Sandy Bridge-E Details Revealed

[pantsaregood]

SB-E isn't even a die shrink

Per-core (per-clock) performance should be very, VERY similar (closer than say Thuban and Llano). Multithreaded performance should increase because of more cores, and singlethread performance should increase due to higher clock speeds.

Of course, if you buy a 'K' series you can get some of those higher clock speeds in an easier and much cheaper manner...

I was speaking of IB.

 

[Arachnotronic]

So those of you chanting that SNB-E = SNB clock-for-clock, core-for-core...do you not understand the magic that is massive memory bandwidth for workstation applications? Also, huge caches are great for workstation environment.

Although I think the memory bandwidth advantage will be more evident on the hex-cores than on the quad-cores, since there's 50% more mouths to feed on the hex-core...still, anybody who thinks these factors are not important obviously uses their PC for just gaming (and I do! Hence Lynnfield! )

 

[ed29a]

So those of you chanting that SNB-E = SNB clock-for-clock, core-for-core...do you not understand the magic that is massive memory bandwidth for workstation applications? Also, huge caches are great for workstation environment.

Caches might give performance boost for sure, but dual channel vs tri channel tests have shown very minor gains for triple channel over dual channel. Quad channel might get another minor boost, but nothing major for games for instance, definitely not "massive".

 

[Arachnotronic]

Sure, but for games, I would tell people to just go buy a 2600K now and be done with it.

 

[Edrick]

Caches might give performance boost for sure, but dual channel vs tri channel tests have shown very minor gains for triple channel over dual channel. Quad channel might get another minor boost, but nothing major for games for instance, definitely not "massive".

The issue here is a quad core CPUs, like the i7 920 for example, may only need dual channel memory to keep it feed. So the increase to tripple channel would only show a very minimal improvement.

However, when we start seeing 8 core SB-E cpus and even Haswell, it will be safe to assume that dual channel memory will not keep those fed, especially when the new instructions are used. So I expect to see a nice jump in performace using a quad channel IMC.

And granted, like Intel17 said, certain users (workstation) will see larger performance increases than gamers.

Also, Intel once stated (I will look for the article), that it wants to keep the minimum memory channels set to 1 channel per 2HT cores. So thats dual channel for 4 cores and quad for 8 cores (tripple for 6 cores). So if this holds true, we will see more than quad channel memory eventually. And we all know Intel is not going to invest the money if there is no increased performance to be gained.

 

[exar333]

The issue here is a quad core CPUs, like the i7 920 for example, may only need dual channel memory to keep it feed. So the increase to tripple channel would only show a very minimal improvement.

However, when we start seeing 8 core SB-E cpus and even Haswell, it will be safe to assume that dual channel memory will not keep those fed, especially when the new instructions are used. So I expect to see a nice jump in performace using a quad channel IMC.

And granted, like Intel17 said, certain users (workstation) will see larger performance increases than gamers.

Also, Intel once stated (I will look for the article), that it wants to keep the minimum memory channels set to 1 channel per 2HT cores. So thats dual channel for 4 cores and quad for 8 cores (tripple for 6 cores). So if this holds true, we will see more than quad channel memory eventually.

The extra memory channels are very welcome for those utilizing multiple VMs.

Dual-channel - 4C
Triple-channel - 6C
Quad-channel - 8C

With memory so cheap, a quad-channel 4x4GB system is VERY cheap to build. Even if you start with a 4C or 6C CPU, its ready for a 8C drop-in without any other changes.

 

[Edrick]

With memory so cheap, a quad-channel 4x4GB system is VERY cheap to build. Even if you start with a 4C or 6C CPU, its ready for a 8C drop-in without any other changes.

That is exactly what I will be doing.

 

[greenhawk]

I have no idea why the "X58 must use triple-channel!" dumb myth came about

I would not call it a myth, but I would call someone that did it stupid as you pay extra for the cpu/motherboard, and disable it's main feature. Might as well have gone for a i7-8xx and save a few hundred.

 

[LOL_Wut_Axel]

Who am I to argue with an expert.

The thing is, nobody knows. You can't claim to know with 100% certainty what the performance will be like in a wide variety of tests with varied hardware configurations. You could be right on some and way off on others. As a Software Load Tester, the number one thing I know is to never assume with 100% certainty what kind of results to expect.

Alright, then. I guess it's settled: Intel will pull much more performance from SB-E in comparison to SB out of thin air.

Pretty confident I'm 99% right if you look at how CPU architectures work.