Forget Intel Broadwell, Skylake On the Way

Nice. But that pricing... over $2000 US dollars with just the 970m, over $2600 for 980m... a bit steep. But I guess that's the price for being an early adopter.

Sent from my potato running on Android 5.0.1

 

can you please link?

 

When's the NDA for skylake expiring?

 

The new Intel Core i7-6500U (Skylake) is already in our rankings

 

Skylake has been meh so far. Hopefully they at least show better battery life and better performance per watt. But I am not holding my breath. Intel's delivery has been meh over past few years.

 

Hm... I wouldn't dismiss it out of hand however Skylake should be some improvement.. It will be 10-15% over Haswell hopefully...

 

Average 5% IPC improvement over Haswell per 6700K/6600K benchmarks

 

Wasn't talking about IPC... Talking more about the performance/W and battery life...

 

This is not an indication of CPU performance but one of laptop performance.

For instance article shows Cinebench 11.5 result for i7-6500U as 3.28 while i7-5500U @2.9GHz and 15W limit scores 3.32.

Spoiler

It also makes a comparison to dual core 4200m, I only have results for 4600m.

Spoiler

5.02 vs 3.28

Similar for the i7-6700HQ results, no mention about extra unlocked bins and the score of 6.93 is disappointing even at stock. IMO would be nice to see how far it could go, maybe when the 6700HK is released. An i7-4700MQ for comparison if that happens.

Spoiler

Not being a battery user I personally don't care much for battery life but for those that do the simple answer is to use a higher capacity battery. A 15W Skylake throttled to 15W is going to use as much power as a Haswell throttled to 15W. For idle powers there are other things to consider such as peripheral and display power.

 

Will we be getting 6X00 MQs? rPGA is the way... not BGA.

 

dont think so.. new unlocked K chip might be bga though.

 

No. All mobile CPUs are soldered since Broadwell.

 

Desktop processr Clevos all the way!!

 

There are some important next generation features missing: HDMI 2.0, USB 3.1 and VISA Adaptive Sync.

I'm a bit disappointed... going to wait for Kaby Lake.

 

Skylake for laptops: Faster Core M and Ultrabook GPUs with eDRAM
Intel officially announces the full initial lineup.

 

FWIW seems like the new Skylake laptops are still using 1600 DDR3L such as
HP Envy 17-n109tx i7-6700HQ
HP Envy 15-AE103TX - i7-6500U

 

So if I want a quad-core i7, is there anything of interest here at all? It seems all of the clock speeds are very similar to Haswell as is the IPC. I was hoping for the TDP to drop to 28W or at least 35W, but no, it is still 45W. I guess I will have to upgrade at some point, but Skylake is not very much of an inspiration for doing so. Hopefully this laptop will last until the next generation of GPUs.

 

Why would you want a TDP drop, the performance is tied directly to TDP, the lower the TDP the lower the performance.

 

..well, I suppose some of us were expecting optimizations that allowed similar synthetic performance on lower tdp in this iteration. Or, for example, more cores and better thread handling, over the same tdp envelope. Like.. a 17w quad-core with different clock limitations on the same hardware, at the very least. Or an actual increase in processing power on the "ultra-low" variants on two cores.

Not that that was realistic. But with a new "model", you would on the face of it expect at least something to change.

 

This is not necessarily true when comparing across architectures. We already knew that the difference in performance was going to be marginal so I was hoping that instead of trying to increase performance, they tried to lower the TDP. Obviously, that is not the case so I'm wondering if this new architecture makes any practical difference at all.

 

Althernai,

Not picking on you specifically, but I do need to address this issue (again).

The IPC increase is marginal, granted. But that is only when artificially limiting the testing to our old standards (i.e. same clocks, same RAM, same s&%*).

But for anyone (like me) that doesn't care about O/C'ing and actually sees it as a detriment to sustained productivity over time and simply compares the previous platform to the new, Skylake is offering about 20% improvement over Haswell. That is huge. And it will be especially so on mobile platforms that are geared for business, not gaming. Over Sandy Bridge? A 40% gain in performance.

See:
http://www.anandtech.com/show/9575/the-anandtech-podcast-episode-35-windows-10-and-skylake

(Listen around the 48 minute mark on...).

Speaking of gaming, yeah; small improvements there. But as I've already mentioned on this site; gaming doesn't drive performance innovation (and I don't believe it ever has, except maybe for GPU's - but to me GPU's play an insignificant part of my current workflows except to simply drive the monitor... igpu's are more than enough...).

