Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

Page 172 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

lolfail9001

Golden Member
Sep 9, 2016
1,056
353
96
As for CB on 8 cores and 16 threads - do you mean with affinity set? Because it's like this:
Hm, looks like affinities do not work at all.
Of you mean without thread affinity... imagine 100% CPU usage for all cores for a blip, because it's done before you can sneeze.
More like without thread affinity but with 4,5,8 or 9 worker threads. 5 and 9 are particularly curious.
 

looncraz

Senior member
Sep 12, 2011
722
1,651
136
Hm, looks like affinities do not work at all.

More like without thread affinity but with 4,5,8 or 9 worker threads. 5 and 9 are particularly curious.

Only one I have for now: four threads, no affinity...

W10_HighPerformance_3GHz_Ryzen_CR15_4T_SMT_ON.jpg


I tried everything I could in the BIOS to try and disable SMT and msconfig just uses X logical cores when you use it to limit core count, so I am at the mercy of ASRock to get their BIOS in order.

EDIT:

Never mind, they just added support with BIOS 1.42 Beta... grabbing it and going to bed. Will disable SMT tomorrow and see what that does!
 

imported_jjj

Senior member
Feb 14, 2009
660
430
136
Never mind, they just added support with BIOS 1.42 Beta... grabbing it and going to bed. Will disable SMT tomorrow and see what that does!

Can you measure memory latency with only one module (so in single channel) with 1 CCX disabled and then with the other CCX disabled? Curious if there is a material difference.
 

lolfail9001

Golden Member
Sep 9, 2016
1,056
353
96
Can you measure memory latency with only one module (so in single channel) with 1 CCX disabled and then with the other CCX disabled? Curious if there is a material difference.
Fairly certain you cannot control which CCX is disabled in core disable options.

At least i only ever see the choice between 4+0 and 2+2, without 0+4 option.
 
  • Like
Reactions: looncraz

Trender

Junior Member
Mar 4, 2017
23
1
16
Guys read what is written. Loonc explained why it will be fixed for games.
Its the same old core 2 quad is "not a true quad core" crap again.
Stop the 2x ccx drama for games. Its nonsense. It just have to work within a 144fps limit. Not perfect but just not trashing. Hardly nano seconds we talk here. It will be okey.
Yeah it is but I wanted them to shit on Intel, now ppl keep saying bad things about Ryzen in games its so annoying. Anyways should I buy the Corsair 3200 MHz CL16 or Corsair 3000 MHz CL15 for my 1700?(sorry if I shouldn't ask here?)
 

Puffnstuff

Lifer
Mar 9, 2005
16,015
4,785
136
Yeah it is but I wanted them to shit on Intel, now ppl keep saying bad things about Ryzen in games its so annoying.
I was hoping that David would slay Goliath but that didn't happen. Lisa and the gang hyped up expectations in gaming and once real people got the new gear in their hands they were disappointed with the results. Yes there are many teething problems ritght and its up to AMD and their partners to work through them quickly to restore confidence. I'm hoping that everything gets sorted out and performance jumps across the board to force Intel to drop prices and enthusiasts and customers in general win no matter who's gear you want to build with.
 

piesquared

Golden Member
Oct 16, 2006
1,651
473
136
Kind of makes you wonder why AMD released it on March 2nd and not April 1st, because putting out a half baked product is something better done on April Fools.

Besides



Sorry, what.

I can answer that. It's highly likely that is the only way AMD could encourage motherboard makers to make some AMD boards in volume along with intel boards. Early adopters know there are growing pains, as was the case with X99 and many others in the past.
 

looncraz

Senior member
Sep 12, 2011
722
1,651
136
Can you measure memory latency with only one module (so in single channel) with 1 CCX disabled and then with the other CCX disabled? Curious if there is a material difference.

Latency didn't change depending on how many CCXes were enabled or how many channels I used (102.0ns every time with very little variation at DDR4-2133 CL15 2T, 92ns at DDR4-2667 CL15 1T).

There are some BIOS settings regarding how memory interleaving happens. I will continue to experiment with them - we have a lot more control than normal (we can even set L2 cache associativity!).
 

imported_jjj

Senior member
Feb 14, 2009
660
430
136
Latency didn't change depending on how many CCXes were enabled or how many channels I used (102.0ns every time with very little variation at DDR4-2133 CL15 2T, 92ns at DDR4-2667 CL15 1T).

There are some BIOS settings regarding how memory interleaving happens. I will continue to experiment with them - we have a lot more control than normal (we can even set L2 cache associativity!).

Have you checked if M.2 perf changes at all with the data fabric clocks?
 

krumme

Diamond Member
Oct 9, 2009
5,952
1,585
136
Yeah it is but I wanted them to shit on Intel, now ppl keep saying bad things about Ryzen in games its so annoying. Anyways should I buy the Corsair 3200 MHz CL16 or Corsair 3000 MHz CL15 for my 1700?(sorry if I shouldn't ask here?)

Dont know. I chose 3000c15 because it was in stock and cheapest. Kind of a principal thing. Had ordered 3200c16 anotherplace but i had to start ryzen 3 hrs before launch so not possible.

Today I would go with some ram that is proven for 3200.
 
Last edited:

looncraz

Senior member
Sep 12, 2011
722
1,651
136
Have you checked if M.2 perf changes at all with the data fabric clocks?

My NVMe performance is already better than on Intel - can't get much better (though I'm slightly behind in 4K-64Thrd in AS-SSD, I'm well ahead everywhere else).

I can't change bus clocks, I can only change memory multiplier... and I can go from 2133 to 2667 before I get a no-post situation. Will retest AS SSD with DDR4-2667, but I don't expect any changes.
 
  • Like
Reactions: Drazick

imported_jjj

Senior member
Feb 14, 2009
660
430
136
My NVMe performance is already better than on Intel - can't get much better (though I'm slightly behind in 4K-64Thrd in AS-SSD, I'm well ahead everywhere else).

I can't change bus clocks, I can only change memory multiplier... and I can go from 2133 to 2667 before I get a no-post situation. Will retest AS SSD with DDR4-2667, but I don't expect any changes.

Would be enough of a diff in clocks to see if there is an impact.
 

inf64

Diamond Member
Mar 11, 2011
3,686
3,957
136
ht4u.net finally posted their Ryzen review, it is packed with data. They seem impressed ;)
They included 1800X @ stock+ DDR4-3200 memory, along with 2400 and 2666Mhz results. Pretty massive review, took some time to read it through.
2666+MHZ memory really does make a difference in games with Ryzen, in some titles there was 15+% difference in 1080p (resolution they used for their tests).

https://www.ht4u.net/reviews/2017/amd_ryzen_7_1800x_im_test/index37.php