• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

AMD Ryzen 5000 Builders Thread

Page 42 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Det0x

Senior member
Sep 11, 2014
496
584
136
Here's a scenario that DOES happen... a lot, for professional builders.

Customer A says "I want a pure stable PC, there can be nothing overclocking related involved because -I- know that overclocking kills hardware and nothing you say will convince me otherwise... now advise me in those parameters."

Now, say out loud "AMD doesn't overclock their processors, they sneak you extra clocks because they got in trouble last generation and now sneak you free Mhz!"

Say that out loud and think about how the customer who said the first part is gonna react. Honestly.

Edit: Yanno what, nevermind. This is a waste of breath, there are no fellow pro builders here, or if so, they don't care what business they lose for "guessing" specs. I'm clearly out of place.
Really hope you dont use Nvidia graphic cards in any of the computers you build/recommend ? :innocent:
 

eek2121

Senior member
Aug 2, 2005
811
903
136
Really hope you dont use Nvidia graphic cards in any of the computers you build/recommend ? :innocent:
Yeah I don’t understand what all the fuss is about.

The 5950X can reach 5ghz at stock settings. This isn’t an overclock or the board maker cheating. The 5800x has similar headroom above the guaranteed boost clock.

The reason they do this is core quality. Depending on the silicon, not every core can hit the same max boost. Why doesn’t AMD lock them all down? Why would they? It is free performance.

Anyway, ShopBLT is now saying end of April for shipment.
 

Det0x

Senior member
Sep 11, 2014
496
584
136
Have played a bit with the latest CTR 2.04 hotfix that got released earlier tonight. (hybrid oc bugfix)

The next version of CTR, which will be launched in mid-April, will have automatic up and down clocking depending on the workload, so I'm pretty sure that CTR 2.5 will outperform PBO CO in most, if not everything, so i have started to do a few testruns :)

Following data based on Cinebench r20 with thread/score as the metric

PBO CO benchmode: (ambient ~ 20 degrees)

  • 1 thread = 662 points
  • 2 threads = 1303 points
  • 4 threads = 2444 points
  • 6 threads = 3706 points
  • 8 threads = 4887 points
  • 10 threads = 5974 points
  • 12 threads = 7022 points
  • 14 threads = 7906 points
  • 16 threads = 8645 points
  • 20 threads = 9583 points
  • ...seems like i didn't save 24 thread screenshot, but 105xx score
  • 32 threads = 12238 points

CTR 2.04 benchmode (ambient ~ 20 degrees)

  • 1 thread = 656 points
  • 2 threads = 1301 points
  • 4 threads = 2557 points
  • 6 threads = 3789 points
  • 8 threads = 5023 points
  • 10 threads = 6042 points
  • 12 threads = 7184 points
  • 14 threads = 8316 points
  • 16 threads = 8894 points
  • 20 threads = 9637 points
  • 24 threads = 10474 points
  • 28 threads = 11176 points
  • 32 threads = 12111 points

CTR 2.04 hotfix: (ambient ~ 24 degrees)

Did run all the threadcounts back to back without closing background apps etc. (can gain a few points on each score)
These CTR numbers are daily 24/7 settings intended for summer with high temperatures 😎
  • 1 thread = 652 points
  • 2 threads = 1295 points
  • 4 threads = 2525 points
  • 6 threads = 3752 points
  • 8 threads = 4979 points
  • 10 threads = 6016 points
  • 12 threads = 7171 points
  • 14 threads = 8287 points
  • 16 threads = 8831 points
  • 20 threads = 9539 points
  • 24 threads = 10217 points
  • 28 threads = 11117 points
  • 32 threads = 12032 points
LLC4 = upto 2% vdroop

PX high = from 1 to 2 threads @ 5050mhz
PX mid = from 3 to 4 threads @ 4950mhz
PX low = from 5 to 9 threads @ 4900mhz
P2 = from 10 to 20 threads @ 4800/4675mhz
P1 = from 21 to 32 threads @ 4700/4600mhz

