Really the only thing in terms of performance that significantly gained with the CPU are some apps with very high, out of spec memory bandwidth. I guess this is to be expected until silicon is replaced and/or there is competition for server and general single threaded IPC performance.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Skylake review thread
- Thread starter LordOfChaos
- Start date
At stock speeds it has slower cores. Overclocking it a bit would solve that problem.
Smushroomed
Banned
I5 4670k oc to 4.4 on air sitting pretty
Zeyphersan
Banned
Hmm, true. Actually I totally forgot I get a discount anyway, so I'll grab the new i7
Last_colossi
Member
What brand is your motherboard?, I'm just curious.
Why wouldnt it be? 4790K should be more than enough to push the Titan X to its utilization limit, its SLI where i would have expected some bottlenecking from the CPU. Titan X is fast, but its not extremely fast to the point of CPU bottlenecking, the 980Ti overclocked can beat it.
Numbers prove me wrong here there, so....
That "what if" game can easily lead you to a different conclusion.
What if games 3 years from now are indeed CPU-limited, but are limited by multithreaded performance? What if games 3 years from now are limited by main memory bandwidth? We know that Haswell-E significantly outperforms Skylake in both of those dimensions.
Now you have two pieces of information that lead toward conflicting conclusions. If future games are single-threaded CPU-limited, you should get Skylake. If they are multi-threaded CPU-limited or memory bandwidth-limited, you should get Haswell-E. And of course, if they are GPU-limited, it doesn't really matter, and if you own a recent processor you should stick with it.
So in this case having incomplete knowledge (i.e. only knowing about CPU-limited single-threaded performance if you only read the Eurogamer review) could easily lead you down a worse path than not having that information at all. As I said earlier, that information only helps the consumer when they also have the proper context. In this case, the context is that games are currently GPU limited, and trends like increasing desktop resolution, VR, and DX12 are unlikely to make games more single-threaded CPU-limited in the future.
I have an old ASRock Z68 Extreme3 Gen3. It's served its purpose, and I was able to maintain a 4.5ghz clock stable until recently. (That could be the PSU though, I never really bothered to diagnose it since I'd be replacing it anyways when I had the money for new parts since my build is a 2700k/670 from a few years ago).
TouchMyBox
Member
If only this were the case in Canada, where the price difference is negligible.
The nice thing about buying DDR4 now is that most people could probably bring that forward to their next CPU as well.
Suppose I agree with all of this. The conclusion is that CPU limited benchmarks are useful, but need to include both single and multithreaded tests, and include some synthetic measurements as well.
That's a long way from the claim that such tests are like benchmarking on Mars.
In my case, I have a limited budget so can't stretch to Haswell-E, especially as I am going mini-ITX! Hence the comparison is quite a bit simpler and Eurogamer's benchmarks are helpful to me.
Vulcano's assistant
Banned
Kind of late of my part, but didn't see this until now. Thank you for this.
JoeyJungle
Banned
The digital foundry for the new i7 is up, don't think it's been posted in the thread yet.
http://www.eurogamer.net/articles/digitalfoundry-2015-intel-skylake-core-i7-6700k-review
I noticed in that anandtech review, they underclocked all the processors to 3.0ghz (I assume to force the game to be cpu limited?). Maybe underclocking by ~1.2ghz makes the cpu behave strangely, that's why their results are so diffferent? Looking at the digital foundry video, the 6700k's framepacing looks pretty impressive, though not always a huge difference from devil's canyon.
http://www.eurogamer.net/articles/digitalfoundry-2015-intel-skylake-core-i7-6700k-review
I noticed in that anandtech review, they underclocked all the processors to 3.0ghz (I assume to force the game to be cpu limited?). Maybe underclocking by ~1.2ghz makes the cpu behave strangely, that's why their results are so diffferent? Looking at the digital foundry video, the 6700k's framepacing looks pretty impressive, though not always a huge difference from devil's canyon.
Where is underclocking mentioned and tests ran with underclocked CPU's? All I saw was tests with stock clocks and tests with OC.
BleedOrange
Banned
Would this be a good upgrade for my i7 930?
Well that was not a statement of equivalence. It was a reductio ad absurdum argument to the point that test results are only useful if the test conditions are realistic.
More gaming numbers, nicely formatted by Anad user Sweepr from PCLab's bench. Benches were ran by using normal gaming settings (1080p, Ultra etc.) Check links for sources and full details, but below some picked summary by me.
Used memory in benching: DDR3-2133 CL10 vs DDR4-2666 CL16
Edit: Polish bench on i5 6600K, also showing strong gaming numbers: http://www.purepc.pl/procesory/test_procesora_intel_core_i56600k_skylake_mocne_cztery_rdzenie
Used memory in benching: DDR3-2133 CL10 vs DDR4-2666 CL16
http://forums.anandtech.com/showpost.php?p=37617058&postcount=3255
http://forums.anandtech.com/showpost.php?p=37617084&postcount=3256
Edit: Polish bench on i5 6600K, also showing strong gaming numbers: http://www.purepc.pl/procesory/test_procesora_intel_core_i56600k_skylake_mocne_cztery_rdzenie
Okay, this is starting to look really cool.