Having anything older than Haswell will be a huge performance jump (stock to stock). This is 95% of the way most use a system. Not O'C'd, not modifying the BIOS, not beefing up the cooling system and/or TIM.

Just buy and use.

Anyone that needs the most mobile performance/productivity they can without looking like a mad scientist or a 14 year old geek with an O/C'd gaming 'rig' needs to upgrade to Skylake asap.

Suggesting otherwise is doing your friends, colleagues and clients a disservice.

 

I'm not talking about overclocking. If the clocks were notably different, I would not be asking my question... but they're not. Compare the cheapest of the quad cores:
i7-4700: 2.4 GHz base, 3.2 / 3.2 / 3.3 / 3.4 turbo for quad / triple / double / single active core.
i7-6700: 2.6 GHz base, 3.1 / 3.1 / 3.3 / 3.5 turbo for quad / triple / double / single active core.
The base clock speed is 8% higher, but it's not obvious that it will be that much better in practice given that the turbo speeds for Haswell are more or less the same and actually higher for multithreaded performance.

The high end is even worse for Skylake:
i7-4900: 2.8 GHz base, 3.6 / 3.6 / 3.7 / 3.8 turbo for quad / triple / double / single active core.
i7-6920: 2.9 GHz base, 3.4 / 3.4 / 3.6 / 3.8 turbo for quad / triple / double / single active core.
Base clock here is higher by 4%, but the turbo is equal at best and generally less than. I would be really surprised if they didn't more or less run at the same frequencies in real life workloads.

And note that I'm using the very earliest Haswell CPUs -- the ones from the refresh are clocked higher.

 

Again, you're comparing numbers. That do not tell the whole story.

Listen to that podcast if you don't want to believe me. As a complete platform: the gains are huge, especially for older platforms that 'common sense' would indicate are still as good or even better than Skylake actually is.

Not for gaming...

But gaming has never pushed hardware, imo. As previously stated, to increase gaming performance the 'engines' need to be overhauled, not simply get a new platform by itself.

The gaming developers are sitting on their cash cows because they know that most/all review sites will simply say 'Skylake? zzzz...'. But sooner or later they will have to get the s&%$ together and actually provide a better engine. Blaming others for your laziness (or lack of innovation) is not the way to succeed for long...

 

Disagree completely that "gaming has never pushed hardware." Intel wouldn't have ever even tried to improve integrated graphics if (AMD didn't do it first and if) there wasn't pressure from gamers who could not do anything on systems with integrated graphics. Platform changes can revolutionize the gaming experience when going from an older system to a new one with more powerful processor and graphics card.

 

@Althernai you can power limit your processor below TDP if you're happy to sacrifice performance but you cant have your cake and eat it too.

Don't know yet if these processors offer extra unlocked bins. Being of that 5% Tiller mentions I can give you some CPU numbers to compare between i7-6700HQ and i7-4700MQ.

PassMark CPU Mark i7-6700HQ 5519
PassMark CPU Mark i7-4700MQ 10503

Geekbench3 i7-6700HQ ST 3336, MT 12415
Geekbench3 i7-4700MQ ST 3946, MT 15628

Sisoft Processor Arithmetic i7-6700HQ 104.4GOPS
Sisoft Processor Arithmetic i7-4700MQ 119.4GOPS

Even the older 2nd and 3rd gen overclocked CPU's put up good numbers.

The i7-6820HK should be interesting though and since Intel has the i7-6820HK, i7-6820HQ and i7-6700HQ priced the same who would want an i7-6700HQ, the mind boggles.
http://download.intel.com/newsroom/kits/core/6thgen/pdfs/6th_Gen_Intel_Core-Intel_Xeon_Factsheet.pdf

Seems a thermal time window has being introduced with these new CPU's meaning a CPU can run hotter for a while before throttling, beware.

 

Dufus, your 'scores' are funny.

Comparing 16GB RAM to 8GB RAM, 3.64GHz (turbo) to 2.09GHz (turbo), RAID0 (most likely SSD's...) to a Toshiba HDD 7200RPM model that barely outperforms a Hitachi 5400RPM model.

See:
http://www.nikktech.com/main/articl...q01acf050-500gb-sata-iii-hdd?showall=&start=8

Even the 8GB RAM above is probably a low spec'd DDR3 example...

I'm sure you'll find more 'proof' that Skylake is subpar to over-volted and prehistoric, cpu relics of the past, but the proof is to simply use one.