1 thread @ 5050mhz -> 1475mv set = 1475mv get -> ~58 Tctl/Tdie temp
2 thread @ 5050mhz -> 1475mv set = 1475mv get -> ~63 Tctl/Tdie temp
4 thread @ 4950mhz -> 1450mv set = 1444mv get -> ~74 Tctl/Tdie temp
6 thread @ 4900mhz-> 1435mv set = 1431mv get -> ~72 Tctl/Tdie temp
8 thread @ 4900mhz -> 1435mv set = 1425mv get -> ~74 Tctl/Tdie temp
10 threads @ 4800/4675mhz -> 1350mv set = 1337mv get -> ~65 Tctl/Tdie temp
12 threads @ 4800/4675mhz -> 1350mv set = 1331mv get -> ~67 Tctl/Tdie temp
14 threads @ 4800/4675mhz -> 1350mv set = 1331mv get -> ~68 Tctl/Tdie temp
16 threads @ 4800/4675mhz -> 1350mv set = 1331mv get -> ~71 Tctl/Tdie temp
20 threads @ 4800/4675mhz -> 1350mv set = 1325mv get -> ~76 Tctl/Tdie temp
24 threads @ 4700/4600mhz -> 1275mv set = 1256mv get -> ~70 Tctl/Tdie temp
28 threads @ 4700/4600mhz -> 1275mv set = 1250mv get -> ~71 Tctl/Tdie temp
32 threads @ 4700/4600mhz -> 1275mv set = 1250mv get -> ~72 Tctl/Tdie temp

Settings:
32 threads.png
 

Attachments

Last edited:

B-Riz

Golden Member
Feb 15, 2011
1,295
364
136
Have played a bit with the latest CTR 2.04 hotfix that got released earlier tonight. (hybrid oc bugfix)

The next version of CTR, which will be launched in mid-April, will have automatic up and down clocking depending on the workload, so I'm pretty sure that CTR 2.5 will outperform PBO CO in most, if not everything, so i have started to do a few testruns :)

Following data based on Cinebench r20 with thread/score as the metric

PBO CO benchmode: (ambient ~ 20 degrees)

  • 1 thread = 662 points
  • 2 threads = 1303 points
  • 4 threads = 2444 points
  • 6 threads = 3706 points
  • 8 threads = 4887 points
  • 10 threads = 5974 points
  • 12 threads = 7022 points
  • 14 threads = 7906 points
  • 16 threads = 8645 points
  • 20 threads = 9583 points
  • ...seems like i didn't save 24 thread screenshot, but 105xx score
  • 32 threads = 12238 points

CTR 2.04 benchmode (ambient ~ 20 degrees)

  • 1 thread = 656 points
  • 2 threads = 1301 points
  • 4 threads = 2557 points
  • 6 threads = 3789 points
  • 8 threads = 5023 points
  • 10 threads = 6042 points
  • 12 threads = 7184 points
  • 14 threads = 8316 points
  • 16 threads = 8894 points
  • 20 threads = 9637 points
  • 24 threads = 10474 points
  • 28 threads = 11176 points
  • 32 threads = 12111 points

CTR 2.04 hotfix: (ambient ~ 24 degrees)

Did run all the threadcounts back to back without closing background apps etc. (can gain a few points on each score)
These CTR numbers are daily 24/7 settings intended for summer with high temperatures 😎
  • 1 thread = 652 points
  • 2 threads = 1295 points
  • 4 threads = 2525 points
  • 6 threads = 3752 points
  • 8 threads = 4979 points
  • 10 threads = 6016 points
  • 12 threads = 7171 points
  • 14 threads = 8287 points
  • 16 threads = 8831 points
  • 20 threads = 9539 points
  • 24 threads = 10217 points
  • 28 threads = 11117 points
  • 32 threads = 12032 points
LLC4 = upto 2% vdroop

PX high = from 1 to 2 threads @ 5050mhz
PX mid = from 3 to 4 threads @ 4950mhz
PX low = from 5 to 9 threads @ 4900mhz
P2 = from 10 to 20 threads @ 4800/4675mhz
P1 = from 21 to 32 threads @ 4700/4600mhz

1 thread @ 5050mhz -> 1475mv set = 1475mv get -> ~58 Tctl/Tdie temp
2 thread @ 5050mhz -> 1475mv set = 1475mv get -> ~63 Tctl/Tdie temp
4 thread @ 4950mhz -> 1450mv set = 1444mv get -> ~74 Tctl/Tdie temp
6 thread @ 4900mhz-> 1435mv set = 1431mv get -> ~72 Tctl/Tdie temp
8 thread @ 4900mhz -> 1435mv set = 1425mv get -> ~74 Tctl/Tdie temp
10 threads @ 4800/4675mhz -> 1350mv set = 1337mv get -> ~65 Tctl/Tdie temp
12 threads @ 4800/4675mhz -> 1350mv set = 1331mv get -> ~67 Tctl/Tdie temp
14 threads @ 4800/4675mhz -> 1350mv set = 1331mv get -> ~68 Tctl/Tdie temp
16 threads @ 4800/4675mhz -> 1350mv set = 1331mv get -> ~71 Tctl/Tdie temp
20 threads @ 4800/4675mhz -> 1350mv set = 1325mv get -> ~76 Tctl/Tdie temp
24 threads @ 4700/4600mhz -> 1275mv set = 1256mv get -> ~70 Tctl/Tdie temp
28 threads @ 4700/4600mhz -> 1275mv set = 1250mv get -> ~71 Tctl/Tdie temp
32 threads @ 4700/4600mhz -> 1275mv set = 1250mv get -> ~72 Tctl/Tdie temp

Settings:
View attachment 42461
So at full load, 32 threads, PBO CO is still a little better, but new CTR does better for lightly threaded things?
 

