Intel Skylake / Kaby Lake

[Sweepr]

Looks like Skylake-S GT2 will close the gap with Kaveri/Godavari as far as iGPU is concerned. If the same GT2 iGPU is present in lower-end chips then AMD's iGPU advantage is almost gone where it hurts the most, the $80-150 niche. Skylake GT4e will add the final insult next year, outclassing Bristol Ridge by a larger margin than Broadwell GT3e does compared to Kaveri/Godavari. Remember this might not be final drivers.

Ps: Can anyone register there? There might be other benchmarks in the next pages.

 

[Fjodor2001]

So the CPU performance of Skylake 6700K is more or less the same as the two generation old Haswell 4790K (except in iGPU bound tests)?

Even a regression in some benchmarks. And all this while bumping the TDP from 88 to 95 W.

Can this really be true!? Maybe it's because they have prioritized the low power mobile segment for Skylake?

 

[mikk]

+35% in 3dmark11, Firestrike is much better than previous leaks. Although I don't think the gap will close with only 24 EUs and 1150 Mhz. I'm more interested in gaming benchmarks.


Btw, this test was a fake given that CB R15 was much better there: http://www.pcfrm.com/intel-i7-6700k-vs-i7-4790k/

 

[ShintaiDK]

So the CPU performance of Skylake 6700K is more or less the same as the two generation old Haswell 4790K (except in iGPU bound tests)?

Even a regression in some benchmarks. And all this while bumping the TDP from 88 to 95 W.

Can this really be true!? Maybe it's because they have prioritized the low power mobile segment for Skylake?

Remember the Haswell uses DDR3 2400 at very tight timings. And since its Asus, it may also use MCE.

Skylake looks exactly what is expected.

 

[coercitiv]

Please don't come up with Broadwell-U NotebookCheck voltage curves, I'm not in the mood for laughing.

With all due respect, I did not see you laughing when Broadwell reviews did not overclock past 4.2Ghz (or didn't even try), nor when actual high overclocks showed a voltage delta over Haswell exactly as predicted by those BW-U voltage curves.

I have a lot of respect for your constant effort on these CPU threads, and I share your optimism about Skylake, but let's not ignore facts because of a forum brawl.

 

[Sweepr]

So the CPU performance of Skylake 6700K is more or less the same as the two generation old Haswell 4790K (except in iGPU bound tests)?

Even a regression in some benchmarks. And all this while bumping the TDP from 88 to 95 W.

Can this really be true!? Maybe it's because they have prioritized the low power mobile segment for Skylake?

Actually total system power went down by 50W.

I'm sure the voltage experts and those who interpret TDP = power consumption might have a good explanation.
Note that in 3DMark's CPU tests it's >10% faster than Core i7 4790K so it seems to vary according to the test you pick, we don't know where the overall improvement will end but it looks like it will vary depending on your main system use.

 

[Sweepr]

With all due respect, I did not see you laughing when Broadwell reviews did not overclock past 4.2Ghz (or didn't even try), nor when actual high overclocks showed a voltage delta over Haswell exactly as predicted by those BW-U voltage curves.

You didn't because actual reviews pointed out to better performance per watt from Broadwell-K compared to Devil's Canyon, despite the constant efforts from certain people to point in the opposite direction at every single Intel launch. Now I'm not saying I expect miracles from Skylake-S but we don't know what flavour of 14nm they are using, neither the kind of tweaks they might have done. Sorry but extrapolating data from Broadwell-U NotebookCheck notebook reviews to call these desktop Skylake-S leaks fake and Intel biased is laughable at best.

 

[coercitiv]

Sorry but extrapolating data from Broadwell-U NotebookCheck reviews to call these Skylake-S leaks fake and Intel biased is laughable at best. I'm sorry if you disagree.

Who used BW-U data to extrapolate Skylake voltage?

PS: if anything, extrapolating from BW-U data and comparing with all the Skylake-S leaks is the main reason for optimism.

 

[Sweepr]

Who used BW-U data to extrapolate Skylake voltage?

I'm assuming that's where the claim of recent leaks (5.2GHz bootable at 1.35V @ air cooling) being fake might come from, otherwise maybe it's a plain lie from someone known to be an Intel hater?

 

[Abwx]

Other than the fact that if true those voltages will put the nail in the coffin of your reputation what kind of evidence do you have to prove them wrong? Please don't come up with Broadwell-U NotebookCheck voltage curves, I'm not in the mood for laughing.

