Intel Skylake / Kaby Lake

[ClockHound]

Remember eSATA?

Yes. And still use it everyday. It was (is, sort of) a widely embraced standard for cheapskate video editors and their cheapskate outboard RAID arrays. Oh, yeah, we're a niche. Never mind.

Less niche than than Sata Express tho. :biggrin:

 

[WaitingForNehalem]

Is there any other M.2 SSDs than Samsungs x4 that overheat?

I believe the overheating is just a consequence of the NAND working at those speeds without a heatsink. Intel has a heatsink on both versions of the 750. This might be remedied by a geometry shrink later on but for the moment these chips need some sort of cooling.

 

[ShintaiDK]

I believe the overheating is just a consequence of the NAND working at those speeds without a heatsink. Intel has a heatsink on both versions of the 750. This might be remedied by a geometry shrink later on but for the moment these chips need some sort of cooling.

Its the controller overheating on the fast Samsung models. The NAND doesnt have any issues. Isolated cooling on the controller fixes it.

But again, this is a Samsung issue on their superfast x4 AHCI/NVME M.2 SSDs. So in short, the amounts of SSDs with the issue is what. Sub 1%?

 

[Dave2150]

Its the controller overheating on the fast Samsung models. The NAND doesnt have any issues. Isolated cooling on the controller fixes it.

But again, this is a Samsung issue on their superfast x4 AHCI/NVME M.2 SSDs. So in short, the amounts of SSDs with the issue is what. Sub 1%?

The drive only overheats during benchmarks.

During real world use, performance does not throttle.

 

[Sweepr]

Quick look at the first retail Carrizo benchmarks from NotebookCheck. Now remmember this AMD chip can be set anywhere from 15W to 35W (cTDP), I guess this particular HP model is closer to the latter as we're talking about a 17'' notebook.

Quick comparison with Broadwell-U Core i5 and estimated Skylake-U Core i5 performance (10% improvement). Core i5 5200U is the lowest bin Broadwell-U Core i5. There is a faster Carrizo model (A10 8800P) and a bunch of faster Broadwell-U models (Core i5 5300U, Core i7 5500U, Core i7 5600U) too, I will include results as soon as new reviews are available.

Cinebench R15 - CPU Single 64Bit
A10 8700P (15-35W TDP): 69
Core i5 5200U (15W TDP): 108
Core i5 6200U (15W TDP, estimated): 118.8

Cinebench R15 - CPU Multi 64Bit
A10 8700P (15-35W TDP): 194
Core i5 5200U (15W TDP): 259
Core i5 6200U (15W TDP, estimated): 284.9

Cinebench 11.5 - CPU Single 64Bit
A10 8700P (15-35W TDP): 0.86
Core i5 5200U (15W TDP): 1.25
Core i5 6200U (15W TDP, estimated): 1.37

Cinebench 11.5 - CPU Multi 64Bit
A10 8700P (15-35W TDP): 2.36
Core i5 5200U (15W TDP): 2.82
Core i5 6200U (15W TDP, estimated): 3.1

WinRAR
A10 8700P (15-35W TDP): 1328
Core i5 5200U (15W TDP): 2326
Core i5 6200U (15W TDP, estimated): 2558

3DMark 11 Physics
A10 8700P (15-35W TDP): 2181
Core i5 5200U (15W TDP): 3349
Core i5 6200U (15W TDP, estimated): 3683

X264 HD Benchmark 4.0 - Pass 1
A10 8700P (15-35W TDP): 74.4
Core i5 5200U (15W TDP): 85.3
Core i5 6200U (15W TDP, estimated): 93.83

X264 HD Benchmark 4.0 - Pass 2
A10 8700P (15-35W TDP): 15.2
Core i5 5200U (15W TDP): 16.2
Core i5 6200U (15W TDP, estimated): 17.82

Geekbench 3 - 64 Bit Single-Core Score
A10 8700P (15-35W TDP): 2170
Core i5 5200U (15W TDP): 2691
Core i5 6200U (15W TDP): 2928

Geekbench 3 - 64 Bit Multi-Core Score
A10 8700P (15-35W TDP): 5105
Core i5 5200U (15W TDP): 5327
Core i5 6200U (15W TDP): 6195

Short story, despite the hype Intel maintains a sizeable CPU performance advantage (especially in ST) and Skylake-U is right around the corner.

 

[mikk]

Looking at the scores this must be a ~20W configured notebook, otherwise it is a debacle for AMD. For 17" this would be a poor choice. Not only HP's fault, this is also AMDs fault with their messy huge cTDP variance.