Skylake is a future looking design (even Win10 currently doesn't support all it's capabilities, but it will - and Win8.1 and lower? Never), a new born, taking baby steps that keep up and surpass it's predecessors when judged fairly platform to platform.

I use some of my old systems (Arrandale) and some of my clients systems that are newer and older and chuckle that I once thought of them as bleeding edge. And today, some of those same systems have been upgraded to the fastest O/S I've seen in a while (Win10), the better SSD's we have today and the best/fastest RAM possible for their respective platforms. And just browsing the O/S, they're still slow and laggy vs. even a T100TA with 64GB eMMC storage and 2GB RAM. Why? Because the platform matters.

I'll eventually get a system to compare to my current workstations and I may change my mind (who knows?). But until then, everything (sane) I've researched on Skylake points to me just giving a blank cheque to my local tech store and revamping my entire operation, asap.

But the caveats (for me) are: no O/C'ing, no GPU's (if I can get away from them) and no need of increasing my 'gaming' results (I don't game). Given the above; Skylake makes upgrading even from Broadwell a worthwhile move (depending on the business). From Haswell or below? If upgrading is not in the plans; I'll be coming to the 'out of business' sale soon...

 

From reading some of the architecture recaps today it seems that Intel is pushing "power efficiency" moreso by implementing better gating and allowing the processor to switch amongst p-states more quickly. So it reaches idle faster and uses less power when idle. They don't seem to have done much on the other end of the spectrum where you're pushing the CPU hard. In that sense, Meaker's conclusion is basically correct, the top end performance will always be quite tied to the TDP.

Intel has kept their TDPs pretty steady over the last few iterations, so I don't see them changing too drastically. They've basically decided ~45W is a reasonable amount of power to dissipate in a notebook, whereas ~90W is a reasonable amount for quad core desktops.

 

they have announced a pair of 25W quad core, namely Core i7-6822HQ and Core i5-6442HQ
seems like a nice upgrade from previous generations of lower voltage (37W) quad core processors (4702MQ, 3612QM, etc)

 

sonyfw550, why would less performance be a nice upgrade?

Intel takes essentially the same cores, and slaps TDP performance limits on them.

What you get with a 25w part is 68% of the performance headroom of the 37w version of the same core, which would be 82% of the maximum performance of the top end 45w part.

 

I mean a nice improvement in power consumption and battery life

 

@sonyfw550 You can run a 4700MQ with 37W limit and get pretty much the same performance as the 4702MQ. My guess would be the 4702MQ is just a 4700MQ with out of spec thermals so has had it's clocks reduced and hence lower TDP to keep within spec while marketed as a "lower power" chip.

As for low voltage chips, what does it really mean. I've seen 4th gen desktop dual cores undervolt to use less voltage than undervolted mobile U chips at similar clocks (2.7GHz). Perhaps it has to do with idle voltages or that the maximum voltage used is less. It certainly doesn't seem to use less voltage for same clocks.

 

...for less performance...

You can get the same thing by use Intel XTU to downgrade your multipliers, and run in power saving mode.

It's not an improvement, it's less of the same thing.

 

when you compare it to the same generation it is certainly less of the same thing.
I'm happy if it can match the performace of a 3610QM with 40% or more improvement in battery life. I think that would be an improvement.
So I can take a 6700HQ and make myself a 6822HQ from it? Would Intel make additional tweaks that I cannot easily do to keep a good balance of performance and battery life, like the dynamic of the clock rate?

 

Closer to 4GHz IIRC and CPU mark is not dependent on disk drives. I can only compare with available results and those are the highest ones for Skylake so far. As I've mentioned before overclocking extends the useful life of my laptops, usually I only plan to upgrade once every 5 or 6 years. If older gen can be overclocked to outperform a newer gen while the newer gen cannot be then CPU performance wise it is a downgrade all other things being equal such as instructions and width. For instance AVX512 will offer substantial gains for software that can make use of it but yet to come on the mobile platform.

As already said results of the unlocked i7-6820HK should IMO be interesting when it arrives. Only seen one result so far for the MSI GT72 6QD Sisoft Processor Arithmetic 107.5GOPS

@sonyfw550 For battery life you cant beat getting a bigger capacity battery.

 

I've tried to google the 6822HQ and while I've found some references to it, there aren't any specs and it's not in Intel's ARK. Do you know what it actually does?

Not necessarily. Sometimes these are binned chips with the same performance at lower TDP. For example, on desktops there is the i7-6700T which actually has higher clocks (2.8 GHz base, 3.4 / 3.4 / 3.5 / 3.6 turbo) than the laptop i7-6700HQ, but its TDP is only 35W. I wonder why they don't just stuff that one in a laptop.

