Intel Skylake / Kaby Lake

[cytg111]

End of the road?
Thanks for playing?
Fun while it lasted?
All good things?

Death of pc performance scaling? At least it will give the consoles a chance to catch up so we wont have to bitch about "crappy console ports" anymore.

Meh.

 

[Sweepr]

so can we expect the usual 15%?
nice opengl score

Just looking at Sandra and Cinebench looks like the improvement is going to vary a lot from benchmark to benchmark. I'm confident Skylake will be a solid tock though.

DXVAChecker detection Information: significant support within HEVC_VLD_Main and HEVC_VLD_Main10, meaning can support HEVC / H.265 Format 8 / 10bit video decode acceleration.

 

[NTMBK]

"Less" than i7 5775C?! Salt mandatory.

5775C has an enormous L4 cache, that's going to help in certain CPU workloads.

 

[TeknoBug]

Oh we'll still get crappy console ports like *cough* Arkham Knight.

5775C might be on my shopping list one day.

 

[Sweepr]

As a reminder, from the original source:

But even Skylake launch plan has been imminent, the hardware that can be achieved at the moment, drivers, software, etc., are not the final official version, including the protagonist Z170-Claymore motherboard versa. Therefore, we believe can only spy or two, is not the time to make absolutely commented, formal and more complete testing and comment, left various manufacturers then the arrival of the official product. Let us wait and see!

 

[coercitiv]

5775C has an enormous L4 cache, that's going to help in certain CPU workloads.

I know, but why should the customer care? If that eDRAM cache brings about more performance than a new arch, why are we still seeing DT i7 CPUs without one?

 

[Madpacket]

If these results are true this is hilarious. Anyone who has a 2500K or newer will be like "well, maybe next node".

 

[NTMBK]

I know, but why should the customer care? If that eDRAM cache brings about more performance than a new arch, why are we still seeing DT i7 CPUs without one?

Because a big hunk of extra silicon increases costs significantly. There's a reason why the 5775C costs so much.

 

[SOFTengCOMPelec]

Surely it's nonsense. I thought going from Haswell to Broadwell (equivalent model chips), gave about a 5% IPC improvement ?
So Skylake should give at least the same (+5%), if not more ?
Unless there are terrible issues with the Skylakes, at the moment.

Maybe the benchmarks are from an early ES (Engineering Sample) chip and/or it has been performance limited (on purpose) by Intel, to stop people benchmarking it before launch.

 

[Skylake]

Just looking at Sandra and Cinebench looks like the improvement is going to vary a lot from benchmark to benchmark. I'm confident Skylake will be a solid tock though.

I see..
the cine bench result looks fishy, no way it's slower than broadwell.. software not final?

but sisoftware looks good

 

[Phynaz]

There's a Skylake thread, you know? Nice try though.

This. Reported for merge.

 

[coercitiv]

Because a big hunk of extra silicon increases costs significantly. There's a reason why the 5775C costs so much.

The eDRAM cache offers a performance improvement bigger than a new arch, or alternatively power savings equal to a new node, and we consider it expensive?

What happens if 6700k uses more power than 5775C to achieve more or less the same performance? As I said, I'll take it with a grain of salt and hope for better figures.

 

[SAAA]

Sysoft is right were Skylake should be, that Cinebench score is worrisome instead.
I hope it's just wrong and there is "some" improvement clock/clock over Haswell because better graphics (and power consumption?) aren't exactly my priority now...

 

[Sweepr]

Sysoft is right were Skylake should be, that Cinebench score is worrisome instead.
I hope it's just wrong and there is "some" improvement clock/clock over Haswell because better graphics (and power consumption?) aren't exactly my priority now...

I agree. Software to blame?

But even Skylake launch plan has been imminent, the hardware that can be achieved at the moment, drivers, software, etc., are not the final official version, including the protagonist Z170-Claymore motherboard versa. Therefore, we believe can only spy or two, is not the time to make absolutely commented, formal and more complete testing and comment, left various manufacturers then the arrival of the official product. Let us wait and see!

Some people are saying Broadwell-K brings a larger IPC boost due to eDRAM, that might be true in some applications but not Cinebench R15.

Broadwell's advantage (@ CB R15) without eDRAM is less than ~3% (Core i3 5005U vs Core i5 4005U, no Turbo, normalized clocks).
Broadwell's advantage with eDRAM is less than 4% (Core i5 5675C vs Core i5 4690, normalized single-core Turbo clocks).

