Official OcUK Ryzen review thread

[CAT-THE-FIFTH]

Eh? That chart shows 5 titles where there is SMT/HT causing degradation on the 6900K:-

F1 2016 -5.5%
GTA V -2.5%
Watch Dogs 2 -6.4%
Civ 6 -2.6%
Myenne jeux 3D -1.9%

So clearly not an issue unique to Ryzen (worse on Ryzen, granted, but not unique). Interesting how Intel have said or done nothing to fix it but AMD have recognised it and vowed to fix it.

Have you noticed how Kaby Lake does not show this on average?? Otherwise we would see performance dropping on average for a Core i5 7600K against a Core i7 7700K??

The problem is 12 games show this(I didn't bother posting some other charts),and even looking at the numbers you posted only two games on the Core i7 6900K show more than a 5% margin of error.

AMD is showing drops of over 5% on nearly a dozen games.

Hardware.fr showed that Ryzen was 10% quicker with no SMT enabled. Another site show around 6% to 7% and this is massive.

What you need to understand is that many reviewers didn't disable SMT,they tested with it on which dropped performance anywhere from 5% to 15% which is enough to make Ryzen look more like Ivy Bridge level for gaming.

Its better to switch SMT off for Ryzen,and if more reviews did this things would have looked better.

 

[CAT-THE-FIFTH]

from where was this chart???

Hardware.fr - a few others show the same. Hardware.fr hinted it was a scheduling issue due to the design of Ryzen being a pair of 4C core units and Windows not swapping threads properly.

OTH,AMD then said it was not a Windows scheduler issue but more a games optimisation issue which made less sense,and made it sound worse than it is since it would take longer to fix.

 

[CAT-THE-FIFTH]

Probably the best analysis of Ryzen on the internet for non-gaming workloads by The Stilt,a well known overclocker:

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/

At 3GHZ and below the efficiency is pretty amazing.

 

[CAT-THE-FIFTH]

Which confirms Ryzen is a server cpu scaled to the desktop.

Naples is going to be interesting when it lands.

It also explains the lack of SMT optimisations in games and less issues in non-gaming. I suspect AMD was more worried in getting that sorted out first for non-gaming.

My main gripe is that they should have made some excuse and said to test with no SMT in gaming as it would make the performance look stronger.

 

[CAT-THE-FIFTH]

See what I mean? With SMT on,the R7 1800X looks more like a Core i7 3770K. With it off,it seems more like a Core i7 4790K.

If AMD had communicated this with reviewers beforehand,more reviews would have tested it with SMT off. Haswell level gaming performance is much better sounding than Ivy Bridge level!!

 

[CAT-THE-FIFTH]

First picture of an AM4 mini-ITX motherboard leaked:

https://www.techpowerup.com/231205/biostar-shows-off-first-mini-itx-socket-am4-motherboard

It uses the X370 chipset.

 

[CAT-THE-FIFTH]

OP updated!!

 

[CAT-THE-FIFTH]

http://www.pcgameshardware.de/Ryzen-7-1800X-CPU-265804/Tests/Test-Review-1222033/

It seems the 4C versions which appear to use only one CCX might be fine.

Phew!!

 

[CAT-THE-FIFTH]

It looks like Hardware.fr has done some more testing with SMT off.

But here is the big kicker!!

They tested in performance mode and found not all the regressions were actually only down to SMT!

Edit!!

PS: Mods, Hardware.fr provide forum links for their review graphs,as they seem fine for direct linking of their images.

 

[CAT-THE-FIFTH]

Another day another potential issue which might have hampered performance at launch:

http://hwbot.org/newsflash/4335_ryz...bias_w88.110_not_allowed_on_select_benchmarks

 

[CAT-THE-FIFTH]

There's a YouTube tech channel ( I can't remember which) that tested at least 12 games, and while some definitely benefited from having SMT off 3 of them actually performed significantly worse with SMT off. It's not a singular picture.

Well there is also the power issues - look at the bottom part of the review page. It seems Windows power management might be part of the reason too.

 

[CAT-THE-FIFTH]

They're going to be waiting for a while until we have decent BIOS, drivers and Windows updated..

Better they do wait otherwise if they show even one negative result(due to some issue which is not their fault),everybody on social media would be accusing them of being paid Intel shills trying to discredit AMD.

At least if they wait a bit,it will be a fairer review for the AMD CPUs.