• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."
  • Community Question: What makes a good motherboard?

Intel Skylake / Kaby Lake

Page 220 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Edrick

Golden Member
Feb 18, 2010
1,909
196
106
Anyone know when the Xeon E3s will be available is the US? And the new C232 chipset mother lads by Gigabyte? Want to try the combination but have been unable to find them anywhere.
 

zir_blazer

Senior member
Jun 6, 2013
939
85
91
Anyone know when the Xeon E3s will be available is the US? And the new C232 chipset mother lads by Gigabyte? Want to try the combination but have been unable to find them anywhere.
Newegg seems to have a third party seller with the E3 1230V5. The same third party seller has the E3 1245V5 at Amazon for 320 U$D, rather high compared to the 284 U$D Intel MSRP of Intel Ark (But that's the Tray version, Box should be 290$ like the 1246V3 it replaces).

I don't recall mentions regarding C232 features. Its obviously going to be crippled compared to C236, but no actual info on what. For reference, the big difference between C222/224 and C226 for Haswell/Broadwell generation, is that only the C226 supports using the integrated GPU, the others are like the Sandy Bridge P67 Chipset.
 

witeken

Diamond Member
Dec 25, 2013
3,876
154
106
For your comparing pleasure:

SRAM cell size comparison

Company / regular density / high density

Intel 22nm = 0.108µm² / 0.092µm²
TSMC 16nm = 0.07µm² (high density)
Samsung 14nm = 0.080µm² / 0.064µm²
Intel 14nm = 0.0588µm² / 0.0500µm²
Samsung 10nm = 0.049µm² / 0.040µm²
TSMC 10nm = 0,0364µm²F (high density)
Intel 10nm = 0.0312µm²F / 0,0272µm²F (with scaling of 1.84x like 22->14 high density)
TSMC 7nm = 0.0207µm²F (high density -43% // TSMC reported minus 40-45%)

F = Forecast
 

Ajay

Diamond Member
Jan 8, 2001
7,584
2,738
136
For your comparing pleasure:

SRAM cell size comparison

Company / regular density / high density

Intel 22nm = 0.108µm² / 0.092µm²
TSMC 16nm = 0.07µm² (high density)
Which TSMC is that 16FF or 16FF+?

Thank for the info!
 

coercitiv

Diamond Member
Jan 24, 2014
4,036
4,642
136
What are the differences between Intel's Speedshift and AMD's AVFS?
My understanding is the following:

  • Speedshift allows the CPU to make quick large jumps between frequency states in order to maximize efficiency in burst load scenarios. (keep CPU in low power states as much as possible)
  • AVFS allows the CPU to make extremely fast minute adjustments to it's frequency in order to maximize efficiency under sustained load scenarios. (keep operating voltage as low as possible)
They complement each other.
 

einhoernchen

Junior Member
Oct 19, 2015
7
1
0
According to Mindfactory (one of the biggest sellers in Germany) 6700K availability won't improve in the next weeks. They don't expect new CPUs until Dec 29th. That would be a real bummer. Prices are currently 400-500 EUR in Europe (~ 400-500 USD).
 

Dresdenboy

Golden Member
Jul 28, 2003
1,730
554
136
citavia.blog.de
My understanding is the following:

  • Speedshift allows the CPU to make quick large jumps between frequency states in order to maximize efficiency in burst load scenarios. (keep CPU in low power states as much as possible)
  • AVFS allows the CPU to make extremely fast minute adjustments to it's frequency in order to maximize efficiency under sustained load scenarios. (keep operating voltage as low as possible)
They complement each other.
Thanks. I'm checking AMD's patents. I remember seeing a related diagram there.

BTW, looking at the old articles, Charlie from S|A mixed AVFS with VAO. :)
 

t0mt0m

Member
Apr 21, 2015
35
2
36


LGA 65W Skylake-S with GT4e Iris Pro iGPU lives. 35W version under evaluation.
Is it normal to have such a time spread between different models coming out? 65W GT4e not for another 13-15 months? Wonder when the 35W will come out.
 

Sweepr

Diamond Member
May 12, 2006
5,151
1,127
131
Is it normal to have such a time spread between different models coming out? 65W GT4e not for another 13-15 months? Wonder when the 35W will come out.
Looks like desktop LGA Iris Pro will lag one generation behind regular GT2 SKUs. Do note that BGA Skylake GT4e is expectec to arrive a lot sooner though, probably Q1-2016 (mobile and Xeon E3 v5).

Talking about Iris, some lucky customers just got their Surface Pro 4 based on Core i7 6650U. Hopefully we will have some benchmarks results any time now.



mikk said:
Thanks mikk, nice improvement. Any idea about Kaby Lake's iGPU specs yet?
 
Last edited:

Sweepr

Diamond Member
May 12, 2006
5,151
1,127
131
Core i7-6650U (15W Skylake-U GT3e)
3DMark 11 Performance Preset score: ~2600



~83.5% the graphics performance of 65W Broadwell-K's Iris Pro Graphics 6200 inside a thin tablet/convertible.

www.3dmark.com/3dm11/10550899

64% faster than Core i5-6300U (15W Skylake-U GT2). Iris Graphics 540 is a little beast.
Compared to Carrizo @ AMD reference platform:

 
Last edited:

mikk

Platinum Member
May 15, 2012
2,955
767
136
60%+ on the same power envelope is pretty good, much better than expected from me.

