More Skylake-X and Kaby Lake-X details tip up

[Katanga]

I'm waiting for Frozen Lake, those CPU's are going to kick some serious ass.

 

[Silent_Scone]

We don't know, ss is just making **** up again.

You gathered that from my post on platforms that are already out?
Or was it my post regarding leakage that confused you? Posting passive aggressive sarcasm will only get you so far. It makes you look simple.

We've seen Ryzen IPC comparisons to BWE. If you can't work out it's slower than SKL from that then I don't know what to tell you.

It's fine, though. Most people's ideas of why KBL is 'identical' is because their understanding of these things is elementary at best. (Some regurgitated medicine for you)

Surely we don't know the IPC with what Ryzen is doing. Other than AMD stated they wanted 40% greater than their previous line of CPU's and they stated they are already beating that figure back in mid December so with that what can we expect it to be upon release.

I have some sources stating that they are just over the 50% greater IPC to their previous line and match up to Intels current IPC's. Just info I have been passed on/read accordingly.

How that matches up to the true info and how this release from Intel will affect what Ryzen does will be more interesting but I don't see it as much of a threat due to the price that Intel will be bringing these chips in at.

It's not true, Ryzen does not match KBL IPC

 

[Silent_Scone]

Dp

 

[shamus21]

Looking like more of the same, a pointless upgrade.

 

[DragonQ]

We've seen Ryzen IPC comparisons to BWE. If you can't work out it's slower than SKL from that then I don't know what to tell you.

We saw two benchmarks, where it matched the Broadwell-E CPU in Blender and beat it in Handbrake by ~9%. Skylake is also 9% faster than Broadwell-E in Handbrake benchmarks, so for that use-case at least it could well be as fast as Skylake/Kaby Lake (in terms of IPC).

This is all multithreaded though, we don't know how it'll perform in low-threaded applications yet.

 

[Silent_Scone]

We saw two benchmarks, where it matched the Broadwell-E CPU in Blender and beat it in Handbrake by ~9%. Skylake is also 9% faster than Broadwell-E in Handbrake benchmarks, so for that use-case at least it could well be as fast as Skylake/Kaby Lake (in terms of IPC).

This is all multithreaded though, we don't know how it'll perform in low-threaded applications yet.

Can you tell me how you've managed to conclude that Skylake is 9% faster than BWE in Handbrake?

 

[Curlyriff]

Can you tell me how you've managed to conclude that Skylake is 9% faster than BWE in Handbrake?

As far as I am concerned about as much of a conclusion as you have had with that of Ryzen compared to anything Intel at the moment.

You've stated that a chip that isn't released or have all the details is going to be slower compared to Skylake based on the fact that Ryzen as pointed out previous outpreformed the current Intel Broadwell by 9% in handbrake.

Can you conclude how you have managed to suggest that Ryzen will be slower than Skylake because of such info. And thus point where the conclusion of Skylake is greater than 9% faster than Broadwell.

I am not saying the statement of Skylake being 9% faster is correct. Just want to know how you got the opposite end of the viewpoint?

 

[Curlyriff]

Looking like more of the same, a pointless upgrade.

Yep I am not seeing a huge amount with this refresh going on but will be interesting to see what gains there are and compare that to what Ryzen is doing then.

I think that would lead us to know really how AMD are fairing against what Intel is pushing out.

 

[DragonQ]

Can you tell me how you've managed to conclude that Skylake is 9% faster than BWE in Handbrake?

Broadwell-E has more cores than Skylake so will obviously be faster but we're talking about IPC here. Broadwell-E uses the same core architecture as Broadwell (hence the name), which means it has the same IPC. There are dozens of comprehensive reviews on the internet of Skylake and most include a Handbrake benchmark showing Skylake is ~9% faster than Broadwell. I checked 4 of them whilst writing that post.

