Intel Skylake / Kaby Lake

[Bouowmx]

Are memory bandwidth (frequency) and latency independently tested? Because typically, 2133 MT/s 15 cycles (1.2 V): 14 ns, and 3000 MT/s 15 cycles (1.35 V): 10 ns.

 

[Head1985]

Ram have same timing cl 14-15-15-35 1T.I also downclock cpu and cache at 3.5Ghz.So its not like cpu is at 3.5ghz and cache at 4.5ghz.

 

[Bouowmx]

Thanks for the info. Bandwidth and latency were not separated. So, comparing "fast" and "slow" RAM gives expected results.

Another interesting test is to keep latency the same but change bandwidth: 2133 MT/s 14-15-15-35 1T and 3000 MT/s 20-21-21-21-49 1T.

 

[LTC8K6]

Not exactly Skylake / Kaby Lake related, but still interesting:

[Various] Overclocker discovers Xeon E5 V3 Errata, Engineers exploit to unlock Turbo



http://www.xtremesystems.org/forums/showthread.php?293290

It's been known for ages that you can get many Intel chips, including Xeon chips, to run all the cores at the single core turbo multiplier. Heck, I've been mentioning it for years.

 

[DrMrLordX]

Does that work with Broadwell-based Xeons though? Or Skylake ones?

Might not make that much of a difference what with the single-socket Xeons using different sockets now.

 

[Head1985]

GTA5 test:
first- GTA5 dont benefits from HT at all.HT works only in benchmark but i test in real gameplay.And it was not GPU limited.GPU was at 80-90%.
second -not that great memory scaling.Only 10% perf increase.And because of that 3500mhz cpu with fast ram is slower than 4500mhz with slow ram.
CPU is still same 6700k

4500mhz+2133mhz ddr4 vs 4500mhz+3000mhz ddr4

3500Mhz+3000mhz ddr4 vs 4500mhz+2133mhz ddr4

HTon vs off

test scene(1920x1080 all max and advanced all max no AA very high grass) https://www.youtube.com/watch?v=G4aEWP2pAEE

 

[Sweepr]

It's been known for ages that you can get many Intel chips, including Xeon chips, to run all the cores at the single core turbo multiplier. Heck, I've been mentioning it for years.

Totally missed it then. Still, very interesting considering how cheap some of these Xeon models are on eBay.

 

[arandomguy]

If only non Z boards allowed faster memory than stock. Would consider a 7700 (non-k) and non Z board.

Curious if anyone knows a review that actually directly compares the 7700 k vs non k or 6700k and non k purely at stock speeds? Along with power consumption numbers?

 

[dullard]

Curious if anyone knows a review that actually directly compares the 7700 k vs non k or 6700k and non k purely at stock speeds? Along with power consumption numbers?

I honestly don't see the point in the 7700 chip. If you need the 8 threads or the 8 MB L3 cache, then the 7700k is ~10% faster than the 7700 for not much more money. If you don't need the 8 threads, then the 7600 is ~2% slower clock speed and 40% cheaper. The 7700 (non-k) just doesn't have many winning situations. That is, unless you need lower power and lots of L3 cache, but how often is that what customers need?

 

[coercitiv]

The 7700 (non-k) just doesn't have many winning situations. That is, unless you need lower power and lots of L3 cache, but how often is that what customers need?

Was it not your opinion that the i7 wins over the i5 in games especially due to the extra L3 cache?

 

[dullard]

Was it not your opinion that the i7 wins over the i5 in games especially due to the extra L3 cache?

It is.

But are there many people who need high power gaming desktops but can't have 26 more Watts of TDP? Laptop yes, but these aren't laptop chips. With the desktop I just don't see it being that common. The 7700k (currently $299 at Microcenter) is just better for gaming and doesn't cost much more than the 7700 (currently $289 at Microcenter). If there was a real price difference, then the 7700 would be a chip that I'd consider. But there just isn't.

 

[coercitiv]

It is.