Sky Diver and 3dmark11 graphics score are ~30% faster compared to my i7-6700k @HD530 with DDR3-3000.
 
Aug 11, 2008
10,457
641
126
Only 30% better at 15W than carizzo. Am I reading that correctly?

With e-dram I would have expected it to be further ahead. That would make skylake-U GT2 about 30% slower than Carizzo? I thought with Skylake they perhaps had caught up to AMD with the mainstream graphics solution, at least in the low power envelope.

Problem is, the e-dram only comes on top of the line chip in a very expensive platform. Now if they would make it standard across the line, I would be impressed.
 

ShintaiDK

Lifer
Apr 22, 2012
20,395
128
106
Problem is, the e-dram only comes on top of the line chip in a very expensive platform. Now if they would make it standard across the line, I would be impressed.
EDRAM isn't top of the line anymore. Its essentially hitting i5 mainstream.

There is roughly 100$ difference between the i3 6100U and i5 6260U for NUCs. Even less for other products. And that's 100$ for EDRAM, GT3, higher clocks etc.
 

rootheday3

Member
Sep 5, 2013
44
0
0
...That would make skylake-U GT2 about 30% slower than Carizzo? ...
Take a look at the bar graph immediately above - see the 3DMark11 Performance score for Surface Pro 4 (1584)? That is SKL U GT2.
So ~1600 for SKL GT2 vs ~2000 for Carizzo FX8800P 15W in the table (SKL GT2 20% slower) and SKL GT2 on par with the A10-8600P Carrizo 15W.

Now consider that SKL GT2 = 24EUs = ~192 sp (Intel EUs are SIMD8). The Carrizo FX8800P is 512sp and the A10-8600P is 384 sp.

I would say the perf/SP for SKL GT2 is quite good - it is likely either TDP limited (turbo up on narrow design hits P ~ f*v^2; f~v in turbo regime; P ~ v^3 ... therefore, design which is narrow but fast can get decent perf but isn't as efficient as running wide and slow) OR else frequency limited (hitting gpu Fmax).

Given the scaling of SKL GT3e (48 EU = 384sp), where it can run wider but slower, the perf/watt and perf/sp for SKL gpus actually looks really good.
 
Mar 10, 2006
11,719
2,003
126
Take a look at the bar graph immediately above - see the 3DMark11 Performance score for Surface Pro 4 (1584)? That is SKL U GT2.
So ~1600 for SKL GT2 vs ~2000 for Carizzo FX8800P 15W in the table (SKL GT2 20% slower) and SKL GT2 on par with the A10-8600P Carrizo 15W.

Now consider that SKL GT2 = 24EUs = ~192 sp (Intel EUs are SIMD8). The Carrizo FX8800P is 512sp and the A10-8600P is 384 sp.

I would say the perf/SP for SKL GT2 is quite good - it is likely either TDP limited (turbo up on narrow design hits P ~ f*v^2; f~v in turbo regime; P ~ v^3 ... therefore, design which is narrow but fast can get decent perf but isn't as efficient as running wide and slow) OR else frequency limited (hitting gpu Fmax).

Given the scaling of SKL GT3e (48 EU = 384sp), where it can run wider but slower, the perf/watt and perf/sp for SKL gpus actually looks really good.
Intel has done a good job improving its iGPUs. I know you work on the Intel GPU team, so props to the folks on the team. Can't wait to see what's in the pipe for Gen 9.5 and Gen. 10 GPUs -- and beyond.
 
Aug 11, 2008
10,457
641
126
Take a look at the bar graph immediately above - see the 3DMark11 Performance score for Surface Pro 4 (1584)? That is SKL U GT2.
So ~1600 for SKL GT2 vs ~2000 for Carizzo FX8800P 15W in the table (SKL GT2 20% slower) and SKL GT2 on par with the A10-8600P Carrizo 15W.

Now consider that SKL GT2 = 24EUs = ~192 sp (Intel EUs are SIMD8). The Carrizo FX8800P is 512sp and the A10-8600P is 384 sp.

I would say the perf/SP for SKL GT2 is quite good - it is likely either TDP limited (turbo up on narrow design hits P ~ f*v^2; f~v in turbo regime; P ~ v^3 ... therefore, design which is narrow but fast can get decent perf but isn't as efficient as running wide and slow) OR else frequency limited (hitting gpu Fmax).

Given the scaling of SKL GT3e (48 EU = 384sp), where it can run wider but slower, the perf/watt and perf/sp for SKL gpus actually looks really good.
But still, Skylake is on 14nm and carizzo is 28nm. So in a TDP limited scenario, seems like intel should be able to pull ahead without having to resort to edram. OTOH, we dont really know the true power consumption of either, and AMD reference platforms generally perform quite well compared to real world devices.
 

ASK THE COMMUNITY