I'm now excited to buy in once the six core unlocked variants come out.
sssstententen
Member
As someone still with a 955 Black edition and a 5850, I guess it boggles me to see people with 49xx cards bemoaning the lack of a huge improvement.
The 2xxx series guys I get. But if you're 3-5, you just bought it a year or two ago, whats the rush to spend another $300-600 on a new mobo/cpu combo.
The 2xxx series guys I get. But if you're 3-5, you just bought it a year or two ago, whats the rush to spend another $300-600 on a new mobo/cpu combo.
BlueSilver
Member
Anything is better than this Pentium 4 system I'm rocking now. =P
This Polish review is so much different from the rest. Here, there is quite a big difference in FPS in games compared to Devil's Canyon. Am I missing something?
Man.. Who would've thought that my first ever PC built in 2011 with the i7 2600k + HD 5970 would last me this long. Only things I've upgraded are the video card (to a 780 Ti last year) and adding two SSD's.
It certainly served me well in college and still serving me well even now.
I'm definitely keeping it as a Unity/UE4 development workstation.
From what I'm seeing though, I'm seeing enough to upgrade. That sweet i7 6700k and 980 Ti SLI is calling to me as well as VR heaven!!
It certainly served me well in college and still serving me well even now.
I'm definitely keeping it as a Unity/UE4 development workstation.
From what I'm seeing though, I'm seeing enough to upgrade. That sweet i7 6700k and 980 Ti SLI is calling to me as well as VR heaven!!
It seems like Skylake is actually a big improvement over Haswell hobbled by either the lack of quad channel memory (bandwidth) or really low latency DDR4 (latency). It might be both. It seems like the real power of this chip won't be unlocked until Skylake-E next year then. If they are more aggressive on the L1/L2/L3 cache sizes on the E refresh this chip might be a real beast.
JoeyJungle
Banned
http://anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/10
Every processor is listed as running at 3ghz.
HUELEN10
Member
So with Skylake fairing worse, at least with what's in the OP and what I've read elsewhere with graphics, should I just get a 2014 Haswell Mac Mini this fall? If a Skylake one gets announced, yields won't be too improved, and the Haswell Mini would be cheaper.
Lol gaming on a Mini aside, it just makes sense to cheap out at this point in some respects, don't it?
That's assuming the next Mini does have discrete graphics.
Lol gaming on a Mini aside, it just makes sense to cheap out at this point in some respects, don't it?
That's assuming the next Mini does have discrete graphics.
Mr. Wonderful
Member
If the Mini is using a laptop processor, I would hold off. The Mini may go back to Quad Core with Skylake, and you'll get HDMI 2.0, USB Type-C, a processor that theoretically throttles less (if that's something that happens on the Minis), and a better GPU.
The real question is, when the heck are they going to upgrade the Mini? It's not like they do that all the time.
HUELEN10
Member
With them using Haswell processors last year, a push for USB type C, and disappointment in last years model, a lot of us are indeed expecting a Mini refresh.
And throttling is actually an issue if you run any OS other than OS X on those things, Linus from TechQuickie explained it rather well once.
Their gaming numbers seem to be somewhat in line with EuroGamer, French benching site (name escaped me atm) etc. when using 2666MHz DDR4 with decent timing and their test platform doesn't seem to favor DDR3 nor DDR4 by that much.
My bad then, for some reason I thought you were referring to EuroGamer bench and missed that you named Anand. Usually CPU's are put to same line, this time via downclock, to get comparable numbers between gens as CPU's run at same frequency.
That is why you read more than just cherry[?]picked OP that OP doesn't bother to update with new data. Have a look at e.g. this; http://www.neogaf.com/forum/showpost.php?p=174830049&postcount=267 and there is more with somewhat similar performance increase in gaming alone. Skylake is improvement, but maybe not by as huge % as some expected for some reason.
Coulomb_Barrier
Member
Don't trust that Polish site, a lot of their stuff is known to be garbage.
The data we really care about is same clock 4790k vs 6700k. Skylake fails here.
ScepticMatt
Member
The real winner is eDRAM
Broadwell is consistently beating Skylake in gaming despite an 500 MHz disadvantage.
A Skylake with 128 MB eDRAM would be amazing apparently. (Using 99% frame times)
The TR Podcast 183: Diving into Skylake with David Kanter
Broadwell is consistently beating Skylake in gaming despite an 500 MHz disadvantage.
A Skylake with 128 MB eDRAM would be amazing apparently. (Using 99% frame times)
The TR Podcast 183: Diving into Skylake with David Kanter
I'm confused by this statement.
Or do you mean with "fail" that i7 6700K outperforms i7 4790K when running same speeds making Skylake solid chip and upgrade over prev-gen (that is or isn't worth of cost based on personal estimate)?
Well that's good to see. I hadn't seen that until you posted it.
Still, able to source a similar outcome from any other site?
Also preferably at higher resolutions (which is what I care about).
I have posted several bench links in this thread, some doing similar 1080p gaming tests with getting similar performance gains for i7 and i5 Skylake. Higher resolution benches mean very little when doing CPU benching because at high resolution (1440p upwards) you will very likely be GPU limited even in CPU heavy games.
We are here to bench CPU, not GPU :b
That said here few sites more "realistic situation" dGPU benches;
http://www.hardware.fr/articles/940-1/intel-core-i7-6700k-i5-6600k-z170-skylake-test.html
http://www.guru3d.com/articles_pages/core_i7_6700k_processor_review_desktop_skylake,16.html
http://www.computerbase.de/2015-08/intel-core-i5-6600k-i7-6700k-test-benchmark-skylake/7/ (no mention of used mem, may have gone way of Anand with garbage 2133's)
Low res bench to show how memory impacts Skylake perfomance in gaming; http://www.hardocp.com/article/2015/08/05/intel_skylake_core_i76700k_ipc_overclocking_review/6
So basically, a complete wash for a person in my situation. Ok thanks.
Coulomb_Barrier
Member
Yeah, seems about right.Others trying to justify the performance by quoting obscure benches from non-English language sites is a stretch.
Also, the idea that the 6700K benchmarks were hamstrung by slow frequency DDR4 memory was also debunked as iirc KitGuru ran their tests with a 3000Mhz memory kit and showed no difference: 5-10% faster CPU and zero gain in gaming over Haswell.
Judging whole release in OP because you don't personally see it as worthy investment is very biased and unfair imo. I'm not saying Skylake is mind blower, but it's far more capable chip than OP leads on and it's why state of OP frustrates me. A lot of people draw conclusion from it and don't google up additional benches.
Edit: To be clear I'm not accusing you of going to assumption from small segment of data. Just seems that a lot of people do.
If bench site isn't in English language it should be ignored as obscure and untrustworthy as they dare type in different language than your native lang? Oh wow. Language of site shouldn't matter if benches are solid work and how they got to results is laid out well.
And other sites show that Skylake gains performance and beats Haswell when running faster DDR4 with reasonable CL rating, 15 being most common atm. I think one bench telling X or Y shouldn't lead to situation where all other benches are ignored or stamped as "invalid" because favorite site says otherwise.
Edit: Sorry for double!
Prophet Steve
Member
You mean a complete wash because a a lack of higher resolution benchmarks or because you think the difference is too low?
The first argument does not make a lot of sense. A better CPU will last longer for everybody.
If bench site isn't in English language it should be ignored as obscure and untrustworthy as they dare type in different language than your native lang? Oh wow. Language of site shouldn't matter if benches are solid work and how they got to results is laid out well.
And other sites show that Skylake gains performance and beats Haswell when running faster DDR4 with reasonable CL rating, 15 being most common atm. I think one bench telling X or Y shouldn't lead to situation where all other benches are ignored or stamped as "invalid" because favorite site says otherwise.
Edit: Sorry for double!
It's just you in both this thread and the other one referencing cherry-picked benchmarks to try and make yourself feel better about your purchase. It's just very obvious, to the point where I've even started to skip your posts just because of the bias you're showing.
I have nothing against the skylake chips. I wanted them to be amazing so I had a reason to upgrade. But they aren't amazing. They're just an average step up from the previous generation. Whether you like it or not, that's disappointing to me and others in this thread who were hoping for better than just an average step up.
One Punch-Man
Banned
See to the average gamer, no one really cares. They just want the best fps possible.
Well, my choices are 6700k or 5820k.
Given the kind of differences we're seeing, they're just too marginal.
I do rendering work, so I might as well go extra cores for that.
Unless the 6700k yields a better upgrade path for my purposes than 5820k (i.e. Skylake-E brings significant and worthwhile improvement, while 2011-3 is deprecated after this current iteration of cpus). But no one has information on that, so I might not worry about it for now.
I have posted full lists of benches for people to read up and draw own conclusions. People kept complaining about gaming performance so I started to link that stuff specificially. A lots of benches out there show gaming performance increases in CPU heavy games, by how large % varies from bench to bench. I would argue that people who keep repeating how Skylake is maybe same and downgrade mostly are ones that are biased.
When Intel has delivered "amazing mind blowing huge performance increase in double digits" last? When looking at benches as a whole, not just gaming performance, Skylake looks to be solid Tock with realistically expected performance increase.
Prophet Steve
Member
Yeah, in that case that should be a perfectly fine choice.
Resolution does not have an impact on CPU performance, so you just have to look at the 980 ti performance on that resolution. Or at whatever CPU benchmarks are provided for the games if you want to know what the limit of the CPU is.
Notor!ousG
Member
I upgraded my i7-950 last friday... leap couldn't get bigger .
Very happy with the results so far.
.Very happy with the results so far.
At 1080P on my 970 I got almost 60fps solid on Crysis 3. At 1440P on my 980 ti I get drops to 30/40 (same CPU - i5 2500 non k)