Ryzen: Strictly technical

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.

set111

Junior Member
Oct 23, 2016
1
1
36
I would be interested in finding out if just limiting the program thread affinity in task manager to only use cores from one CCX makes any difference to the draw call / ST performance.
I assume this would also stop windows wondering the threads between both CCXs which should reduce CCX to CCX data transfers.
 
Last edited:
  • Like
Reactions: MajinCry

PPB

Golden Member
Jul 5, 2013
1,118
168
106
All of the designs had their multithreading (SMT/CMT) enabled during the testing since that is what the end-users would be having.

I've verified some of the individual results without the multithreading being enabled on all of them, and the differences fell within the standard deviation of the results (final results being 3RA).
IIRC NBody illustrated a slight improvement on Excavator from having the CMT disabled, however that's pretty much irrelevant since such configuration is not allowed at default.

Stilt, regarding "OC" mode. You said the voltage bias that Ryzen has at stock gets shut down and thus the actual voltage across every pstate gets higher, probably ending up in more power consumption across the board. Is that correct?

Can you tell for the people that is trying to decide if it's worth to OC Ryzen or not, how much does the OC mode impacts idle and low load scenarios regarding power consumption? Also, does the CPU engage in OC mode too if you undervolt? Have you tried undervolting or does it yield some advantages when running stock and considering turbo pstates?

Lastly. I have seen 1700 max turbo clock on all cores is 3.2ghz. Can you confirm what would be the max turbo clock on all cores for the 1700x? Thanks in advance and sorry for the bunch of question I'm asking. Deciding if it's worth not OCing at all on a 1700X instead of getting a 1700 and make it run at 1700x turbo clocks, and if the impact in power consumption across the board and not only P0 is worth it for my use case and form factor.
 
  • Like
Reactions: french toast

SPBHM

Diamond Member
Sep 12, 2012
5,056
409
126
did anyone test SMT off vs SMT on but with affinity set for the game process only for one thread per core?
kind of curios if the performance loss is the usual, just the game using the wrong threads at the wrong time, or something else with SMT causing some other penalty?!
 

piesquared

Golden Member
Oct 16, 2006
1,651
473
136
restore point AFAIK does not restore BCD. A bootable USB or DVD is needed. if don't boot, use usb or dvd, repair menu, command prompt and bcdedit to restore...

An clone of the harddrive is quick and easy, especially with SSD's.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
Stilt, regarding "OC" mode. You said the voltage bias that Ryzen has at stock gets shut down and thus the actual voltage across every pstate gets higher, probably ending up in more power consumption across the board. Is that correct?

Can you tell for the people that is trying to decide if it's worth to OC Ryzen or not, how much does the OC mode impacts idle and low load scenarios regarding power consumption? Also, does the CPU engage in OC mode too if you undervolt? Have you tried undervolting or does it yield some advantages when running stock and considering turbo pstates?

Lastly. I have seen 1700 max turbo clock on all cores is 3.2ghz. Can you confirm what would be the max turbo clock on all cores for the 1700x? Thanks in advance and sorry for the bunch of question I'm asking. Deciding if it's worth not OCing at all on a 1700X instead of getting a 1700 and make it run at 1700x turbo clocks, and if the impact in power consumption across the board and not only P0 is worth it for my use case and form factor.

Correct.
I haven't measured the actual delta in power consumption, between the normal mode and the "OC-Mode" (without manually undervolting), however I would expect the difference to be quite massive. I'd expect the difference to be anything between 25-40%.
The delta will vary between the specimens since the negative offset is dynamically calculated by the SMU based on the silicon characteristics. Also the load and the temperatures will affect it.
I'll measure that one too, so we'll know the facts (at least for one specimen).

3.5GHz is the maximum ACXFRC for 1700X.
 

piesquared

Golden Member
Oct 16, 2006
1,651
473
136
Not so much when you have 680GB of games and specialized tools installed :p

Of course. When i said quick and easy i meant relatively. :p Maybe it takes 15 minutes to clone but it's there when you need it, and being able to boot off of it if something goes wrong has saved me many of times!
 

iBoMbY

Member
Nov 23, 2016
175
103
86
Okay, I wouldn't recommend my bcdedit processor group settings, because it forces all threads of one process to one of the groups, so you would effectively always only use half the CPU with a single application. The setting is too strong.
 

flash-gordon

Member
May 3, 2014
123
34
101
That question could be a bit pre-mature, but out of all of these specs, on which could we expect they are going to focus on Zen II ?
1 - Memory controller is undoubtedly the first that comes to mind. But from AMA they said quad-channel is not what they are looking for, so I imagine just better latency overall.

2 - Higher frequencies of course...

3 - Infinity Fabric has it's own roadmap as described on "Anand Review pt1". So better interconnect is a sure thing.

4 - Probably better AVX and FMA cycle.

5 - I believe they will include more IO/PCI-E lanes into the SoC, keeping compatibility with current boards, but allowing the development of new revisions. What's good for their partners.

I see that AMD is much more prone to achieve +15% performance gain on high-end than Intel.
 