Look at the voltages of DT BDWs, and in actual reviews that you indeed linked in this thread....

http://www.hardware.fr/focus/112/core-i5-5675c-broadwell-cote-cpu-test.html

The overcklocking part is the one of interest for our subject..

 

[Fjodor2001]

Actually total system power went down by 50W.

Ok, but do we know if the exact same system setup was used in that test except for the CPUs? PSU, motherboard etc influence power consumption a lot.

 

[Fjodor2001]

Skylake looks exactly what is expected.

Well in that case I think many people might as well just get the two generation old Haswell 4790K instead. Especially those that do not intend to use the iGPU anyway.

 

[Sweepr]

Ok, but do we know if the exact same system setup was used in that test except for the CPUs? PSU, motherboard etc influence power consumption a lot.

I agree but other than different MBs and memory kits looks like everything else is the same. 50W @ load is significant. Idle is only 13W lower.

 

[Fjodor2001]

50W @ load is significant.

Yes, in fact it's actually far too much to expect Skylake CPU + chipset to be able to save that much power. That's what made me wonder if there's some other explanation, like other parts of the system being responsible for most of the power savings. But if the Skylake CPU + chipset really can reduce system power consumption by 50 W that'll be good, and I'll be impressed.

 

[ShintaiDK]

Well in that case I think many people might as well just get the two generation old Haswell 4790K instead. Especially those that do not intend to use the iGPU anyway.

I dont think so. Skylake is faster and uses less power.

 

[Arachnotronic]

Well in that case I think many people might as well just get the two generation old Haswell 4790K instead. Especially those that do not intend to use the iGPU anyway.

Why? You get a worse platform and a slower CPU this way.

 

[LTC8K6]

Well in that case I think many people might as well just get the two generation old Haswell 4790K instead. Especially those that do not intend to use the iGPU anyway.

4790K can barely run at stock speed on air, though.

If I could get 4790K performance without using a giant cooler, I'd take it.

 

[Dave2150]

:thumbsup:

Based on how far behind i7 4790K @ 4.7ghz is vs. 5930K @ 4.7ghz with dual 980Tis @ 1300mhz, I doubt even a 5Ghz Skylake-K quad will keep up. Given how long Intel CPUs last nowadays, for anyone contemplating flagship $500+ cards in pairs (and if that's their upgrade path for 2-3 generations forward), it's going to be well worth it to wait for the Skylake-E 5820K successor.

http://forums.anandtech.com/showthread.php?t=2440000

This is especially true this time since the new 10nm micro-architecture from Intel is now only slated for 2018 which means for the next 2.5 years or so, there will hardly be any tangible increases in IPC. With DX12 and Windows 10 though, we should see more games taking advantage of more cores which will allow X99 and its successor to shine!

Those results are questionable - the 4790k rig's 980ti's are using HALF the VRAM of the 5930k's 980ti's. Something wasn't setup right for sure.

 

[Dave2150]

4790K can barely run at stock speed on air, though.

If I could get 4790K performance without using a giant cooler, I'd take it.

Garbage. I've setup several 4790k systems using the stock HSF, in a well ventilated case, and during full gaming load, or full benchmarking load (non avx) temperatures are 85°C at most, at turbo boost clocks.

I did replace the stock thermal material on the heatsink with mx2, however.

 

[Arachnotronic]

4790K can barely run at stock speed on air, though.

If I could get 4790K performance without using a giant cooler, I'd take it.

Depends on the kind of "air" you're talking about. On the stock cooler, my 4790K did run hot and the cooler itself was very loud.

However, with a Hyper 212+ EVO, my 4790K is humming along nicely at good temperatures, relatively low noise levels, and a solid overclock.

 

[CHADBOGA]

Why? You get a worse platform and a slower CPU this way.

Some people don't want to let go of the AMD Experience. :biggrin:

 

[Fjodor2001]

I dont think so. Skylake is faster and uses less power.

Skylake is only marginally faster, and only in some of the benchmarks. Haswell is actually faster in some benchmarks, which is quite remarkable for a two generation old CPU.

So I guess it depends on what you'll be able to get the 4790K vs 6700K for. They are so close in CPU performance that price will be decisive.

I'd probably be monitoring the price of both and see if some good deal pops up on either of them. But to be honest, it feels like none of them are worth upgrading to from my current 3570K.

 

[Arachnotronic]

Some people don't want to let go of the AMD Experience. :biggrin:

:biggrin:

 

[Fjodor2001]

:biggrin:

I'm currently on an "Intel experience" (3570K), so go ahead and laugh at that all you want.

 

[ShintaiDK]

Some people don't want to let go of the AMD Experience. :biggrin:

LOL :awe: