleasantly surprised if BD can put up a good showing against a 2600K, much less a 990X

Thats practically like saying the same thing, aside from thread intensive projects, the 2600K and 990x perform very similarly, sometimes the 2600K wins.

 

[996GT2]

Thats practically like saying the same thing, aside from thread intensive projects, the 2600K and 990x perform very similarly, sometimes the 2600K wins.

2500K and 2600K are very competitive against 990X in most benchmarks. I agree with you there.

But they key thing to note here is that BD was designed with heavy multitasking in mind (even at the expense of single core IPC, I might add). In heavily multithreaded apps, 2600K cannot compete with 990X.

That's what I was referring to. BD may be able to keep up with a 2600K in multithreaded apps, but I highly doubt it will be competitive with a 990X there.

In not so heavily multithreaded apps, my prediction is that BD will not be competitive even against a 2600K.

 

[LOL_Wut_Axel]

2500K and 2600K are very competitive against 990X in most benchmarks. I agree with you there.

But they key thing to note here is that BD was designed with heavy multitasking in mind (even at the expense of single core IPC, I might add). In heavily multithreaded apps, 2600K cannot compete with 990X.

That's what I was referring to. BD may be able to keep up with a 2600K in multithreaded apps, but I highly doubt it will be competitive with a 990X there.

In not so heavily multithreaded apps, my prediction is that BD will not be competitive even against a 2600K.

I think you're being too pessimistic. If AMD can achieve 20% higher IPC in comparison to K10.5 then it should match or beat slightly the 990X in multi-threaded apps. It would still be a good 15% behind Sandy Bridge in IPC, though. That means it would lose in most games by a small amount since most only use three or four cores and it would lose by 15% in audio encoding. In everything heavily multi-threaded, though, if it can match the 990X then it should be 20-25% faster.

Food for thought.

 

[996GT2]

I think you're being too pessimistic. If AMD can achieve 20% higher IPC in comparison to K10.5 then it should match or beat slightly the 990X in multi-threaded apps. It would still be a good 15% behind Sandy Bridge in IPC, though. That means it would lose in most games by a small amount since most only use three or four cores and it would lose by 15% in audio encoding. In everything heavily multi-threaded, though, if it can match the 990X then it should be 20-25% faster.

Food for thought.

Let's wait and see. AMD hasn't even released any real benchmarks of Bulldozer and it is way past the originally scheduled June release date. In contrast, solid Sandy Bridge benchmarks were released 4 months before the official release. So, I'm not holding my breath for BD to be some miraculous super-CPU with all of the recent delays and problems that have happened. If it turns out to be a great performer once it's released, great. But I'm not counting on it.

 

[LOL_Wut_Axel]

Let's wait and see. AMD hasn't even released any real benchmarks of Bulldozer and it is way past the originally scheduled June release date. In contrast, solid Sandy Bridge benchmarks were released 4 months before the official release. So, I'm not holding my breath for BD to be some miraculous super-CPU with all of the recent delays and problems that have happened. If it turns out to be a great performer once it's released, great. But I'm not counting on it.

What I'm trying to say is that if they only increased IPC by 10% in comparison to K10.5 then it pretty much sucks. It wouldn't even match Nehalem in IPC and would be surpassing Core 2 by only 5%.

They NEED a big IPC increase.

Otherwise, really, what's the point? If it's only 10% higher IPC it would be slower than Gulftown in multi-threaded apps and only around 10% faster than Sandy Bridge, while losing severely in gaming and audio encoding.

 

[Edrick]

Otherwise, really, what's the point? If it's only 10% higher IPC it would be slower than Gulftown in multi-threaded apps and only around 10% faster than Sandy Bridge, while losing severely in gaming and audio encoding.

Well, it would still be much cheaper than Gulftown, so it may not have to beat it in order to still be a good selling chip.

 

[LOL_Wut_Axel]

Well, it would still be much cheaper than Gulftown, so it may not have to beat it in order to still be a good selling chip.

Yeah, but then only people that use extremely multi-threaded apps would buy it over a 2600K. 10% higher multi-threaded performance coupled with 25-30% slower single-threaded performance is not gonna strike a chord for most people.

The 2600K, even having only four cores, is an extremely fast chip.

 

[AtenRa]

BD may be able to keep up with a 2600K in multithreaded apps, but I highly doubt it will be competitive with a 990X there.