 

[WaitingForNehalem]

Its the controller overheating on the fast Samsung models. The NAND doesnt have any issues. Isolated cooling on the controller fixes it.

But again, this is a Samsung issue on their superfast x4 AHCI/NVME M.2 SSDs. So in short, the amounts of SSDs with the issue is what. Sub 1%?

Do you have a source for that? Legit Reviews claims it is both the controller and NAND: http://www.legitreviews.com/samsung-sm951-512gb-m-2-pcie-ssd-review_161689/3

Wouldn't it make sense that the controller throttles the speed to decrease the temps of the NAND packages?

 

[WaitingForNehalem]

The drive only overheats during benchmarks.

During real world use, performance does not throttle.

A 2 minute sustained write is not that uncommon in the real world. I wouldn't spend that much money on a drive that is prone to overheating.

 

[nvgpu]

https://twitter.com/TMFAeassa/status/613165391280480256

Skylake does not support fixed function HEVC 10bit hardware decoding it seems.

 

[ShintaiDK]

Do you have a source for that? Legit Reviews claims it is both the controller and NAND: http://www.legitreviews.com/samsung-sm951-512gb-m-2-pcie-ssd-review_161689/3

Wouldn't it make sense that the controller throttles the speed to decrease the temps of the NAND packages?

Example:
http://forums.anandtech.com/showpost.php?p=37576577&postcount=24

Its not the NAND. It simply doesnt use anywhere near that power. So forget that.

 

[Sweepr]

I've wondered about this same thing, but I'm not sure I see the point. Personally, I'm not a big fan of dual-booting and neither are average consumers.

Are they really going to do multiple architecture thing yet again.

I think it would be an interesting way to differentiate the iPad Pro from iPad Air and iPhone. IMHO iOS looks a bit limited for a productivity oriented 12.2'' device that should carry Apple prices. If they choose this route (ARM+iOS) I hope they prove me wrong. Otherwise I'd probably opt for a $799 Core i3 6100U (Skylake-U) Surface Pro 4.

Looking at the scores this must be a ~20W configured notebook, otherwise it is a debacle for AMD. For 17" this would be a poor choice. Not only HP's fault, this is also AMDs fault with their messy huge cTDP variance.

Either way, a disappointing debut. It only gets worse when you think they won't have any new mobile chip till 2017.

 

[Sweepr]

Core i7 6700K Geekbench Windows 64-bit Score
Single-thread: 5217
Multi-thread: 20774

http://browser.primatelabs.com/geekbench3/3044491

Can someone run a Haswell Core i7 at 4.0 or 4.1GHz and share some results?
Average Core i7 4790K scores under Windows 64-bit are ~4700/18500.

 

[Abwx]

Quick look at the first retail Carrizo benchmarks from NotebookCheck. Now remmember this AMD chip can be set anywhere from 15W to 35W (cTDP), I guess this particular HP model is closer to the latter as we're talking about a 17'' notebook.

Fortunately we have the numbers that suggest that the ST scores are achieved at 2.8GHz, NBC could do nothing about it but to repeat their infamous review of Kaveri whose ST scores where later updated....

FYI ST scores in CB11.5/R15 should be about 0.95 and 80 respectively, any other score point to a butchered methodology...

And contrary to your innacurate claims the CB scores say that it s running barely at 15W, and screen size has nothing to do with TDPs, HP previously used 19W Kaveris and 14W Beemas for these kind of laptops.

Cinebench R15 - CPU Multi 64Bit
Core i5 5200U (15W TDP): 259

Cinebench 11.5 - CPU Multi 64Bit
Core i5 5200U (15W TDP): 2.82

Not only you didnt provide any link to those numbers but the scores are performed at much higher TDP than what is stated in your post, that said if it makes you happy to believe those fairy taled Intel TDP numbers, why not...

 

[Arachnotronic]

Core i7 6700K Geekbench Windows 64-bit Score
Single-thread: 5217
Multi-thread: 20774

http://browser.primatelabs.com/geekbench3/3044491

Can someone run a Haswell Core i7 at 4.0 or 4.1GHz and share some results?
Average Core i7 4790K scores under Windows 64-bit are ~4700/18500.

Seems a little high for the 4790K; I get more along the lines of 4400/16200 at stock in Windows 7 64-bit mode.

 

[poofyhairguy]

https://twitter.com/TMFAeassa/status/613165391280480256

Skylake does not support fixed function HEVC 10bit hardware decoding it seems.

