Intel Skylake / Kaby Lake

[dark zero]

OK... it managed to get GTX 750 levels... still they failed their main target... Maxwell GTX 750Ti tier... is decent, but now Intel must work in their drivers.

BTW, the lack of OCing the iGPU hurts them.

 

[frozentundra123456]

AnandTech: The Intel Skull Canyon NUC6i7KYK mini-PC Review

www.anandtech.com/show/10343/the-intel-skull-canyon-nuc6i7kyk-minipc-review

Each to his own I guess, but like I have said before, not my thing. So expensive for middling performance just to get a small form factor. If you want small and portable, you can get a gaming laptop for 1000.00 (or even 750.00 at MC with a slow HDD) with a quad mobile processor and a GTX960M which will destroy this NUC in anything gpu related, with very similar cpu performance.

And as the reviewer said, seems like sort of disappointing performance for GT4e.

 

[AtenRa]

Nice review,

720p benchmarked on Performance (you know, we have to make this GT4e look awesome at 80+ fps)

then benchmarking 1080p at Extreme because nobody will ever try those settings on those iGPUs.

I just stopped reading the review after i show this.

:thumbsdown:

 

[monstercameron]

Its so tdp limited...kinda makes me regret buying the parts for it...but i still kinda want it. Atleast now ill wait a bit.

 

[LTC8K6]

Nice review,

720p benchmarked on Performance (you know, we have to make this GT4e look awesome at 80+ fps)

then benchmarking 1080p at Extreme because nobody will ever try those settings on those iGPUs.

I just stopped reading the review after i show this.

:thumbsdown:

You skipped one that kinda' busts your point, I think. It certainly doesn't make GT4e look good.

 

[IntelUser2000]

The Iris Pro 580 is 40-50% faster in gaming compared to the Brix Pro's Iris Pro 5200.

Brix Pro gets 2330 in 3DMark11.

2330 x 1.45 = 3380

That's why I like 3DMark11 for Intel GPU comparisons. Every few years you find a version of 3DMark that's useful for the games of that generation. 3400 points is about the level Iris Pro 580 gets, or 45% faster than Iris Pro 5200, and you get that in games.

You add 10% for 65W and you end up with 3700. At 15% it'll be 3900. 20-30% faster than Broadwell. That's called disappointing. If they weren't able to fully utilize 50% extra compute resources, there's very small chance that Kabylake will do even what Skylake did, nevermind better.

 

[Qwertilot]

Depends slightly - if its something 'simple' (ha!) gone wrong then it could of course give them something to fix in Kaby

It does all seem a bit odd. Underperforming a little, but also simply putting something with all these GPU resources into such a small case/such a small TDP that the chip can't really hope to use them.

 

[HiroThreading]

AnandTech: The Intel Skull Canyon NUC6i7KYK mini-PC Review

www.anandtech.com/show/10343/the-intel-skull-canyon-nuc6i7kyk-minipc-review

Such a ridiculous decision to not have the Alpine Ridge (or two) controller run off the CPU's PCIE lanes.

 

[majord]

OK... it managed to get GTX 750 levels... still they failed their main target... Maxwell GTX 750Ti tier... is decent, but now Intel must work in their drivers.

BTW, the lack of OCing the iGPU hurts them.

Not quite.. It's not even at r7-250x levels, letalone GTX750.

As mentioned - Thermally limited @ 45w. 65w SKU might run close to GTX750 / 250x at least.

Too little too late though really. 14/16nm Discrete Desktop, and Mobile GPU's are literlaly just around the corner, and will completely wipe the floor with it. Even with the inherent perf/watt handicap a discrete GPU has over integrated.

 

[NTMBK]

It gets demolished by the AsRock VisionX- that thing has an ancient 28nm Cape Verde GPU, clocked down to ~750MHz. And it still finds space to fit in an optical drive, meaning it's actually a useful HTPC.

Can't wait to see the next generation of HTPCs- small 14nm GPUs (like Polaris 11, or the NVidia equivalent) plus a Skylake or Zen CPU. Or Zen APU with HBM, when they eventually come out.

 

[IntelUser2000]

It gets demolished by the AsRock VisionX- that thing has an ancient 28nm Cape Verde GPU, clocked down to ~750MHz. And it still finds space to fit in an optical drive, meaning it's actually a useful HTPC.

At the current power level and cost, the only way it could have been justified was if they had the possible configuration posed on a slide - 2TFLOP device.

That might have been OK for performance. Not stellar. GT2 Gen 9 seems fine, but this is a fail. I was hoping it would at least achieve 50% perf/watt improvement. That's a no go for sure. If AMD needs 2x perf/watt to be competitive with Nvidia, Intel needs 3x. I am going to justify this by saying its likely due to 14nm failures, but we may have to realize they really don't have the capability to make decent mid range GPUs.