Then how is it that 7600 is, in your view, only 2% slower than 7700? The poster you replied to is clearly looking to use it in games or some other memory intensive apps.

 

[dullard]

Then how is it that 7600 is, in your view, only 2% slower than 7700? The poster you replied to is clearly looking to use it in games or some other memory intensive apps.

I just posted clock speed differences which are 2%, since I didn't know the use case. I will edit it to say if you need 8 threads or 8 MB L3 cache then the 7700k is just so much better. I still stick with the point that the 7700k is just so much better for almost no price increase which makes the 7700 nearly useless for the vast majority of people.

Cut $30 to $40 off and the 7700 makes sense. It is just the most mis-priced intel consumer chip.

 

[Sweepr]

According to Digital Foundry a stock Core i7-5960X (up to 3.5 GHz) from 2014 is ahead of Ryzen 7 1800X (up to 4GHz+) in CPU limited gaming tests. Meanwhile Broadwell-E (especially OC vs OC) can post a significant lead in some titles, like Ashes of the Singularity and Rise of the Tomb Raider (>40% faster). i7-7700K, aside from some Crysis 3 scenes, delivered better framerates across the board and got DF's recommendation for gamers. Bodes extremely well for Skylake-X with new cache structure launching in August. I hope the entry-level 6C/12T part is agressively clocked and priced. CPU comparisons are fun again.

http://www.eurogamer.net/articles/digitalfoundry-2017-amd-ryzen-7-1800x-review

 

[Ajay]

Great. More Ryzen posts in this thread

 

[inf64]

Great. More Ryzen posts in this thread

It's ok, 4.4Ghz OCed 6900K that costs 1K dollars is 25% faster than stock 1800 that costs 1/2 of that, even counting games that obviously have issues with thread scheduling on Ryzen . OC Ryzen 10% from all core turbo of 3.7Ghz and difference drops to ~13%. Bodes well for X399 platform , hopefully they can reach the same Turbo clocks on those SKUs as they can with 1800X.

 

[Sweepr]

A $400 6C/12T Skylake-X >4.5 GHz overclock (probably doable) will leave any Broadwell-E behind in games inf64. And that will be a direct competitor to Ryzen's fastest SKUs, whether you like it or not. Point in case, Intel HEDT still has a significant advantage in certain titles even with older Haswell/Broadwell cores (more overclocking headroom on top), next generation will extend the lead.

 

[Arachnotronic]

A $400 6C/12T Skylake-X >4.5 GHz overclock (probably doable) will leave any Broadwell-E behind in games inf64. And that will be a direct competitor to Ryzen's fastest SKUs, whether you like it or not. Point in case, Intel HEDT still has a significant advantage in certain titles even with older Haswell/Broadwell cores (more overclocking headroom on top), next generation will extend the lead.

I really look forward to seeing how Skylake-X performs...IPC and clocks. My expectations are pretty high, but we'll see.

 

[Azuma Hazuki]

Inf64 said what I was going to but better.

People...perspective. For the first time in almost a decade, we have a competitive marketplace again, and that is very possibly the most important thing that came out of this. Intel has been arrogant and high-handed toward its customers for entirely too long now because of its advantage, and they got there more by scheissty shoulda-been-illegal tactics than skill until Penryn hit.

AMD has pulled off a miracle here. Do remember that most of the benchmarks showing Intel's HEDT chips beating the 17/1800(x) series are comparing chips that cost $500-$1,700 to those that cost $250-$500 respectively.

So celebrate Intel all you want, I'm looking forward to Skylake-X too, but do it for the right reasons: the return to a competitive marketplace. If the more...ahem...enthusiastic Intel boosters ever actually got their way, the entire market would be a monopolistic disaster.

