Skylake: Intel's Core i7-6700K And i5-6600K

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
I'm wondering if we'll get a desktop Skylake with Iris Pro at all. Broadwell was an exception. A dedicated Broadwell desktop die was never developed. The i7-5775C and i5-5675C are actually Broadwell-H (mobile) CPUs stuffed into a desktop package. Maybe we'll get lucky and Intel will do the same with Skylake-H, but it's not like they could have any data on the Broadwell desktops that show it would be worth doing again for Skylake.

Interestingly, I heard a rumor that they're finally going to stuff a Skylake H-series quad core with Iris Pro (72 EUs !!!) into a NUC. If there's any reason to want kick-ass integrated graphics in a desktop, an ultra-small form-factor PC is it.
 

Assuming the table on this page on wccftech is accurate, we can expect a Skylake-S/K with 64MB eDRAM at some point in the future. Don't wait for 128MB eDRAM desktop variants though, since the 128MB models will apparently be BGA-only.
 


Thanks for the links. I feel more informed now.

The biggest differences in any of these reviews takes place when frame rates exceed 100 fps for all processors listed. For the most part, the differences are a fraction of a frame up to a few frames at most. FCAT differences are around 3ms max.
 


Nothing here seems to support what you are saying. Most games seem to be at the same performance level across all the CPUs, which I would expect of most Intel CPUs

You made a statement like the Broadwell CPUs had a massive advantage over Skylake and while they do for the IGP they do not for gaming as of yet.That might change if the developers start to develop using the ability in DX12 to utilize all GPUs to render frames but as it stands nothing does that right now so the L4 eDRAM is for the most point useless in a high end gaming system with a discrete GPU.

That is what we are, have and will be seeing for the next few years. Once DX 12 becomes more highly adopted and games start utilizing the feature set then we can worry about having a better iGPU if it truly will benefit us in the same way.

Remember a lot of synthetics show great promise until it comes to the real world where we get little to no benefit in comparison.

I could just as easily pick and choose benchmarks showing that Broadwell and Skylake for the most part perform the same in most games and I would have vastly more to back that up than does your point, which most are fractions of FPS which is considered margin of error.
 
This could be very big. Overclocking seems to not be affected at all. Granted, it's 100 MHz lower than a stable Haswell-e, but Tom's might have gotten a lower end chip. As the manufacturing process gets more mature, expect the numbers to go up by 100-200 MHz.

Enthusiasts mustn't look at just overclock speed, though. Broadwell had an approximately 5% ipc over Haswell. I suspect that Skylake will have about a 10% over Broadwell, a total of 16% ipc increase over Haswell. That makes a 4.7 GHz Skylake overclock equivalent to 5.4 GHz Haswell overclock or a 6.4 GHz Sandy Bridge overclock. Raw bandwidth continues to increase with decent results.

I'll withhold my judgement on whether it's a massive improvement until I see the microarchitecture, but the results look promising. Something with Windows 10 is helping out Skylake. Maybe Microsoft and Intel collaborated to improve performance?

The graphics look just like another average generational improvement. It is an improvement, though, and not to be bashed. I'm sure that Intel will release an Iris Pro i7 with a 95w tdp.

Overall, I'm satisfied. I was really worried that the die shrink would limit clock rates and hamper performance increases in other ways. I just wish that Intel would do something big to the architecture. I mean, i5's have always had 6MB of cache and 4 cores. The only changes have been tweaks in microarchitecture, clock rates, and integrated graphics. That's just one example. Nothing big has been changes, just little tweaks here and there. I guess if Intel goes all boss on its CPU's, then AMD would be completely roasted and Intel would be broken up because they would have a monopoly.

I can't wait to see Skylake-e and Xeons. If they can fit 18 cores on a Haswell Xeon, then they should be able to fit 24 cores in a Skylake Xeon.
 
I'm so severely disappointed! Only 4 cores, instead of the 8 as my previous-generation i7-5960X has? !!! Why would I replace my 8-cores and downgrade to only 4 ?
 
Oh Intel. You're drunk. Go home.
Who in their right mind would buy a Skylake processor at 350 when the price is similar to the 5820K that has 6 cores and can reach 4.5Ghz in most cases. Given the push on DX12 for spreading the workload to more cores (more than 4) I don't see any compelling reason to buy this processor, especially if that someone is considering CF/SLI.

Realizing this, I should probably go buy a 5820K now before they replace it with something that has a higher price at a marginal performance difference.
 


I'm glad it's on their planned list then. It originally looked like they were just going to stop GT4e; knowing that it might cripple AMD even worse.
 


I agree. Paid $425 for my 5930k, and I am very happy I did not wait for Skylake.



 


I actually posted the worst link of them all first to see if you would fall for it. You did. Pathetic. Stop wasting my time if you cant even be bothered to read FACTS stated from several different sites.
 
To all the AMD haters that knew how to buy only Intel and recommend only Intel.. you are already seeing how slow the CPU market is going now... and it`s going to be worse.
 
Good upgrade for 1st Gen i5/i7 users. Though I think they targeted it at the 2nd Gen i5/i7 users, doesn't seem like a huge improvement for them though.
Click to expand...
meh.. we will see. my 980x still runs quite nicely. And I love having 12 threads to throw at various applications all at once.

If the benchmark programs weren't so $$ i could get the same ones and run on my system for a 1:1 comparison.

but the chipset upgrades sound like they have possibilities.
 
