Skylake SKU

[mikk]

Pretty much confirms CPU-World.

 

[*NixUser]

I wonder why no 95w sku with edram? Maybe because they expect consumers who buy top end won't use it at all?

 

[positivedoppler]

its all good. slap a water cooler and see how far it oc like the good old day.

 

[Ajay]

I wonder why no 95w sku with edram? Maybe because they expect consumers who buy top end won't use it at all?

Heck, I'm just glad that there is a 95W SKU. That should mean a nice boost in performance over Haswell for the Skylake Tock. Or, it could mean that AVX512 will be available on some at least the top i7 bins.


mikk - where is this slide from, there's no logo?

 

[jpiniero]

I wonder why no 95w sku with edram? Maybe because they expect consumers who buy top end won't use it at all?

I imagine OEMs don't want the extra cost. The K model will be 4+4e.

 

[III-V]

Heck, I'm just glad that there is a 95W SKU. That should mean a nice boost in performance over Haswell for the Skylake Tock. Or, it could mean that AVX512 will be available on some at least the top i7 bins.

I would not put any faith in there actually being a 95W SKU at launch. Any performance and power numbers are worthless at this point.

 

[Ajay]

Any performance and power numbers are worthless at this point.

They are not worthless if that's an official Intel slide, if not, they yeah, you're right.

 

[IntelUser2000]

I wonder why no 95w sku with edram? Maybe because they expect consumers who buy top end won't use it at all?

That's probably coming in 2H 2016, for a successor to Broadwell-K.

I would not put any faith in there actually being a 95W SKU at launch.

Core i7 4771 is a 84W SKU that's locked. It's not hard to see 95W Skylake is a DIRECT successor to that.

 

[Fjodor2001]

4+4e < 4+2 max TDP-wise. 65 W vs 95 W. Interesting... :hmm: Will the 4+4e part perhaps be clocked lower, or what could explain it?

 

[mikk]

I wonder why no 95w sku with edram? Maybe because they expect consumers who buy top end won't use it at all?

I wouldn't read too much into the TDP at this point, it can be placeholder and might change. In the past that was the case for desktop. Even Intel can't be sure until later advanced ES models are available.

mikk - where is this slide from, there's no logo?

I don't think it's a good idea to reveal it because Intel tries to hunt them down, so I cut the logo.

Btw CPU-world claimed that for Xeon E3 GT4 SKU is coming late 2015 or early 2016. Xeon E3 are locked usually though. And it's a BGA (SKL-H)...

At the end of 2015 or at the beginning of 2016, Intel will introduce BGA versions of Xeon E3-1200 v5 chips with Premium graphics, featuring GT4 GPU and 128 MB eDRAM. The processors will have 4 CPU cores, and support only DDR4 memory.

http://www.cpu-world.com/news_2014/...s_of_Intel_Xeon_E3-1200_v4_v5_processors.html

 

[witeken]

Core i7 4771 is a 84W SKU that's locked. It's not hard to see 95W Skylake is a DIRECT successor to that.

But you forgot that there's a shrink between those 2, which would come out as 65W. And without more EUs it's hard to imagine why there would have to be a SKU with 1.5x the TDP.


Btw, what's the source?

 

[mikk]

Haswell and Ivy Bridge @desktop were all classified with a 95W TPD in early Roadmaps as well, look how it turned out in the end.

 

[NTMBK]

Hopefully there is a 2+4e SKU- that would be great for gaming HTPCs.

 

[ShintaiDK]

Dont put too much into TDP, since its standard family values.

SB/IB/HW are "95W" as well. Even tho they was 95/77/85W.

 

[ShintaiDK]

Hopefully there is a 2+4e SKU- that would be great for gaming HTPCs.

Seems the 4+4e is the target for HTPCs, SFF, etc. Tho I wouldnt be surprised to see the 2+3e in NUCs instead.

 

[meloz]

I am struggling. Can someone explain what 4+2 and 2+2 refer to? Thread and core count, respectively?

If so, why does the 4 core (and 4 thread?) variant with eDRAM and stronger iGPU only goes upto 65 watt whereas the 'lesser' 4+2 gets a 65 watt variant and a 95 watt variant?

 

[ShintaiDK]

I am struggling. Can someone explain what 4+2 and 2+2 refer to? Thread and core count, respectively?

If so, why does the 4 core (and 4 thread?) variant with eDRAM and stronger iGPU only goes upto 65 watt whereas the 'lesser' 4+2 gets a 65 watt variant and a 95 watt variant?

Its cores and graphics.

2+2 means dualcore with GT2.
4+4e means quadcore with GT4 and eDRAM.

4+4e may be targetted to other form factors that cant use 95W TDP.

Also remember, 95W etc is just a family type TDP. The final TDP can be anywhere between 66 and 95W for example. SB was 95W, IB 77W, HW 84/88W. All was in the 95W category.

 

[pw257008]

first number is cpu cores, second is gpu units

edit: beat me to it.

 

[meloz]

Thanks guys!

And like Ajay I am just glad there is a 95 watt family type. Seems Intel have not forgotten us desktop users after all.

 

[Fjodor2001]

Dont put too much into TDP, since its standard family values.

SB/IB/HW are "95W" as well. Even tho they was 95/77/85W.

I think it's not the specific TDP number that is interesting, but that 4+4e has lower TDP than 4+2. All TDP numbers taken from the same slide.

So it might turn out that 95 W will actually be 75 W or whatever in the end. But Intel still considers the 4+4e part to be a higher TDP part than the 4+2 part. That is kind of odd, unless the 4+4e part is intended to be clocked lower.

 

[pw257008]

I think it's not the specific TDP number that is interesting, but that 4+4e has lower TDP than 4+2. All TDP numbers taken from the same slide.

So it might turn out that 95 W will actually be 75 W or whatever in the end. But Intel still considers the 4+4e part to be a higher TDP part than the 4+2 part. That is kind of odd, unless the 4+4e part is intended to be clocked lower.

If they want to protect the Broadwell K's, it makes sense for them to clock the Skylake 4e's lower

 

[jpiniero]

If they want to protect the Broadwell K's, it makes sense for them to clock the Skylake 4e's lower

Skylake 4+4e isn't coming out until 2016. It's not really competition for Broadwell-K.

 

[witeken]

I'm still hoping the number of EUs of GT4 will be closer to 128 than 72. Exponential trends like Moore's Law don't work by adding a new (linear) slice every generation!

Cannonlake should allow a 256EU SKU instead of 4 24EU slices (96EUs), which would be 5 (Gen7.5) TFLOPS.

 

[III-V]

They are not worthless if that's an official Intel slide, if not, they yeah, you're right.

Even so, the official slide on Broadwell-E we've seen has it listed as a 140W part, yet that is essentially nonsensical, as that is the same TDP as Haswell-E. The move to 14nm alone would reduce the TDP significantly, and core counts aren't changing. Basically, what I'm saying is that even Intel hasn't nailed down the final specifications for Skylake, or Broadwell-E for that matter.

 

[Fjodor2001]

I'm still hoping the number of EUs of GT4 will be closer to 128 than 72. Exponential trends like Moore's Law don't work by adding a new (linear) slice every generation!

Cannonlake should allow a 256EU SKU instead of 4 24EU slices (96EUs), which would be 5 (Gen7.5) TFLOPS.

I agree. And also, mainstream Broadwell should have 16 CPU cores and Cannonlake 32, since Sandy Bridge had 4.

Sure, the cores have gotten a bit fatter due to uArch changes, but it should at least be 8-12 CPU cores for Broadwell and 16-20 for Cannonlake.