Does CPU-Z actually 'read' stuff like that, I thought for most such details it basically just does a database lookup of whatever details it knows about the chip based off Family, Model and Stepping (plus a few other factors) and tries to find the best match?
Found something interesting in the comments section of some random HardWare.fr contest:
This was posted by Doc TB (Samuel) of CanardPC, formerly x86-secret.com.
He's basically saying that he has an R0 stepping of Skylake, nominally running at 2.6/3.4GHz (Base/Turbo) but in practice always stuck at 3.3GHz. He says that performance is "strange" and that paradoxically, the A0 stepping [that presumably he's played with too] behaved more logically [more consistently?].
In any case, the chip is very stable, and performance isn't exactly spectacular. He hasn't been able to test the IGP yet.
http://www.hardware.fr/news/14211/concours-18-ans-reponses.html#commentaires → page 7.
This was posted by Doc TB (Samuel) of CanardPC, formerly x86-secret.com.
He's basically saying that he has an R0 stepping of Skylake, nominally running at 2.6/3.4GHz (Base/Turbo) but in practice always stuck at 3.3GHz. He says that performance is "strange" and that paradoxically, the A0 stepping [that presumably he's played with too] behaved more logically [more consistently?].
In any case, the chip is very stable, and performance isn't exactly spectacular. He hasn't been able to test the IGP yet.
Blazkowicz
Legend
The Skylake K is announced at 95 watts, while non-K are Skylake-S at 65 watts and also -T models at 35 watts.
So, 6700K is at 95W while 6700 is at 65. (doesn't mean 6700K uses up the full 95 watts)
65 watts is kind of that magic number for OEMs, it's good for a stock heatsink fan too while the 35 watt seems to obviously work in SFF (slim micro-ATX please!)
What maybe changed is we were kind of supposed to only get Skylake-S CPUs, not the K ones yet.
So, 6700K is at 95W while 6700 is at 65. (doesn't mean 6700K uses up the full 95 watts)
65 watts is kind of that magic number for OEMs, it's good for a stock heatsink fan too while the 35 watt seems to obviously work in SFF (slim micro-ATX please!)
What maybe changed is we were kind of supposed to only get Skylake-S CPUs, not the K ones yet.
There's a function in x86 CPUID instruction which returns cache and TLB information including size, set associative, etc. So if CPU-Z gets these data this way, it should be accurate.
Yikes even at only 2.4Ghz that CPU would pack over 4 TFLOPS!
Very disappointing results tbh. Especially considering this is compared the 2 generations old Haswell rather than Broadwell so this IPC difference of 6.9% is what we get from 2 architecture revisions, not just one. Add to that that the biggest performance jumps appear to be attributable to the IGP - Cinbench R15 (OpenGL) and Sandra 2015 or merely memory bandwidth tests that would naturally be higher on account of using faster memory and the results are actually pretty pathetic.
Sure the 4790K has a 200Mhz boost advantage but that's no excuse - I mean why the hell should the 14nm 6700K with a 95w TDP have a lower top clock speed than the 22nm 4790K with an 88w TDP?? Oh and no AVX 512 either and naturally we are still stuck at 4 cores. Outside of the IGP it seems Intel is going backwards on the CPU front. At least my 2500K won't need upgrading for a few more years. Zen can't come fast enough.
It only really needs Sandybridge level IPC if it can hit clock speeds around 4Ghz and come with 8 cores at a mainstream price. That would be a very compelling product vs Skylake.
UniversalTruth
Veteran
Indeed. An 8-core Zen FX will come with 16 threads. I think a 4-core 8-thread Skylake will have terrible times to compete with these.
Intel will inevitably need to put the igp-free chips to lower market segments in order to compete adequately.
Intel is sleeping right now, because AMD refused to be a real competitor with those several-old 32-nm FXs.
Should feel lucky that still have those profit margins. To enjoy them while they are still there.
DX12 should also greatly help scaling across multiple CPU cores which will hopefully start to make Intels old quad core standard look very outdated, very quickly.
This is something that needs to be repeated every time someone benchmarks an ES chip: Determining the top clock speeds and power envelopes for retail chips involves a complicated process of speed and power testing and binning. Typically absolutely none of that is done for ES chips -- they just pick some random clock speed/TDP that the chip will certainly work at and call it a day. ES chips that leave Intel/AMD are not meant for benchmarking and have no need to be fast to work in the tests that the people they are given (typically board partners) need to do, so there is simply no need to rate them at proper clock speeds/power.
Absolutely the only speed numbers that are interesting for ES chips are IPC and other mostly clock-speed independent measures. There is no reason to expect that 3.3GHz is the top what skylake chips can do at based on this chip. There is no reason to expect that 3.3GHz is the top what this specific chip can do while staying within the parameters Intel considers acceptable for stock speeds.
A long time ago, when AMD introduced Hammer, the initial batch of ES chips that went out were all rated at 800MHz. This caused some amount of sky-is-falling-for-AMD rumormongering before chips with higher speeds came out. There just isn't any reason to rate ES chips at competitive clock speeds.
8 sandy-comparable cores at a mainstream price? Color me sceptical, but maybe AMD doesn't want to make any profit off of the chip; it COULD happen...
Frankly I'd expect AMD to torpedo its profit margins before I expect zen's performance to rival Intel's... *shrug* Maybe I'm being too pessimistic, who can say. At this point though I must admit I'm wondering if zen is actually going to get released. AMD has a disturbing tendency to just cancel products these days because of lack of resources/funds. Maybe this project will be prioritized. It seems to be somewhat of a hope-filled symbol which company management clings to, heh.
UniversalTruth
Veteran
You surprise. But nothing good.
It's like saying that they will cancel Radeons' evolution.
It's like saying that they will cancel Radeons' evolution.