 

@Althernai probably means EQ embedded chips

http://ark.intel.com/products/90615/Intel-Core-i7-6822EQ-Processor-8M-Cache-up-to-2_80-GHz

http://ark.intel.com/products/90617/Intel-Core-i5-6442EQ-Processor-6M-Cache-up-to-2_70-GHz

 

Thanks, that makes sense. Embedded chips serve a rather different purpose (though it's still interesting that they'd have a quad-core one like that).

 

Yeah; we're good. Even if we're at opposite ends of the usage spectrum.

If O/C'ing is a way of life; then it's hard to argue with your perspective. If using new platforms at close to stock settings and comparing them to older platforms at their stock settings; upgrading is a no brainer for almost every platform Intel has introduced since 2006.

That 4GHz clock rate you indicate is not what was in your original post (it did state 2.09GHz 'turbo').

Is that 107.5GOPS 'score' good? Or is it something to be ashamed of?

 

I listened to the podcast and read the architecture summaries and indeed, they introduce quite a few new tricks. However, I haven't seen any practical applications of this so far. What do all of these changes add up to? Will Skylake laptops get more battery life?

 

That depends on your uses and the actual notebook in question.

If you are having a light workflow; the battery life should be impressive when all things are taken into consideration (battery size, weight of device, screen size/resolution, igpu or GPU, etc.).

If you have a constant and demanding workload; the battery life will depend on the size of the battery (as always).

For the latter; I would expect more 'work' to be done for the same workload/time period. But effectively extending the battery life is not one of the benefits I would be looking for in such a scenario.

 

Guys, the GT80 group is looking forward to the i7-6820HK in the new release, so I thought I would do a run on my GT80 with i7-5950HQ for comparison.

SiSoftware Sandra
Benchmark Results
Aggregate Native Performance : 122.52GOPS
Dhrystone Integer Native AVX2 : 169.34GIPS
Whetstone Single-float Native AVX : 104.5GFLOPS
Whetstone Double-float Native AVX : 75.18GFLOPS
Results Interpretation : Higher Scores mean Better Performance.
Decimal Numeral System (base 10) : 1GOPS = 1000MOPS, 1MOPS = 1000kOPS, 1kOPS = 1000OPS, etc.

Performance per Thread
Aggregate Native Performance : 15.31GOPS
Dhrystone Integer Native AVX2 : 21.17GIPS
Whetstone Single-float Native AVX : 13GFLOPS
Whetstone Double-float Native AVX : 9.4GFLOPS
No. Threads : 8
Results Interpretation : Higher Scores mean Better Performance.
Decimal Numeral System (base 10) : 1GOPS = 1000MOPS, 1MOPS = 1000kOPS, 1kOPS = 1000OPS, etc.

Performance vs. Power
Processor(s) Power : 200.000W
Aggregate Native Performance : 612.58MOPS/W
Dhrystone Integer Native AVX2 : 846.71MIPS/W
Whetstone Single-float Native AVX : 522.53MFLOPS/W
Whetstone Double-float Native AVX : 375.89MFLOPS/W
Results Interpretation : Higher Scores mean Better Performance.

Performance vs. Speed
Aggregate Native Performance : 42.35MOPS/MHz
Dhrystone Integer Native AVX2 : 58.54MIPS/MHz
Whetstone Single-float Native AVX : 36.12MFLOPS/MHz
Whetstone Double-float Native AVX : 25.99MFLOPS/MHz
Results Interpretation : Higher Scores mean Better Performance.

Benchmark Status
Result ID : Intel(R) Core(TM) i7-5950HQ CPU @ 2.90GHz (4C 8T 2.89GHz/4GHz, 2.9GHz IMC/3.6GHz, 4x 256kB L2, 6MB L3, 1.8GHz 128MB L4)
Computer : MSI GT80 2QE (MSI MS-1812)
Platform Compliance : x64
NUMA Support : No
No. Threads : 8
Processor Affinity : U0-C0T0 U2-C1T0 U4-C2T0 U6-C3T0 U1-C0T1 U3-C1T1 U5-C2T1 U7-C3T1
System Timer : 2.83MHz
No. Runs : 64000 / 640