Defiantly an interesting read. Not a big improvement in speed, but the ability to use DDR4 RAM and have multi M.2 SSDs is a nice feature.

Skylate defiantly has some advantages, namely lower heat but don't let the TDPs fool you. For you readers thinking, "oh no, more TDP! What are you thinking intel!" don't be so quick to come to conclusions. TDP isn't the only area that tells if the CPU is sucking up a ton of power or not, you have to see what voltages is being put into the CPU as well. Plus, this is an enthusiast class CPU so having less heat but a little more TDP is far more acceptable than vise versa.

I am defiantly pleased with Skylate so far, however I do have some disappointments:

Intel seems to be loosening up their speed percentages every new generation, it's not like before when AMD was good competition and Intel released very fast CPUs compared to previous gen. I would like to see a 30% increase in speed from the previous generation so that users that even have Ivy Bridge have a decent upgrade from their systems.

Would it be wise to upgrade from haswell? No in most circumstances. If you have a nice i5 K SKU and a Z97 motherboard there isn't really a reason to upgrade unless your a tech junky that desires all the new toys. But, if your a content creator then the Z170 boards and a i7 broadwell with DDR4 overclocked might be a good upgrade.

However, besides my disappointments I'm still pleased with Intels new cpus. It just needs some competition from AMD Zen before I'm happy at the pace they keep pumping out new stuff. 😀
 
Parts of Adobe CC are GPU accelerated, which could explain the 7850K's performance (and come to think of it, the other APUs) in Illustrator.

From the looks of things, if you're just talking CPU performance, Haswell and perhaps Ivy Bridge owners shouldn't feel that they need to upgrade, but it definitely sounds worthwhile if you've got Sandy Bridge/FX and below.
 

Such major performance leaps have always been attached to drastic changes in CPU architecture that fundamentally changed the execution pipeline. The biggest such change since Core2 was integrating the memory controller which yielded a ~60% IPC improvement largely from slashing DRAM controller latency in half.

When branch prediction, caches, out-of-order execution and other fundamental tweaks like those were new, incremental improvements to them produced several percents worth of overall performance gain. Now that all of those structures are mature and just about as advanced as modern fabrication processes can accommodate, the same amount of effort put in improving them only yields fractions of a percent in further performance improvements. If pushing IPC further was so easy that Intel could still squeeze out 30% leaps per generation, AMD should not be struggling so much to achieve better than 10% improvements per generation either.

In mature CPU architectures, per-core throughput is fundamentally limited by the data-dependent nature of software: the CPU cannot execute instructions any faster than the software's data dependencies can be resolved.
 
AMD's inability to post notable IPC increases for the Bulldozer derivatives is partly (mainly?) down to the fact that the architecture demands a large amount of resource sharing, and whilst adding L1 cache, improving prediction, enhancing decoders, supporting new instructions and such like can help, it's still hobbled by a very handicapped memory subsystem, not to mention a huge gap to Intel in terms of production node. You can't very well tear each underperforming part of an architecture out; there comes a point where you need to start from scratch, and that's what they're doing with Zen.

It also has to be considered that AMD designed Bulldozer to be, essentially, as fast as Phenom II per core but offering up to double the performance overall. Additionally, they bet that the FPU would become less important. Going back to a dedicated FPU per core would probably do a lot for performance. There were people who wanted a 32nm Phenom II X8 instead of Bulldozer, presumably because of that reason. Whilst AMD could theoretically throw out a simple rehash of K10, wouldn't it stand to reason that they would have a fair idea of how to significantly improve IPC over a core family whose origins go as far back as the Athlon?

Just my thoughts on the whole AMD-stuck-in-IPC-rut thing.
 
What the heck Intel? So, you provide great integrated graphics into Broadwell, then nerf it for Skylake? I guess you had to find a way to help sell your 'paper launch' of Broadwell. I really hope Xen makes you guys wake up; although it more than likely won't.
Click to expand...
Why you complaining. This is the highend . Overclocker. If you Want Skylake refresh With K models you have to wait for the answer to devils canyon. in the Kabylake update . My wifes brother inlaw says this one is the one to buy . Why Its name is likely going to be Angel Canyon . This one will have GT4e 64 bit. Its won't be generation 9 Igpu either it will be Gen 10 . . He says this one should be 100% faster than Broadwell c He did say up to 150% faster . But I find that hard to buy with only 64bits of Edram Use DDR4 3800 and you should be set . I myself will stay with 1080 p Res. By the way Broadwell c was not a paper launch . You just can't buy one in the states . Others can buy, So no paper launch . You can also BUY PCs prebuilt with Broadwell C in the states. Skylake 530 Beats AMDs APU. What more do you want . Buy a Dgpu like you likely say you would anyway. The Skylake GT3e and gt4e are on the way. As well as an xeon with gt4e graphics 2x 128 edram on socket 1151. The way your complaining maybe you should just Buy AMD . Be Happy!
 


Ahhh the personal insults.

I did read through each site and most games the 5775C and 6700K are pretty neck in neck (as are the other 4 core or better with them). A few show slight variances (within margin of error) and only in the Project Cars at 800x600 on the German site do we see Broadwell show any sort of advantage over Skylake but without further testing there is nothing to back up that the eDRAM is the cause of it because they didn't test any other chips with the eDRAM cache against it they just assumed.

I miss the days when people could actually have discussions followed up by actual facts.



Upgrading to anything from Sandy Bridge or newer from a Q6600 would be a good idea. A 2500K would easily crush a Q6600. It is just a much better uArch than Conroe was.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom