*** AMD "Zen" thread (inc AM4/APU discussion) ***

[Silent_Scone]

Those physic scores don't correlate to the parts we do know about, let alone the ones that people want to know about.

 

[Apache14]

Ooooh

https://videocardz.com/65913/how-fast-is-ryzen

The fact the "per core" score does not change for the 4.0/3.2/3.3 speeds show you how dodgy that is. From 3.2Ghz to 4.0Ghz on the same CPU arch you should see some reasonable improvement "per core" regardless of anything else :/

 

[humbug]

Ooooh

https://videocardz.com/65913/how-fast-is-ryzen

I don't understand this, it makes no sense......

Why is the 6800K faster than the 6900K in MT?

Again its detracting from AMD to how the Intel scores make no sense at all, yes, again.... slower CPU's are again scoring higher than faster CPU's

Its yet another troll attempt.

 

[CAT-THE-FIFTH]

I don't understand this, it makes no sense......

Why is the 6800K faster than the 6900K in MT?

Again its detracting from AMD to how the Intel scores make no sense at all, yes, again.... slower CPU's are again scoring higher than faster CPU's

Its yet another troll attempt.

It could be a combination of poor score scaling and higher clockspeeds on the Core i7 6800K??

The scores for the Ryzen 6C and 8C CPUs at stock speeds compared to the BW-E ones seems very close.

Even the 4C/8T scores are close enough to the Core i7 4790K/Core i5 5775C.

 

[eddyr]

The fact the "per core" score does not change for the 4.0/3.2/3.3 speeds show you how dodgy that is. From 3.2Ghz to 4.0Ghz on the same CPU arch you should see some reasonable improvement "per core" regardless of anything else :/

There are plausible causes. The 8 core may be having to fetch data from the other CCX's L3 cache. The slower 4 core may benefit from all data being in its local L3 cache. While the 6 core from a better L3:core ratio (more cache per core) in comparison to the 8c, and perhaps from reduced inter-CCX traffic (when it has to fetch data).

 

[eddyr]

It could be a combination of poor score scaling and higher clockspeeds on the Core i7 6800K??

The scores for the Ryzen 6C and 8C CPUs at stock speeds compared to the BW-E ones seems very close.

Even the 4C/8T scores are close enough to the Core i7 4790K/Core i5 5775C.

Seems the 6800k result is from the writers personal system. The guy felt the need to mention in comments.

 

[humbug]

Seems the 6800k result is from the writers personal system. The guy felt the need to mention in comments.

Must be overclocked then.

 

[Cromulent]

If I can overclock an 1800X to 4.4Ghz it'll beat out my current CPU easily. I'm just waiting to hear some news about motherboards so I can start to plan which one I am going to get.

 

[Apache14]

There are plausible causes. The 8 core may be having to fetch data from the other CCX's L3 cache. The slower 4 core may benefit from all data being in its local L3 cache. While the 6 core from a better L3:core ratio (more cache per core) in comparison to the 8c, and perhaps from reduced inter-CCX traffic (when it has to fetch data).

That shouldn't be such a big hit to performance that the 3.2 and 4.0 ghz cores perform the same. I'm trying to say how they calculated per core performance is a little strange / wrong.

 

[SiDeards73]

Welp this has also just popped up

 

[Apache14]

Welp this has also just popped up

Taking with a pinch of salt, but looking very promising indeed. Really can't wait for the official reviews, hope the NDA drops this week for some good weekend reading.

 

[gavinh87]

If true, then that is matching kabylake clock for clock. Impressive.

 

[Silent_Scone]

 

[SiDeards73]

Watched that last night, end of the day they like everyone else kinda guessed at rough ballpark for Ryzen, he hinted it would be between Ivy and Haswell Performance towards the end, but a lot of unknowns in how the memory structure and stuff will actually pan out in real world for AMD. Was a fairly decent video, that Ryan Shrout dude makes my skin crawl listening to him though and the other dude who chimed in so often basically contributed almost zero to the piece other than just questions.

The only thing i took away from that is AMD has some tradeoffs with Zen and that it should get better with iterations, SMT might start off not as strong as Intels HT which is a given, but could potentially improve over time, no one is sure how AMD's architecture design will actually perform yet in real world as there are some unknowns and that he *thinks* its potentially Ivy to Haswell performance.

 

[Silent_Scone]

Yeah, however keep in mind in essence it's his job to predict performance, so he's probably fairly good at it (Kanter).

 

[humbug]

Welp this has also just popped up

Thats pretty good.

It puts the IPC 3% ahead of KabyLake and 54% ahead of Piledriver given the FX-8370 boosts to 4.3Ghz.

 

[humbug]

Thats pretty good.

It puts the IPC 3% ahead of KabyLake and 54% ahead of Piledriver given the FX-8370 boosts to 4.3Ghz.

Just run another math.... at 4.5Ghz, the same boost clock as that 7700K Zen would score 2721, which is just over 3% faster.... it all matches up.

I'm getting a bit giddy now....

 

[d_brennen]

Just run another math.... at 4.5Ghz, the same boost clock at that 7700K Zen would score 2721, which is just over 3% faster.... it all matches up.

I'm getting a bit giddy now....

I'd rein it in a bit, there's not long until we get proper, repeatable benches. All we have at the moment are click-bait snippets

 

[humbug]

I'd rein it in a bit, there's not long until we get proper, repeatable benches. All we have at the moment are click-bait snippets

Can't help it... but don't worry, i'm not pinning any hopes on this, i still have the salt here.

 

[SiDeards73]

PCGamer have put their neck on the line with this one http://www.pcgamer.com/new-amd-ryzen-details-and-pricing-leaks/

If you scroll through the benchmark slides it makes interesting reading thats for sure... so many leaks with similar results showing up now on price and now it seems potential performance. Its definitely getting interesting.