[Various] Intel Core i9 (14nm Skylake-X)

[StyM]

AnandTech -> http://www.anandtech.com/show/11550/...7-7800x-tested
Bit-Tech -> https://www.bit-tech.net/hardware/20...hipset-revie/1
GamerNexus -> http://www.gamersnexus.net/hwreviews...ng-vr-premiere
Guru3D -> https://www.guru3d.com/articles-page...-review,1.html
HardwareCanucks -> http://www.hardwarecanucks.com/forum...0x-review.html
HotHardware -> https://hothardware.com/reviews/inte...40x-cpu-review
Hexus -> http://hexus.net/tech/reviews/cpu/10...4nm-skylake-x/
LinusTechTips -> https://www.youtube.com/watch?v=3w-G85O8n3I
Overclock3D -> https://www.overclock3d.net/reviews/...ake_x_review/1
PcPer -> https://www.pcper.com/reviews/Proces...ocessor-Review
TomsHardware -> http://www.tomshardware.com/reviews/...ke-x,5092.html
TweakTown -> http://www.tweaktown.com/reviews/822...iew/index.html

[chessman]

Also another review is here

[vario]

Ive read both.
I was expecting that clock for clock, skyx can be slower than bde at times.
BUT, its worse than i thought. Even with clock advantage, it STILL can be slower.And with higher temperatures thanks to the thermal paste!
If someone pays top dollar for a new high priced platform, they at least should not be slower (yes i know, only at times).
Im hoping were gonna get at least V2 of this with solder.As it stands. Nobody with 6900K or 6950X should have the urge to change...Well not without a delid.And i was hoping that 6900K and 6950X would radically drop in price...

Peeps, look at the memory ops speed. Somethings rotten with skyx, and thats a score with skyx at 2666 and BDE at 2400...

[AliG]

I was surprised Intel decided to reduce the cache. I get they optimized it quite a bit for Skylake, but many applications benefit from just physically having more.

[sergiu]

Skylake has 2.375 MB cache (1MB L2 + 1.375MB L3), Broadwell 2.75 MB (256KB L2 + 2.5MB L3) and Ryzen 2.5 MB (512KB L2 + 2MB L3). For heavily threaded applications, this should translate in better average memory latencies due to larger L2 that might compensate for lack of L3. This looks to be designed to perform very well in heavily threaded, data intensive applications while trading single thread performance (poor single/light threaded applications always benefit from more L3). I'd not rush to judge it until I see WinRAR & 7zip scores. Those are much better representatives for server loads.

[drmrlordx]

Overall a bit weird, and with TR coming out . . . kind of an ouch for Intel.

[qcmadness]

Skylake has 2.375 MB cache (1MB L2 + 1.375MB L3), Broadwell 2.75 MB (256KB L2 + 2.5MB L3) and Ryzen 2.5 MB (512KB L2 + 2MB L3). For heavily threaded applications, this should translate in better average memory latencies due to larger L2 that might compensate for lack of L3. This looks to be designed to perform very well in heavily threaded, data intensive applications while trading single thread performance (poor single/light threaded applications always benefit from more L3). I'd not rush to judge it until I see WinRAR & 7zip scores. Those are much better representatives for server loads.

I expect the larger L2 cache is to reduce the burden of inter-core L3 bandwidth and traffic (ringbus > mesh).

[AliG]

I understand the benefits of increasing L2 to reduce memory latency. But why not expand L3 on the Skylake-X parts to match Broadwell-E?

The only thing I can think of is it justifies their 6950X pricing. The 7900k is much cheaper (at launch), but doesn't necessarily win across the board.

[qcmadness]

I understand the benefits of increasing L2 to reduce memory latency. But why not expand L3 on the Skylake-X parts to match Broadwell-E?
The only thing I can think of is it justifies their 6950X pricing. The 7900k is much cheaper (at launch), but doesn't necessarily win across the board.

You still cannot address the issue of latencies for mesh-L3 cache.

[StyM]

[StyM]

[StyM]

[StyM]

[StyM]