JaKno

Junior Member
Mar 3, 2017
1
0
1
I've used Win 7 about 95% of the time on Ryzen.
Only the actual performance evaluation was done with Win 10, as it wasn't up to me.
The performance differences are minor, but very constant regardless (in favor of Win 7).
Which motherboard did you use for your Win7 tests?
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
Curious about SMT's perf per W, would be nice if you could add that, guess 8C/8T would be easiest.

Regarding the subject:

yMXFJxi.png


8C/8T with SMT turned off achieved somewhat higher PPW than with SMT enabled.
121.426 RPS/W (54.804W) or 249.694% relative PPW.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
Stilt, regarding "OC" mode. You said the voltage bias that Ryzen has at stock gets shut down and thus the actual voltage across every pstate gets higher, probably ending up in more power consumption across the board. Is that correct?

Can you tell for the people that is trying to decide if it's worth to OC Ryzen or not, how much does the OC mode impacts idle and low load scenarios regarding power consumption? Also, does the CPU engage in OC mode too if you undervolt? Have you tried undervolting or does it yield some advantages when running stock and considering turbo pstates?

Lastly. I have seen 1700 max turbo clock on all cores is 3.2ghz. Can you confirm what would be the max turbo clock on all cores for the 1700x? Thanks in advance and sorry for the bunch of question I'm asking. Deciding if it's worth not OCing at all on a 1700X instead of getting a 1700 and make it run at 1700x turbo clocks, and if the impact in power consumption across the board and not only P0 is worth it for my use case and form factor.

wBivhdh.png


18.74% average difference between "Normal" and "OC Mode".
Exactly the reason why manual undervolting is mandatory when "OC Mode" is enabled.
The actual voltage delta is 121.5mV (the negative offset applied by the SMU in normal mode).
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
Would you be able to run that draw call benchmark in Win7, with a couple other tweaks?

Apparently there's been motherboard BIOS updates that have improved performance (faster DDR4 speeds), as well as disabling SMT also giving better performance. Would be good to see if there's a cumulative effect that boosts the draw call perf.

Glad you asked...

YTN3SVH.png


Same driver version, same everything else.
17.8% faster than Win 10 :rolleyes:
 

Agent-47

Senior member
Jan 17, 2017
290
249
76
have you tried disabling 2 cores from BIOS (if that is possible) and see if you can achieve a higher clock?

im might decide to wait for 6c12t if that is true.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
have you tried disabling 2 cores from BIOS (if that is possible) and see if you can achieve a higher clock?

im might decide to wait for 6c12t if that is true.

They can be disabled (3+3 config), however Zeppelin isn't limited by the thermals or power consumption. I expect the ceiling to be relatively same regardless of the configuration.
 

Rifter

Lifer
Oct 9, 1999
11,522
751
126
They can be disabled (3+3 config), however Zeppelin isn't limited by the thermals or power consumption. I expect the ceiling to be relatively same regardless of the configuration.

This will not be good for the 4-6 cores if true, hopefully they can get another stepping that clocks higher out for at least the 4 core parts.
 

imported_jjj

Senior member
Feb 14, 2009
660
430
136
Regarding the subject:

yMXFJxi.png


8C/8T with SMT turned off achieved somewhat higher PPW than with SMT enabled.
121.426 RPS/W (54.804W) or 249.694% relative PPW.

Thanks.
What i wanted to see was 8C/8T vs 8C/16T efficiency and from that perspecive it looks good as SMT not only increases perf but also leads to higher efficiency.
Not unexpected but you never know.
 

PPB

Golden Member
Jul 5, 2013
1,118
168
106
wBivhdh.png


18.74% average difference between "Normal" and "OC Mode".
Exactly the reason why manual undervolting is mandatory when "OC Mode" is enabled.
The actual voltage delta is 121.5mV (the negative offset applied by the SMU in normal mode).

But is it possible to set manual voltage for different pstates? P0 should require either stock volts or be overvolted to handle the overclock, and the rest will need that 121.25mv undervolt you are talking about. If this is possible, then there is some light into getting a 1700, overclocking to something sensible (3.6), then undervolt anything but P0 to "Normal Mode" levels.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
But is it possible to set manual voltage for different pstates? P0 should require either stock volts or be overvolted to handle the overclock, and the rest will need that 121.25mv undervolt you are talking about. If this is possible, then there is some light into getting a 1700, overclocking to something sensible (3.6), then undervolt anything but P0 to "Normal Mode" levels.

In "OC-Mode" there is no need to undervolt any other PState but the P0. This is because P0 PState is the only one which is used in "OC-Mode". Basically Ryzen is overclocked the same way all of the previous chips were, just remember not to increase the voltage immediately.

E.G.
"Normal-Mode" - P0 PState VID = 1.36250V, SMU voltage offset = ~ -120mV, effective voltage = 1.24250V.
"OC-Mode" - P0 PState VID = 1.36250V, SMU voltage offset = ±0mV, effective voltage = 1.36250V
 
  • Like
Reactions: Drazick
Status
Not open for further replies.