Processor
Model : Intel(R) Core(TM) i7-5950HQ CPU @ 2.90GHz
Speed : 2.89GHz (72%)
Min/Max/Turbo Speed : 800MHz - 2.89GHz - 4GHz
Maximum Power : 47.000W - 200.000W
Cores per Processor : 4 Unit(s)
Threads per Core : 2 Unit(s)
Front Side Bus Speed : 100MHz
Type : Laptop/Netbook
Revision/Stepping : 47 / 1
Microcode : MU0647010D
Latest Version : MU06470111
L1D (1st Level) Data Cache : 4x 32kB, 8-Way, 64bytes Line Size, 2 Thread(s)
L1I (1st Level) Code Cache : 4x 32kB, 8-Way, 64bytes Line Size, 2 Thread(s)
L2 (2nd Level) Data/Unified Cache : 4x 256kB, ECC, 8-Way, 64bytes Line Size, 2 Thread(s)
L3 (3rd Level) Data/Unified Cache : 6MB, ECC, 12-Way, Fully Inclusive, 64bytes Line Size, 16 Thread(s)
L4 (4th Level) Data/Unified Cache : 128MB, 16-Way, 64bytes Line Size, 16 Sectors, 16 Thread(s)

Memory Controller
Integrated in Processor : Yes
Speed : 2.9GHz (80%)
Min/Max/Turbo Speed : 800MHz - 3.6GHz

Features
SSE - Streaming SIMD Extensions : Yes
SSE2 - Streaming SIMD Extensions v2 : Yes
SSE3 - Streaming SIMD Extensions v3 : Yes
SSSE3 - Supplemental SSE3 : Yes
SSE4.1 - Streaming SIMD Extensions v4.1 : Yes
SSE4.2 - Streaming SIMD Extensions v4.2 : Yes
AVX - Advanced Vector eXtensions : Yes
FMA3 - Fused Multiply/Add eXtensions : Yes
AVX2 - Advanced Vector eXtensions v2 : Yes
HTT - Hyper-Threading Technology : Yes

 

Not sure if this is a common approach; however, in my case, I tend to use battery for typing notes, replying to emails, etc., and only run demanding applications when plugged in to AC.

I suppose this kind of usage would benefit from certain features of Skylake mobile processors, how about a laptop with desktop Skylake CPU? All other factors aside, would there likely be any longer battery life with i7-6700K compared to laptop with i7-4790K, for example?

 

For the particular tricks that I mentioned (faster to switch p-states + gating), it should potentially lead to a longer battery life or equivalently a marginally lower power bill (if plugged in). The workload of a regular user has long periods of nothing. You load a webpage, then it takes you awhile to read. You type a few sentences in an email, then you pause for awhile, then type a bit more. It might be a couple of seconds to us (or only a minute), but the quicker the processor ramps down, the more power it saves. "gating" also saves power in the processor can actually shut off parts that aren't being used during these periods of idle.

Way back in the day, your processor used to just go all out all the time. Modern processors aggressively ramp down (1 ms for skylake is quite impressive) and can selectively shut off various sections that it deems it doesn't need.

That's all a long-winded way of saying that the practical application is battery life. Back when Intel was first talking about ultrabooks, they wanted to take battery life from 2 hours to 12 hours or something. This is how they've worked towards accomplishing that: the processor essentially shut downs as soon as you stop (even for a moment). For a gaming laptop that you're running full out in a game, there are no periods of idle, so those particular features don't do very much.

A quick caveat: I'm talking only about switching p-states and better gating. There are a number of other features of the architecture, but the above doesn't apply to them.

 

What Skylake offers is for many more 'P' states and they will also be hardware controlled (about 30x faster - Speed Shift Technology, Hardware P states). This should help battery life a lot in specific workflows.

In addition, they offer Duty Cycling of the cores too - this is to offset the loss (yeah; loss) of efficiency when the cores are running too slow and at too low a voltage from the optimum for the chipset/materials. Yeah; balance is needed even at this fine grained level too.

See:
http://www.anandtech.com/show/9582/intel-skylake-mobile-desktop-launch-architecture-analysis

 

Thanks for the extra info, tilleroftheearth. While features like these don't carry the same jazz as higher clockspeed and the like, I still quite enjoy reading about them. The sophistication of the design is quite interesting in its own right.

 

tilleroftheearth, that's on paper; most people here need to see real-world results to be convinced.

 

Marginal IPC improvement, basically non existent clock speed improvement, no AVX 512, just plain neutered unless you buy Xeon.

Skylake is the worst tock of the Core line.

 

Doesn't look bad to me...



Pretty significant gains in games that are not GPU bound. Since all 4 chips are running the same frequency, its easy to see the improvement.

 

That means X-Plane 10 would benefit greatly. Sweet.