• 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."

Info New 5950x box for DC ** Its live now ! See the bottom for details. **

Page 7 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,128
9,258
136
Thanks for posting the comparison. It sounds like the 5950x will be a great cruncher.

If you already know about the this just disregard what I'm about to post, but depending what your goals are(max speed vs max efficiency) you could probably get those machines down to something like 300 watts(maybe 350) by setting eco mode in the bios, and then lowering the power limit of the 2080ti. It will lower the performance somewhat, but it should be a much lower percentage of performance lost than power efficiency gained. Eco mode should be in the bios somewhere like Overclocking --> Advanced CPU Options --> Precision Boost Overdrive and have settings for 65w, 95w, etc. The higher the wattage the better performance generally, but it uses more power of course. There's usually a PPT limit in a similar spot, that can be set manually instead of eco mode if that's preferable, and should perform similarly.

As far as changing the Nvidia card power limit you would just open a linux terminal and this is the command(the last number is what you're setting the power limit to in watts)

sudo nvidia-smi -i 0 -pl 200
Thanks for that. The Ryzen command line ? or bios only ? Its doing fine though. My EPYC boxes save a lot of power.

WOW, it dropped the ppd by 20%. I think I will leave it at 260.(stock) Losing 1 million ppd for 60 watts ? no way,
 
Last edited:

cellarnoise

Member
Mar 22, 2017
121
54
71
O.k. Since I can't seem to find any AMD 5950x cpu's to order since inception. I plan on now overclocking my Zen1 versions until AMD gives up the goods. I may even overclock them until they just barely fail every so often, hoping that they report "home" their failures :)

I hope for the Zen 3 "refresh" later this year that they are able to use the Zen 3 launch demand to calibrate their factory ask? I know that these things are often planned several years in advance, but maybe AMD has some flexibility built into their manufacturing contracts? Or else this is a lot like the last time AMD had some advantage or parity to the 800 pound Gorilla and ran into capacity issues to fill demand. Though I think that athlete's foot was the major bacteria / virus issue back then :)

Stay safe everyone.

This is my first focused attempt at torturing silicon because of lack of "its" later generational supply, but someone or thing should have to pay :()

This is could lead to machine paybacks?
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,128
9,258
136
O.k. Since I can't seem to find any AMD 5950x cpu's to order since inception. I plan on now overclocking my Zen1 versions until AMD gives up the goods. I may even overclock them until they just barely fail every so often, hoping that they report "home" their failures :)

I hope for the Zen 3 "refresh" later this year that they are able to use the Zen 3 launch demand to calibrate their factory ask? I know that these things are often planned several years in advance, but maybe AMD has some flexibility built into their manufacturing contracts? Or else this is a lot like the last time AMD had some advantage or parity to the 800 pound Gorilla and ran into capacity issues to fill demand. Though I think that athlete's foot was the major bacteria / virus issue back then :)

Stay safe everyone.

This is my first focused attempt at torturing silicon because of lack of "its" later generational supply, but someone or thing should have to pay :()

This is could lead to machine paybacks?
Since microcenter + B&H photo plus who knows who else is recently getting 5000 series CPUs and ebay scalpers are getting less every day, it won't be long before you can get these retail for MSRP. Not it looks like any day.
 

StefanR5R

Diamond Member
Dec 10, 2016
3,898
4,218
136
<off topic>
depending what your goals are(max speed vs max efficiency) you could probably get those machines down to something like 300 watts(maybe 350) by setting eco mode in the bios, and then lowering the power limit of the 2080ti. It will lower the performance somewhat, but it should be a much lower percentage of performance lost than power efficiency gained.
As far as changing the Nvidia card power limit you would just open a linux terminal and this is the command(the last number is what you're setting the power limit to in watts)

sudo nvidia-smi -i 0 -pl 200
WOW, it dropped the ppd by 20%. I think I will leave it at 260.(stock) Losing 1 million ppd for 60 watts ? no way,
Reducing the power limit of GPUs at Folding@home is a double-edged sword because of F@H's overemphasis of the quick return bonus. Still, at least with 1080Ti I remember to have gained a little power efficiency in terms of PPD/W when I lowered the power limit somewhat. Alas WU variability makes this hard to measure, therefore I don't have hard numbers.

(PS, the default board power limit of most 2080Ti cards is 250 W.)
</off topic>
 

Endgame124

Senior member
Feb 11, 2008
587
395
136
<off topic>



Reducing the power limit of GPUs at Folding@home is a double-edged sword because of F@H's overemphasis of the quick return bonus. Still, at least with 1080Ti I remember to have gained a little power efficiency in terms of PPD/W when I lowered the power limit somewhat. Alas WU variability makes this hard to measure, therefore I don't have hard numbers.