If you really want me to copy and paste the links from Google, here you go:

http://techreport.com/review/28751/intel-core-i7-6700k-skylake-processor-reviewed/10
http://www.legitreviews.com/intel-core-i7-6700k-skylake-processor-review_169935/8
http://www.techspot.com/review/1041-intel-core-i7-6700k-skylake/page8.html
http://www.guru3d.com/articles-pages/core-i7-6700k-processor-review-desktop-skylake,11.html

IIRC two are ~9%, one is ~3%, and another is ~15%. Obviously it depends what you're encoding and what settings you're using, which is why this is a crap-shoot until we get more reliable information on Ryzen. However, to suggest it might match Skylake/Kaby Lake IPC in some scenarios is not unreasonable.

 

[Silent_Scone]

Broadwell-E has more cores than Skylake so will obviously be faster but we're talking about IPC here. Broadwell-E uses the same core architecture as Broadwell (hence the name), which means it has the same IPC. There are dozens of comprehensive reviews on the internet of Skylake and most include a Handbrake benchmark showing Skylake is ~9% faster than Broadwell. I checked 3 of them whilst writing that post.

If you really want me to copy and paste the links from Google, here you go:

http://techreport.com/review/28751/intel-core-i7-6700k-skylake-processor-reviewed/10
http://www.legitreviews.com/intel-core-i7-6700k-skylake-processor-review_169935/8
http://www.trustedreviews.com/intel...eview-performance-analysis-and-verdict-page-5

We already know Skylake has higher single threaded performance compared to Broadwell. I thought you meant compared to BWE. Handbrake will use all available cores. In practicality saying Skylake is 9% faster than Broadwell-E simply isn’t true.

As far as I am concerned about as much of a conclusion as you have had with that of Ryzen compared to anything Intel at the moment.

You've stated that a chip that isn't released or have all the details is going to be slower compared to Skylake based on the fact that Ryzen as pointed out previous outpreformed the current Intel Broadwell by 9% in handbrake.

Can you conclude how you have managed to suggest that Ryzen will be slower than Skylake because of such info. And thus point where the conclusion of Skylake is greater than 9% faster than Broadwell.

I am not saying the statement of Skylake being 9% faster is correct. Just want to know how you got the opposite end of the viewpoint?

If you best emulate AMDs Broadwell-E results on Blender 27*, a stock 6900K with speedstep enabled is faster than the time AMD has shown. 31 seconds or less at 150 Samples. I’ve tried this over and over with the same version AMD used during their presentation.

Even if it transpires it matches Broadwell-E or is marginally quicker – the chances of it having a higher IPC than Skylake are slim…

By all means if you have done any testing to refute this, feel free to show it.

 

[eddyr]

Broadwell-E has more cores than Skylake so will obviously be faster but we're talking about IPC here. Broadwell-E uses the same core architecture as Broadwell (hence the name), which means it has the same IPC. There are dozens of comprehensive reviews on the internet of Skylake and most include a Handbrake benchmark showing Skylake is ~9% faster than Broadwell. I checked 4 of them whilst writing that post.

If you really want me to copy and paste the links from Google, here you go:

http://techreport.com/review/28751/intel-core-i7-6700k-skylake-processor-reviewed/10
http://www.legitreviews.com/intel-core-i7-6700k-skylake-processor-review_169935/8
http://www.techspot.com/review/1041-intel-core-i7-6700k-skylake/page8.html
http://www.guru3d.com/articles-pages/core-i7-6700k-processor-review-desktop-skylake,11.html

IIRC two are ~9%, one is ~3%, and another is ~15%. Obviously it depends what you're encoding and what settings you're using, which is why this is a crap-shoot until we get more reliable information on Ryzen. However, to suggest it might match Skylake/Kaby Lake IPC in some scenarios is not unreasonable.

WRT handbrake tests, those reviews are the mainstream platforms with broadwell using DDR3 v skylakes DDR4. Considering Broadwell E is DDR4, this could change the comparison between BW and SL.