I would reconsider if I were you,

http://www.anandtech.com/bench/Product/203?vs=287

If Phenom II X6 1100T can beat Core i7 2600K in a lot of Multithreaded applications, then a 4-Module 8 Core Bulldozer with more than 50% more transistors than 4-core SandyBridge, a new micro architecture and 2 more Integer Cores than Phenom II X6 will trash that puny(transistor count and multithread performance) Core i7 2600K.

Just a food for thought 😉

 

[996GT2]

I would reconsider if I were you,

http://www.anandtech.com/bench/Product/203?vs=287

If Phenom II X6 1100T can beat Core i7 2600K in a lot of Multithreaded applications, then a 4-Module 8 Core Bulldozer with more than 50% more transistors than 4-core SandyBridge, a new micro architecture and 2 more Integer Cores than Phenom II X6 will trash that puny(transistor count and multithread performance) Core i7 2600K.

Just a food for thought 😉

You might want to look at those benchmarks again. Or get some new reading glasses. Or both.

Hint: some of those graphs say "lower is better" next to them.

What you just showed in that link is the 2600K winning in every single one of those benchmarks while drawing 30W less at load.

 

[Rifter]

I would reconsider if I were you,

http://www.anandtech.com/bench/Product/203?vs=287

If Phenom II X6 1100T can beat Core i7 2600K in a lot of Multithreaded applications, then a 4-Module 8 Core Bulldozer with more than 50% more transistors than 4-core SandyBridge, a new micro architecture and 2 more Integer Cores than Phenom II X6 will trash that puny(transistor count and multithread performance) Core i7 2600K.

Just a food for thought 😉

are you high? your own comparison shows SB raping that 1100T in EVERY benchmark including the heavily threaded apps.

 

[AtenRa]

ahahahaha

sorry guys, got it wrong 😉

 

[LOL_Wut_Axel]

You might want to look at those benchmarks again. Or get some new reading glasses. Or both.

Hint: some of those graphs say "lower is better" next to them.

What you just showed in that link is the 2600K winning in every single one of those benchmarks while drawing 30W less at load.

LOL, owned.

AMD really needs to make a huge jump in performance.

 

[bryanW1995]

Quoted for future reference... namely, when it's released. We'll see how well this holds up. I'll quote this post and post my numbers once released. You have my word on that.

Only problem is, that mythical 8c, 3.6/4.2 BD won't be released until they get all the bugs worked out in 2014.

 

[dma0991]

Intel Core i7 3960X vs Core i7 990X Benchmarks
http://www.donanimhaber.com/islemci...lemcisi-Core-i73960Xin-ilk-test-sonuclari.htm
http://vr-zone.com/articles/intel-core-i7-3960x-vs-core-i7-990x-benchmarks-leaked/13062.html
http://www.fudzilla.com/processors/item/23517-core-i7-3960x-is-about-47-faster-than-core-i7-990x

 

[LOL_Wut_Axel]

Intel Core i7 3960X vs Core i7 990X Benchmarks
http://www.donanimhaber.com/islemci...lemcisi-Core-i73960Xin-ilk-test-sonuclari.htm
http://vr-zone.com/articles/intel-core-i7-3960x-vs-core-i7-990x-benchmarks-leaked/13062.html
http://www.fudzilla.com/processors/item/23517-core-i7-3960x-is-about-47-faster-than-core-i7-990x

Not really surprising at all. Sandy Bridge has 15% higher IPC; therefore, it's 15% faster overall than Nehalem.

Intel claiming much higher performance when due to AVX instructions is true. I find their claims of higher performance coming from higher memory bandwidth due to the additional memory channel to be absolute BS. It makes less than a 1% difference when comparing Bloomfield to Lynnfield. Bandwidth is definitely not an issue.

 

[AdamK47]

Not really surprising at all. Sandy Bridge has 15% higher IPC; therefore, it's 15% faster overall than Nehalem.

Intel claiming much higher performance when due to AVX instructions is true. I find their claims of higher performance coming from higher memory bandwidth due to the additional memory channel to be absolute BS. It makes less than a 1% difference when comparing Bloomfield to Lynnfield. Bandwidth is definitely not an issue.

You people really do dismiss memory bandwith, cache size, PCI-E lanes, and all the other goodies that add up in overall performance. It's a bit myopic to always focus on IPC and ignore the rest of what Intel is doing.

You can bet I'll be posting a slew of benchmarks once I get a 3960X. Too bad nVidia's next gen cards won't be available until next year. I'll have to put up with my old 3-way GTX 580s.