(PS, the default board power limit of most 2080Ti cards is 250 W.)
</off topic>
I’m running my 1080ti at a 160w power limit, and the points are all over the board. I jump between 1.4 mill and 2.6 mill ppd depending on the work unit. I ran it at full power during the last contest and honestly couldn’t tell you if it was any different - which WU you get is dramatically more important than power limit it seems.
 

Assimilator1

Elite Member
Nov 4, 1999
23,739
292
126
Is it possible to copy the WU (and whole install if necessary) so you can run it twice for benchmark comparisons?
 

Endgame124

Senior member
Feb 11, 2008
587
395
136
Is it possible to copy the WU (and whole install if necessary) so you can run it twice for benchmark comparisons?
Probably, but that PC is a pain to work with given my current office setup, so I don’t think I’ll be doing it. I would be interested in the numbers on something new, like a 3070, if anyone that has one is experimenting.
 

Icecold

Senior member
Nov 15, 2004
650
476
136
I've ran my 2080ti's at 200 watts at F@H and didn't notice a huge points difference, but like everybody else has mentioned the WU's are so variable it's hard to gauge. The side effect of the lower power usage is everything runs cooler, which is probably a larger benefit to me than the actual power savings. Running the processor on eco mode or with a manual PPT setting is a huge difference - at stock settings I was struggling to keep temps where I wanted them on a 3900x even with a pretty large Noctua heatsink, but once I set eco mode it ran cool even on AVX2/FMA3 work loads. I don't notice a huge difference in PPD on various projects with eco mode or a manual PPT limit set on a 3900x or 3950x as long as I don't try to go too low with it. My 3900x with a low power usage video card and 65 watts eco mode uses 130 watts from the wall, and still performs well, so I'm happy with that.
 

Assimilator1

Elite Member
Nov 4, 1999
23,739
292
126
I did some experimenting with my Ryzen 3600 and PPT, noting power draw increase along with clock speed and temp rises going from 53w PPT to the default 88w.

With the 65w PPT setting I decided to settle on, vs the default Auto/88w PPT, I gave up 3-4% on all core clock speed, which cut wall power consumption by 12-18% and lowered CPU temps by 13-16C for Rosetta and LHC!
 
  • Like
Reactions: Icecold

cellarnoise

Member
Mar 22, 2017
121
54
71
Well some might say ("they should be gentle")... That I like to complain...
I want new sillycone to play with but as long as Covid keeps virusing I and friends don't really have much use. Though my samsung virtual Google's are not getting used as much without better gpu. I tried three times to change Google's to goggels? And gave up...

I'm so fortunate to have more cpus and gpus to play with and not enough time. I'm thankful, but would like more.. bastard, no offense to bastards... Everybody needs someone to love.? Queen.?
 

cellarnoise

Member
Mar 22, 2017
121
54
71
I will add that if I lose fiber internet to my commode / er household that 4 individuals including my extended household would lose income. I might just order a replacement fiber router out of fear! Redundancy has real value quick when you go from 100 to zero in a blank...
 

Assimilator1

Elite Member
Nov 4, 1999
23,739
292
126
Well some might say ("they should be gentle")... That I like to complain...
I want new sillycone to play with but as long as Covid keeps virusing I and friends don't really have much use. Though my samsung virtual Google's are not getting used as much without better gpu. I tried three times to change Google's to goggels? And gave up...

I'm so fortunate to have more cpus and gpus to play with and not enough time. I'm thankful, but would like more.. bastard, no offense to bastards... Everybody needs someone to love.? Queen.?
Huh? You've totally lost me! :tearsofjoy:
PS Don't need full stops with ? ;)
 

Endgame124

Senior member
Feb 11, 2008
587
395
136
I have those 2 working. 2:36 for an open pandemics for 2990wx, 1:34 for 5950x, no comparison. But the 2990wx does twice as many tasks.
So it takes 3:08 for a 5950x to do as much work as the 2990wx does in 2:36. However, I imagine the 2990 uses more power, making the 5950 more efficient?
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,128
9,258
136
So it takes 3:08 for a 5950x to do as much work as the 2990wx does in 2:36. However, I imagine the 2990 uses more power, making the 5950 more efficient?
I think so. I don't have exact usage numbers now, but after I recover from this, I may take more tests. I know the 5950x blows away the 3950x.
 
  • Like
Reactions: Endgame124

cellarnoise

Member
Mar 22, 2017
121
54
71
I am watching also. From Zen 1 or Zen+ IPC, it seems that Zen 3 in fully loaded multi core is from 8% to up to 40% depending on the work load and instructions used?

Amazing generational improvements, and it seems to be invoking competition!

After so many 2010? to 2017 year timeframe (zen 1), there seems to be competition and price point back to the last AMD failure to advance (huge growth vs demand risk to be certain!!!!) time period (lack of supply and foreseeing the future 3 years +) A lot like now. These decisions are made 3 and 4 years out...

All of this is based on what you can realistically back with real wealth. AMD is building, but does not have much bank in the bank... It is growing though!
 

ASK THE COMMUNITY