Udgnim

Diamond Member
Apr 16, 2008
3,658
81
91
had a CO curve of -13 on a 5800X and after trying to play GTA 5 SP, I'm at no CO curve offset at all now

game consistently crashed within less than a minute at the start of the game where Franklin & Lamar were repoing cars

reducing the offset only helped to extend the time before crash to some # of minutes.

went down to -4 and still got a crash, so at that point, I felt there's no point to trying to maintain a negative CO offset and haven't got a crash since going back to no CO offset
 
  • Like
Reactions: lightmanek

Det0x

Senior member
Sep 11, 2014
496
584
136
So at full load, 32 threads, PBO CO is still a little better, but new CTR does better for lightly threaded things?
I still think a properly optimized PBO CO setup is better for "light workloads" such as games as PBO lets the core boost above what they need to be stable with on a "medium workload". (CTR is tuning and maximizing itself for a medium workload like Cinebench)
In medium and heavy workload i could very well se a static/CTR OC be faster.

But its much harder to finetune and get stable what i would call a "optimized PBO CO setup", so for most people is probably better to just run CTR.

And next release of CTR should also be even better as it have automatic up and down clocking depending on workloads, the same way as PBO work.. Think CTR prettymuch always will be faster then, if 1usmus can get it to work properly :)

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Have been benching some games lately, these are my 5950x highscores so far.

SotTB 1080p lowest, 300 average CPU Game:
1617627235822.png

SotTB 1080p highest, 284 average CPU Game:
1617627307529.png

SotTB 1440p highest, 284 average CPU Game:
1617627353102.png

F1 2020 Australia, Cycle, 1080p high = 446 average fps
1617627497972.png

F1 2020 Australia, Cycle, 1440p high = 408 average fps
1617627553920.png

Farcry 5, 1080p ultra = 200 fps
1617627636533.png

Horizon Zero Dawn, 1080p performance, 50% res scale = 281 average CPU fps
1617627764236.png

Horizon Zero Dawn, 1080p performance, 100% res scale = 263 average CPU fps
1617627797629.png
 

Makaveli

Diamond Member
Feb 8, 2002
4,284
474
126
had a CO curve of -13 on a 5800X and after trying to play GTA 5 SP, I'm at no CO curve offset at all now

game consistently crashed within less than a minute at the start of the game where Franklin & Lamar were repoing cars

reducing the offset only helped to extend the time before crash to some # of minutes.

went down to -4 and still got a crash, so at that point, I felt there's no point to trying to maintain a negative CO offset and haven't got a crash since going back to no CO offset
i'm running -15 on my best cores and -5 on the rest and I was playing this last night for a couple hours on single player with no crashes.
 

B-Riz

Golden Member
Feb 15, 2011
1,295
364
136
I still think a properly optimized PBO CO setup is better for "light workloads" such as games as PBO lets the core boost above what they need to be stable with on a "medium workload". (CTR is tuning and maximizing itself for a medium workload like Cinebench)
In medium and heavy workload i could very well se a static/CTR OC be faster.

But its much harder to finetune and get stable what i would call a "optimized PBO CO setup", so for most people is probably better to just run CTR.

And next release of CTR should also be even better as it have automatic up and down clocking depending on workloads, the same way as PBO work.. Think CTR prettymuch always will be faster then, if 1usmus can get it to work properly
Interesting; I have yet to mess with CTR, still kinda wonder if leaving stuff "at stock" is easiest for Zen2 and Zen3, then adjust PPT for ones thermal requirements.
 

bigboxes

Lifer
Apr 6, 2002
32,414
11,208
146
I know I am taking my time, but this is taking longer than I thought. I hooked up all the front panel connections to the mother board, all the case fans to the Nexus+ fan hub, all the power cables and the 3.5" drives (with brackets) and the video card. Then I did all the wire management. The only thing left would be the data cables. Since it has an M.2 SSD, It's ready to fire up. *crosses fingers* Maybe tomorrow! :cool:



 

DrMrLordX

Lifer
Apr 27, 2000
17,031
6,001
136
Yeah it's a beefy cooler. Am I insane for wishing I had an HSF that's even bigger?

Yes, probably.

btw interesting that it blocks your m.2 screw. My m.2 sockets get blocked by other things like video cards. I don't think my nh-d15 blocked the m.2 socket on my x370 Taichi back in the day.
 
  • Like
Reactions: lightmanek

ASK THE COMMUNITY