Take the techreport review mem benchmarks. With one exception skylake outclasses in both cache and mem bandwidth.
http://techreport.com/review/28751/intel-core-i7-6700k-skylake-processor-reviewed/4

 

[DragonQ]

We already know Skylake has higher single threaded performance compared to Broadwell. I thought you meant compared to BWE. Handbrake will use all available cores. In practicality saying Skylake is 9% faster than Broadwell-E simply isn’t true.

You're right, I meant in IPC terms not absolute terms, as that's what you were talking about in the post I quoted. I forgot to say that in my original post.

WRT handbrake tests, those reviews are the mainstream platforms with broadwell using DDR3 v skylakes DDR4. Considering Broadwell E is DDR4, this could change the comparison between BW and SL.

Take the techreport review mem benchmarks. With one exception skylake outclasses in both cache and mem bandwidth.
http://techreport.com/review/28751/intel-core-i7-6700k-skylake-processor-reviewed/4

Yep, memory is another factor and it could well be a factor for Ryzen too. That's why I said it was a crap-shoot until we get more information but I stand by my statement that "to suggest Ryzen might match Skylake/Kaby Lake IPC in some scenarios is not unreasonable". IPC is not a fixed thing, it can vary wildly between applications. I've seen people argue that Skylake has 10% better IPC than Broadwell, which is just nonsense - 10% might be the "up to" figure but on average it's far lower (3-5% IIRC). Again though, it depends what benchmarks you include in the average and what RAM is being used in the test, so it's very tricky to nail these things down.

 

[Holst1981]

Will these chips work on x99 chipset?
Or will you have to use x99?

I'm happy with 5820k at the moment, but it's nice to know there might be something worth upgrading to if I need it.

Pointless to comment on ryzen performance until release.
It looks positive to me, but until we see retail prices and benchmarks there is no way to know.
Worth waiting to see how good it is.

 

[darket]

Will these chips work on x99 chipset?
Or will you have to use x99?

I'm happy with 5820k at the moment, but it's nice to know there might be something worth upgrading to if I need it.

Pointless to comment on ryzen performance until release.
It looks positive to me, but until we see retail prices and benchmarks there is no way to know.
Worth waiting to see how good it is.

No is the simple answer, BW-E is the last cpu family for x99 , SL-E/X is a new socket.

 

[Silent_Scone]

Will these chips work on x99 chipset?
Or will you have to use x99?

I'm happy with 5820k at the moment, but it's nice to know there might be something worth upgrading to if I need it.

No X299 is an entirely new platform for both KBL-X and SKL-E. These two will drop in on the same platform/motherboards

 

[DragonQ]

Whoops.

 

[Silent_Scone]

Jesus Christ, don't pretend you don't know what I'm saying.


Which is why I didn't say that.

lol, you're lost. You already replied to that post.

You're right, I meant in IPC terms not absolute terms, as that's what you were talking about in the post I quoted. I forgot to say that in my original post.

 

[Orcvader]

I think Skylake X would also be on 14nm+

But then that would make it Kabylake X ... Unless Intel actually tweaked something in the CPU core to make it slightly better than Skylake X, but then it would seem odd to limit it to just the quad core configuration.

 

[eddyr]

But then that would make it Kabylake X ... Unless Intel actually tweaked something in the CPU core to make it slightly better than Skylake X, but then it would seem odd to limit it to just the quad core configuration.

I think the KabylakeX 4c makes sense for the platform. Assuming its around 1151 i7 price, it enables a lower entry point into the platform. Get what would be higher end mainstream i7 buyers onto the enthusiast platform where they can provide upgrade path (that is set to become more viable with competition with AMD in the 6core+ market). This way they don't have to really alter the mainstream platform drastically and can keep it a cost effective option, ie no need to be upping connectivity and board complexity.

 

[Mauller]

People talking about skylake being 9% faster than broadwell in terms of IPC etc.

At the same clocks, kabylake is only 20% faster on average compared to Sandy bridge.

http://www.hardocp.com/article/2017/01/13/kaby_lake_7700k_vs_sandy_bridge_2600k_ipc_review

The major performance differences between each generation comes more from base clock increases rather than IPC when performance is averaged.