 

[witeken]

Fool: Intel Corporation CEO Brian Krzanich Answers Two Important Questions

Ashraf Eassa

he first question that I submitted was the following:

Your competition have said that they will transition to 10-nanometer manufacturing by the end of the year, earlier than Intel will. Can investors have confidence that Intel has a clear manufacturing lead?

CEO Brian Krzanich responded with the following (emphasis mine):

My commitment to you is that yes, you can, as an investor, count on Intel having a clear manufacturing lead. You have to remember these measurements don't necessarily mean all the same thing and how people are measuring the 10-nanometers versus 14-nanometers is different.

You really need to take a look at the performance of the devices and what's being delivered. And, so, we believe our 14-nanometer technology will still outperform any of the other technologies that'll be introduced in the next several years. Our 10-nanometer technology will again make another large leap in performance and be even further ahead than we think the 14-nanometer will be.

Pretty much what I said: Intel's 14nm should be very competitive with 10nm.

But take his statement with a grain of salt. Even if their lead was shrinking, they would never admit that of course. I don't think CEOs are that honest.

Edit: His confidence about 10nm performance gives me a slight optimism that 10nm could still be III-V (I mean, of course it is already decided what the 10nm recipe will be, but only Intel knows that right now of course -- the fact they had to delay 10nm also adds to that, because III-V also has major yield difficulties).

 

[frozentundra123456]

"Another large leap in performance"? What, another 5 to 10 percent like we have seen for every generation since Sandy bridge?

 

[VirtualLarry]

"Another large leap in performance"? What, another 5 to 10 percent like we have seen for every generation since Sandy bridge?

I think he's talking about density and performance/watt, but yeah, I get what you mean. Basically a performance snore-fest.

 

[frozentundra123456]

Of course, but did 14 nm even bring that much gain in performance per watt? I am not seeing any great increases in battery life or elimination of throttling.

 

[Arachnotronic]

"Another large leap in performance"? What, another 5 to 10 percent like we have seen for every generation since Sandy bridge?

Transistor performance, much more evident in low power products these days than in high performance ones.

 

[SAAA]

Well 14nm allowed sub 5W core-m products and 24-cores server chips while 22nm stopped at 11W and 18-cores CPUs.
But there is still some margin for both to improve on 14nm: ~30cores under 650mm^2 seems possible after 24 cores at 450mm^2. Then at that same 5W TDP the models evolved from 2.6GHz turbo to 3.1GHz for Skylake, IGP got massive increase too.

Performance are coming, but just a little to mainstream CPUs. Single thread stagnation is entirely another problem and won't be solved by 10-7-5nm unless they clock much, much higher than current parts. Who knows, maybe III-V would allow 5GHz base clock and then everyone'll be happy

 

[AtenRa]

You skipped one that kinda' busts your point, I think. It certainly doesn't make GT4e look good.

You missed my point, nobody will buy the GT4e HD580 to play at 720p, you can already do that with the $130 Core i3 and HD530.
Its the 1080p we want to see what this huge iGPU + eDRAM can do and what people spending 1K USD would use this iGPU in games.

The reviewer should have used high Image Quality settings at 720p and as he was increasing the resolution he had to lower the IQ settings to maintain playable performance, not increase them.

Can you tell me your evaluation on the 1080p performance of the HD580 ?? No you cannot, because the 1080p data gathered in this review are useless for evaluating the performance of the product in a meaningful manner.

Unless I missed something and the goal of the review was to compare the HD580 to the GTX970

 

[Phynaz]

"Another large leap in performance"? What, another 5 to 10 percent like we have seen for every generation since Sandy bridge?

You need to get out of the mindset that performance is determined only by execution speed.

Think of it this way, is your job performance only measured by how fast you work?

 

[frozentundra123456]

Transistor performance, much more evident in low power products these days than in high performance ones.

Yea, sure made Cherry Trail a dynamo didnt it?

 

[Arachnotronic]

Yea, sure made Cherry Trail a dynamo didnt it?

Cherry Trail was a crappy SoC. Don't blame the process for a bad product.

 

[dark zero]

Cherry Trail was a crappy SoC. Don't blame the process for a bad product.

At least people, except the ultra fans admits that truth. Cherry Trail is the Bulldozer of Intel.

 

[Arachnotronic]

At least people, except the ultra fans admits that truth. Cherry Trail is the Bulldozer of Intel.

Bulldozer did better in the marketplace than Cherry Trail did, so that's kind of being too harsh on ol' Dozer.

 

[HiroThreading]

Bulldozer did better in the marketplace than Cherry Trail did, so that's kind of being too harsh on ol' Dozer.

Haha!

Furthermore, Cherry Trail's fate was determined years ago when Intel failed to take the ultra mobile market seriously.

 

[dark zero]

Intel had reasons to kill SoFia, but killing Broxton was alright stupid.... Intel could have went with Windows Mobile and finally getting the marketshare they want... If AMD or VIA manages to make an alliance with MS to bring a cheap X86 phone with Win 10 it would be epic to watch.