Cinebench 11.5 usually scales better with faster cores, too bad they didn't test it too.
Looking at Broadwell it wouldn't surprise me if Skylake-S (retail MBs and software) is less than 10% faster per clock in this particular benchmark, and we have to consider the Turbo clocks difference.

 

[ShintaiDK]

Because a big hunk of extra silicon increases costs significantly. There's a reason why the 5775C costs so much.

The eDRAM cost is very small. It was estimated to be 3$.

 

[mikk]

They did use an old graphics driver for Skylake, that's why the final naming scheme HD Graphics 530 wasn't in use. 3dmark looks rather poor, Cinebench much better. Overall a poor benchmark selection with annoying chinese letters. It's also confusing they say DDR3-1600 has been used but GPUz says DDR-2133.

 

[IndyColtsFan]

If these results are true this is hilarious. Anyone who has a 2500K or newer will be like "well, maybe next node".

Unfortunately that's the boat I'm in with my overclocked i7-2600K. I was initially estimating an average improvement in the 30%-40% at the same clock speed but it doesn't look like that will happen if these benches are accurate. True, in certain tasks, it will be much faster than the 2600K but those tasks probably aren't important enough to me at this time to upgrade. That likely means I'll be using my 2600K for another year. I love it, but I am itching for new toys and an upgrade.

 

[frozentundra123456]

Agreed, we should take these results with a grain of salt. But we are starting to get to the point where the "preponderance of evidence" as they say is starting to add up that the cpu performance improvement for the desktop will be mediocre at best. Since this is two generations, anything less than 10% will be disappointing to say the least.

The only thing I see hopeful is apparently better power consumption and gpu (we have heard that before) so maybe it will be a nice advance for mobile.

I know some will disagree (Shintai), but Intel *really* needs to bring out a hex core mainstream cpu with the latest architecture, especially if they cant seem to get really any significant performance increase per core over 2 generations. If they really want to drive enthusiasts to upgrade, that would be a lot more effective than a few percent ipc and no clockspeed increases, along with a better igp which nobody really cares about in the desktop.

 

[SAAA]

There's still the variable of overclocking though, if it has Haswell/+ IPC and reaches 5GHz easily then everyone will be happy.

That or 6 cores mainstream in the not so distant future, heck they are so small now...

EDIT:

Could that 133mm2 be the actual die size or just some random value from Broadwell parts?

 

[richierich1212]

There's still the variable of overclocking tough, if it has Haswell/+ IPC and reaches 5GHz easily then everyone will be happy. That or 6 cores mainstream in the not so distant future, heck they are so small now...

Haswell had a tough time reaching 5GHz...

 

[Sweepr]

There's still the variable of overclocking though, if it has Haswell/+ IPC and reaches 5GHz easily then everyone will be happy.

That or 6 cores mainstream in the not so distant future, heck they are so small now...

EDIT:

Could that 133mm2 be the actual die size or just some random value from Broadwell parts?

I think that's Broadwell 2C+GT3 die size, maybe CPU-Z is reading Skylake wrong.

 

[jpiniero]

but Intel *really* needs to bring out a hex core mainstream cpu with the latest architecture, especially if they cant seem to get really any significant performance increase per core over 2 generations.

That's not happening. You'll have to settle for the 5820K and friends if that's what you want.

 

[dooon]

Wow

 

[frozentundra123456]

That's not happening. You'll have to settle for the 5820K and friends if that's what you want.

Ehhh.... it is just an interesting hobby for me to follow the forums and performance of AMD/Intel. I am perfectly satisfied with my Sandy i5. If anything, I need to upgrade my gpu a lot more than the cpu. I am getting older now, and although I still game a lot, I play mostly older games. The only games my comp has struggled with are DA:I and Witcher 3, and that is because I still have a wimpy HD7770.

All I am saying is that a mainstream hex core would be cheap now, especially if they only added a small igp, and I think it would motivate a lot of users to upgrade. Lets see what DX12 and Zen bring. Maybe if AMD can bring out an 8 core that is no longer beaten or equaled by a hyperthreaded quad, intel will change their strategy. Heck, they could even disable some instructions or something if they wanted to keep it from eating into their server market.

 

[Sweepr]

Just like Xeon-D came out of nowhere and killed any hope ARM would have in servers I'm pretty sure Intel will react if AMD delivers next year (after their failures I don't buy the hype though). Mainstream 6C+GT2 @ 3.5GHz would be probably be small and easy to do at a mature 14nm process. Another immediate solution would be lowering the price of Broadwell-E hexa-cores to LGA115x Core i7 levels ($330), and octo-cores down to $600. They are playing safe right now because they can.