 

[inf64]

I would like to see SKL-X 12T parts that cost 400$. They will go against unlocked Ryzen 1600X (up to 4Ghz Turbo) that costs 250$. Even if it has 20% higher fps max OC vs max OC, it will be REALLY hard to justify the cost for such a system,especially knowing that B350 boards that cost peanuts can push 8C R7 1700 to stable 3.8-3.9Ghz. Fun times ahead

 

[Sweepr]

I would like to see SKL-X 12T parts that cost 400$.

Unless you're suggesting there's no Core i7-6800K successor, you will.

Even if it has 20% higher fps max OC vs max OC

Broadwell-E already leads Summit Ridge (with HT off) by ~20% in games according to Hardware.fr. Skylake-X at iso clocks and core/thread count will go past that. And that's before you factor any possible frequency advantage for the latter in OC vs OC comparisons.

It will be REALLY hard to justify the cost for such a system,especially knowing that B350 boards that cost peanuts can push 8C R7 1700 to stable 3.8-3.9Ghz. Fun times ahead

Ryzen 7 1700 is an interesting choice from a price/performance POV, but rest assured many will not bother to pay an extra ~$150-200 for a motherboard + CPU upgrade that lasts for years if that provides more consistent gaming performance in CPU limited scenarios today and in the near future. Fun times ahead for both camps.

 

[Ajay]

Ahh - stop!!!! There are like 40 Ryzen threads already. Go post in one of those

Screw it - unsubbed.

 

[raghu78]

Sweepr you can always pick a review which suits your narrative but we are seeing less than 10% between 6900k stock and 1800x stock (With High performance profile and HPET off).

https://www.computerbase.de/2017-03/ryzen-windows-7-benchmark-core-parking
http://www.techspot.com/review/1348-amd-ryzen-gaming-performance/page5.html

Granted the 6900k has higher OC headroom and will widen that gap to 20%. But in the bigger scheme of things a person with a limited budget who wants to game and does content creation has no choice other than Ryzen. You could build an entire PC with R7 1700 and GTX 1070 for roughly the price of 6900k. For gamers who do serious work on their PC a 1700 or 1700x OC to 3.9-4.0 Ghz is definitely better than a 7700k. 7700k makes sense only for the 1080p 144 Hz gamer who does not do any serious work on his PC. For all other cases Ryzen is the better CPU including gamers who livestream .

eurogamer says the same thing too

http://www.eurogamer.net/articles/digitalfoundry-2017-amd-ryzen-7-1800x-review

"It's also worth stressing that while gaming performance does lag behind Intel by and large, this does not make Ryzen 7 processors 'bad for gaming'. The Witcher 3's Novigrad City is one of the most processor-intensive areas of the game and the 1800X is still delivering a 119fps average. Our results show that the i7s AMD is targeting with Ryzen 7 are potentially longer-lasting for gaming - a key factor in any CPU purchase - but in the here and now, it's abundantly clear that Ryzen still does the job. Clearly PCs are used for more than just gaming, so if play is just one part of your requirement from a processor, Ryzen's undoubted capabilities elsewhere may make it the more logical choice than an i7 - and at these prices, the Ryzen 7 line's excellent overall performance simply cannot be ignored"

 

[Sweepr]

Already made my point Ajay. Now here's the latest Skylake-SP leaks @ Geekbench 4.

Intel Xeon Platinum 8160 @ 3.70 GHz
2 processors, 48 cores, 96 threads

Intel Pentium II/III @ 3.20 GHz
2 processors, 56 cores, 112 threads

https://browser.primatelabs.com/v4/cpu/2144813
https://browser.primatelabs.com/v4/cpu/2144889

 

[mikk]

So some websites are assuming this CPU is based on Skylake-SP:

http://browser.primatelabs.com/v4/cpu/2119144



Cache sizes and family/model name don't match though. Could it be a custom Haswell SKU?

GenuineIntel Family 6 Model 63 should be Haswell based and in your last posting we can see a CPU string from a Purley based Skylake: GenuineIntel Family 6 Model 85

Also the L2 Cache size is correct in this: L2 Cache 1024 KB x 28

So we can safely assume this wasn't a Skylake based server SKU.