I am sad to learn this fact, but also kinda happy because now I can buy a few more Haswell HTPCs without worry.

 

[Sweepr]

Seems a little high for the 4790K; I get more along the lines of 4400/16200 at stock in Windows 7 64-bit mode.

64-bit? That would make the Core i7 6700K score more impressive.

 

[frozentundra123456]

Quick look at the first retail Carrizo benchmarks from NotebookCheck. Now remmember this AMD chip can be set anywhere from 15W to 35W (cTDP), I guess this particular HP model is closer to the latter as we're talking about a 17'' notebook.

Quick comparison with Broadwell-U Core i5 and estimated Skylake-U Core i5 performance (10% improvement). Core i5 5200U is the lowest bin Broadwell-U Core i5. There is a faster Carrizo model (A10 8800P) and a bunch of faster Broadwell-U models (Core i5 5300U, Core i7 5500U, Core i7 5600U) too, I will include results as soon as new reviews are available.

Cinebench R15 - CPU Single 64Bit
A10 8700P (15-35W TDP): 69
Core i5 5200U (15W TDP): 108
Core i5 6200U (15W TDP, estimated): 118.8

Cinebench R15 - CPU Multi 64Bit
A10 8700P (15-35W TDP): 194
Core i5 5200U (15W TDP): 259
Core i5 6200U (15W TDP, estimated): 284.9

Cinebench 11.5 - CPU Single 64Bit
A10 8700P (15-35W TDP): 0.86
Core i5 5200U (15W TDP): 1.25
Core i5 6200U (15W TDP, estimated): 1.37

Cinebench 11.5 - CPU Multi 64Bit
A10 8700P (15-35W TDP): 2.36
Core i5 5200U (15W TDP): 2.82
Core i5 6200U (15W TDP, estimated): 3.1

WinRAR
A10 8700P (15-35W TDP): 1328
Core i5 5200U (15W TDP): 2326
Core i5 6200U (15W TDP, estimated): 2558

3DMark 11 Physics
A10 8700P (15-35W TDP): 2181
Core i5 5200U (15W TDP): 3349
Core i5 6200U (15W TDP, estimated): 3683

X264 HD Benchmark 4.0 - Pass 1
A10 8700P (15-35W TDP): 74.4
Core i5 5200U (15W TDP): 85.3
Core i5 6200U (15W TDP, estimated): 93.83

X264 HD Benchmark 4.0 - Pass 2
A10 8700P (15-35W TDP): 15.2
Core i5 5200U (15W TDP): 16.2
Core i5 6200U (15W TDP, estimated): 17.82

Geekbench 3 - 64 Bit Single-Core Score
A10 8700P (15-35W TDP): 2170
Core i5 5200U (15W TDP): 2691
Core i5 6200U (15W TDP): 2928

Geekbench 3 - 64 Bit Multi-Core Score
A10 8700P (15-35W TDP): 5105
Core i5 5200U (15W TDP): 5327
Core i5 6200U (15W TDP): 6195

Short story, despite the hype Intel maintains a sizeable CPU performance advantage (especially in ST) and Skylake-U is right around the corner.

Yea, there is a post about that laptop in the Carizzo thread as well. It is really hard to interpret since we dont know the configuration of the Carizzo laptop. If it is 35 watt, it is a disaster for AMD. If it is 15 or 20 watts, it is competitive, but certainly nothing that lives up to the hype. Broadwell is clearly more efficient, and Skylake will only widen the gap.

 

[Sweepr]

Not only you didnt provide any link to those numbers but the scores are performed at much higher TDP than what is stated in your post, that said if it makes you happy to believe those fairy taled Intel TDP numbers, why not...

You can find them in NBC's reviews. I'm sorry but there's nothing I can do about the disappointing A10 8700P scores, they simply reflect a reality that you're trying very hard to dismiss with your made up numbers, Carrizo is still considerably slower than even the lowest bin Core i5 Broadwell-U, not to mention soon-to-be-launched Skylake-U and likely won't change AMD's situation in the mobile market.

Too bad there's no contra revenue to argue this time, it would make a great excuse wouldn't it?

 

[csbin]

 

[frozentundra123456]

You can find them in NBC's reviews. I'm sorry but there's nothing I can do about the disappointing A10 8700P scores, they simply reflect a reality that you're trying very hard to dismiss with your made up numbers, Carrizo is still considerably slower than even the lowest bin Core i5 Broadwell-U, not to mention soon-to-be-launched Skylake-U and likely won't change AMD's situation in the mobile market.

Too bad there's no contra revenue to argue this time, it would make a great excuse wouldn't it?

I am sure he will have no problem finding some excuse, if nothing more that just making up some numbers for Intel that fit his agenda.

 

[Sweepr]

I am sure he will have no problem finding some excuse, if nothing more that just making up some numbers for Intel that fit his agenda.

Expect Core i3 5005U (slowest Broadwell-U Core i3) results and overblown Intel TDP estimates in the coming posts. I admit there's some entertaining value in watching him try hard to come up with excuses, carefully selected tests/numbers and straight up lies that somehow (magically) always put AMD always in the better light.

 

[Abwx]

You can find them in NBC's reviews.

So far i did find a single one with a Broadwell scoring 2.8 and being at 15W, indeed it s no wonder that you re not prodviding the link of the scores you posted, i guess that power numbers must be really annoying, am i wrong..??.

I'm sorry but there's nothing I can do about the disappointing A10 8700P scores, they simply reflect a reality that you're trying very hard to dismiss with your made up numbers, Carrizo is still considerably slower than even the lowest bin Core i5 Broadwell-U, not to mention soon-to-be-launched Skylake-U and likely won't change AMD's situation in the mobile market.

Too bad there's no contra revenue to argue this time, it would make a great excuse wouldn't it?

You are deluding yourself by using numbers that you know are not at all relevant, provide the links rather than hiding behind wall of arrogant texts and products that do not exists...

 

[Abwx]

I am sure he will have no problem finding some excuse, if nothing more that just making up some numbers for Intel that fit his agenda.

Lol, while finding an excuse to not post the compromising links, so much for the insight..

You would have at least provided something yourself if you were confident about Broadwell fakes TDPs, but looks that no one wants to take the challenge...

 

[Sweepr]

Are you lazy too? Both are average/median results for Core i5 5200U (I could easily find better scores), unlike your handpicked AMD numbers.

www.notebookcheck.net/Intel-Core-i5-5200U-Notebook-Processor.127831.0.html

 

[IntelUser2000]

NotebookCheck got 2.64 points. I consider ASUS T300 Chi and Apple's 12'' Macbook the best Broadwell-Y Core M implementations.

Again, I said it performs 2.6 points in certain reviews. But if you look at more reviews, it varies substantially. It was up to Core M's introduction Cinebench was a reliable indicator of improvement. Not anymore.

I do not count 2600K as that was the last great improvement. A relevant one too, since it was on single thread, not corner case multi-threading only like Nehalem and minimal on single thread after you account for Turbo.

IPC

"Good gains"
Core Duo to Core 2 Duo: 15-20%
Core 2 Duo to Core 2 Penryn: 5%
Core 2 Duo to Nehalem: 5% in ST(not Turbo), 25% in multi
Lynnfield vs Sandy Bridge: 13.2%

"Bad gains"
Sandy Bridge vs Ivy Bridge: 3.7%
With Haswell Anandtech mentions an improvement of 8.3%
4770K: Same chip with factory overclocking

I'm talking about stock performance. Sorry but changing the goalpost to include OCing won't turn the 5% a year (stock) lie to true.

Stock or not most complaints about Ivy Bridge and Haswell was much about lack of IPC improvements as overclocking regression. That certainly matters to this crowd. This also tells so-called Intel process superiority that was long thought to be transistor performance matters extremely little if at all because every high performance CPU crowds at the same frequency levels and TDP/Power limits are the biggest determining factor of how CPUs perform. No wonder they shut up about that and is only mentioning 0.7V and density comparisons at IEDM and such.

Sandy Bridge did 4.6-4.7GHz in 32nm process back in 2011, 4.5 years ago. Skylake isn't doing much better at 5GHz if the rumors are true. Sure, power use is lower at that frequency, however the top end performance has pretty much stalled. If we take for account 11% IPC gain, while in itself is decent, Kabylake will not have a "Devils Canyon" part where it factory overclocks it to look better than it really is. Perhaps they can, but it'll end up like some Phenom chips where it couldn't do above 200-300MHz before getting errors.

I could overclock my 2600K if my motherboard supported it and suddenly I'll have a 4.5GHz chip. Doesn't mean I have something drastically different. Just like that, because Intel factory overclocks it(aka Devils Canyon) and sells only parts that work at the frequency they want to sell it at and puts a stamp, doesn't make it much different than enthusiasts doing overclocks. Heck, for years Intel chips were so overclockable that such gains were thought to be normal.