 

[Idontcare]

You people really do dismiss memory bandwith, cache size, PCI-E lanes, and all the other goodies that add up in overall performance. It's a bit myopic to always focus on IPC and ignore the rest of what Intel is doing.

You can bet I'll be posting a slew of benchmarks once I get a 3960X. Too bad nVidia's next gen cards won't be available until next year. I'll have to put up with my old 3-way GTX 580s.

Maybe I missed it along the way, but until llano came along I haven't seen a CPU review in years which depicted much dependence on ram bandwidth for dual-channel versus tri-channel.

Call it myopic if its not true but at least show us something that backs up the assertion. You just did an amazing 2-4-6 core gaming review, take any one of those benchmark configs and pull some of your ram (reduce it from tri-channel to dual-channel) just to rerun one test config to prove or disprove whether or not the loss in bandwidth results in considerable loss in performance.

 

[AdamK47]

Increased memory bandwidth isn't the only thing going for X79. I know that if you were to compare dual to quad you'll be seeing single digit percentage increases. I'm saying it's myopic to not look at everything as a whole that you'll be getting with X79 & SB-E.

 

[AdamK47]

Can't see the forest for the trees.

 

[pantsaregood]

People dismiss memory bandwidth because it doesn't serve any purpose in most cases. Quad-channel DDR3 1600 provides a theoretical 51,200 Mb/s bandwidth. That just isn't something that serves any real purpose on current CPUs in most cases. It won't result in a "5% or so" performance increase in most applications because it simply won't result in ANY performance increase in most applications.

Excess L3 cache similarly serves little purpose. I'd like to see a comparison between a 2500K and 2600K at the same clock speed with HT turned off. You'd see a ridiculously low difference in performance, if any. Six-core SB-E will likely make effective use of more L3 than 2500K/2600K do, but I don't see either requiring a full 15 MB to be functional.

Extra PCI-E lanes, however, may ultimately be quite useful. I've heard complaints of LGA 1155 being a mediocre workstation platform because of the lack of available PCI-E lanes. SB-E corrects that issue. It also eliminates the need for an NF200 chip, which does decrease performance slightly.

 

[RussianSensation]

I'd like to see a comparison between a 2500K and 2600K at the same clock speed with HT turned off.

1-2%

 

[Edrick]

People dismiss memory bandwidth because it doesn't serve any purpose in most cases. Quad-channel DDR3 1600 provides a theoretical 51,200 Mb/s bandwidth. That just isn't something that serves any real purpose on current CPUs in most cases. It won't result in a "5% or so" performance increase in most applications because it simply won't result in ANY performance increase in most applications.

What people do not realize is that when most comparisons were done between dual and tri channel RAM, in most cases, the memory bandwidth was not the bottleneck. Therefore we saw very little performance difference.

Sure on a quad core machine with 4/8 GB of RAM (running off a HDD), 18,000Mb/s to 32,000Mb/s makes little difference. But when we start looking at machines with 6-8 cores, raided SSD's or even PCIe SSDs, with 16+ GB RAM, then I am sure that 51,200Mb/s will come into play much more. And I am not talking about games.

And some people just dismiss memory bandwidth for the simple fact that BD will not have quad channel RAM.

 

[LOL_Wut_Axel]

You people really do dismiss memory bandwith, cache size, PCI-E lanes, and all the other goodies that add up in overall performance. It's a bit myopic to always focus on IPC and ignore the rest of what Intel is doing.

You can bet I'll be posting a slew of benchmarks once I get a 3960X. Too bad nVidia's next gen cards won't be available until next year. I'll have to put up with my old 3-way GTX 580s.

And when was I arguing over the extra features on the SB-E platform? I already mentioned it in the past: the real reason why you should buy SB-E is if you need the more feature-rich platform and/or the extra cores. Anything apart from that and there's simply no reason.

No, memory bandwidth is not an issue. When comparing Lynnfield to Bloomfield it was an academic difference in programs (under 1% performance increase), and given the fact that Sandy Bridge has even more memory bandwidth than Lynnfield I expect this fact to be accentuated.

Extra cache size, again, will make little to no difference. This is the case going from the 2500K's 6MB L3 cache to the 2600K's 8MB. Just because Intel loves to show off huge numbers in SB-E doesn't mean anything in the real-world. They put them in for marketing purposes, nothing else. It's 15% faster overall than Bloomfield in IPC.