• 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 3000 Builders Thread

Page 77 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

JimKiler

Diamond Member
Oct 10, 2002
3,360
74
91
Anyone had good or bad luck with the Gigabyte x570 Aorus Elite? It seems to be a good candidate for use now with an 8core and later with a 12core or 16core that doesn't break the bank with features I don't want.
Gamers Nexus argues that both the Gigabyte Aorus Elite and Asus Tuf Gaming Plus are so good, unless you spend more than $300 there is nothing else worth buying. I personally got the Asus Tuf Gaming Plus (Wifi) for the bluetooth and Q LED's. I am sure the Gigabyte is solid as well. Personally i think all motherboards should have boot up/ Q LEDS. It should not be a premium feature.

go to the 21:02 mark on this video:

 
Feb 4, 2009
22,824
4,510
126
Hey guys sort of easy question. To set xmp memory is that in the bios section labeled OC/over clocking?
I changed my memory speed to 3200 in bios however the cas is at 24. I know I can do better than that. Crucial Ballistic micron e die.
Also is 50C(ish) good temp for moderate gaming?
Temp taken from Ryzen Master.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
18,796
6,127
136
Hey guys sort of easy question. To set xmp memory is that in the bios section labeled OC/over clocking?
I changed my memory speed to 3200 in bios however the cas is at 24. I know I can do better than that. Crucial Ballistic micron e die.
Also is 50C(ish) good temp for moderate gaming?
Temp taken from Ryzen Master.
All the board BIOS that I have, specifically say something like "XMP" and if you select that, some allow you to choose the memory speed.50ish temps at moderate load ? seems fine to me. In doubt of your cooling solution ? try prime95. It you don't go over 80c, then you have good cooling.
 
  • Like
Reactions: Drazick

DrMrLordX

Lifer
Apr 27, 2000
13,507
2,949
136
Is it everyone's understanding that if you run a 3900x in ECO-Mode that it's basically a 3900.
Haven't tried it yet, actually. Might be better to volt-starve the chip via offsets than to try ECO mode though. Hmm, might have to play with that later.
 
Feb 4, 2009
22,824
4,510
126
All the board BIOS that I have, specifically say something like "XMP" and if you select that, some allow you to choose the memory speed.50ish temps at moderate load ? seems fine to me. In doubt of your cooling solution ? try prime95. It you don't go over 80c, then you have good cooling.
Cool thanks man.
Been many hurdles to this build but I’m finally there, except windows but I suspect I’ll solve the activation issue with a long call to Microsoft.
*Tried the online account activation but it didn’t like so many hardware changes.
 

JimKiler

Diamond Member
Oct 10, 2002
3,360
74
91
Hey guys sort of easy question. To set xmp memory is that in the bios section labeled OC/over clocking?
I changed my memory speed to 3200 in bios however the cas is at 24. I know I can do better than that. Crucial Ballistic micron e die.
Also is 50C(ish) good temp for moderate gaming?
Temp taken from Ryzen Master.
My Asus board calls it D.O.C.P.. Tom's says you should manually set the XMP settings to shave a fraction of a second off your boot time.
 

daperl

Member
Feb 15, 2016
63
18
51
Haven't tried it yet, actually. Might be better to volt-starve the chip via offsets than to try ECO mode though. Hmm, might have to play with that later.
If things go as planned I should have a system built by early next week and I'm gonna give this a try. I definitely want to dial-in the right cool-'n'-quiet profile, but wouldn't it be nice if we could just press a button. Please give it a go if you have time; I'm interested in what you find out. I read somewhere that ECO-mode being available is motherboard dependent. Do you know anything about this? And if so, is there some BIOS switch that needs flipping?
 

Kenmitch

Diamond Member
Oct 10, 1999
7,533
689
126
Is it everyone's understanding that if you run a 3900x in ECO-Mode that it's basically a 3900.
I'd lean towards the 3900x being the faster of the two still. Locking the 3900x to 65w with ECO mode won't kill the boost speeds until higher core count loads. How far it drops with all cores loaded? I'd imagine higher than the 3900 still.
 

daperl

Member
Feb 15, 2016
63
18
51
I'd lean towards the 3900x being the faster of the two still. Locking the 3900x to 65w with ECO mode won't kill the boost speeds until higher core count loads. How far it drops with all cores loaded? I'd imagine higher than the 3900 still.
Makes sense. A better binned chip should be more efficient with the same power target.
 

DrMrLordX

Lifer
Apr 27, 2000
13,507
2,949
136
If things go as planned I should have a system built by early next week and I'm gonna give this a try. I definitely want to dial-in the right cool-'n'-quiet profile, but wouldn't it be nice if we could just press a button. Please give it a go if you have time; I'm interested in what you find out. I read somewhere that ECO-mode being available is motherboard dependent. Do you know anything about this? And if so, is there some BIOS switch that needs flipping?
I haven't tried ECO mode yet, so I'm not sure if the UEFI needs tweaking. I suspect that using LLC tweaks and offsets along with ECO mode will screw it up royally. But I could be wrong . . . anyway, the main problem with volt starvation is clock stretching. You have to test meticulously to make sure you aren't getting to that point.
 

DrMrLordX

Lifer
Apr 27, 2000
13,507
2,949
136
@DrMrLordX

I was just wondering how the 3900x performed with ECO-mode selected and all other CPU tweaks set to Auto or Normal or Default.
Fair enough. Let me try poking around a bit . . . one thing I can tell you is that setting Eco Mode with Ryzen Master makes the PC want to reboot itself. Hmmm! Odd.

edit: okay, did some quick testing. I found Eco Mode under the AMD Overclocking options in the UEFI of this x570 Aorus Master motherboard. First test is with LLC 0/LLC Off, no voltage offsets, and Eco Mode activated. Note that I left my memory OC in place which may or may not lead to excessive amounts of the package power being committed to the I/O die during MT testing. But really, running DDR4-2133 is ridiculous, and I don't feel like dialing in a completely different memOC for the test.

Single-threaded testing was conducted with SuperPi 1.5 mod XS. It's an SSE2 benchmark, and it is very good for isolating single-core boost behavior. Result: ~9.02 seconds, which is consistent with the performance that I get running the same config without Eco Mode. The weird thing is that I saw the highest temps I've ever seen in a single-core benchmark! The chip showed temps in excess of 55C. CPU-z reported really high voltages, too - 1.48v and (slightly) higher. I observed the same behavior in ST Cinebench R20.

Then I went for an MT test: CBR20 MT. Temps were stupidly low (44C-ish) and Ryzen Master reported the chip hitting a cap of 87W. Result was ~6550, which is quite a bit lower than the ~7200-7250 I usually get running this config without Eco Mode.

Pondering this for a bit, I decided to slap on negative voltage offset to see if I could force the chip to clock higher. This works without Eco Mode in MT scenarios, so would it work with Eco Mode?

Running SuperPi 1.5 mod XS again, the -.1v offset drove voltage down to 1.392v and produced a score of 9.045s which is very close to what I had without the offset. Again, this result is consistent with what I get using LLC Off + offset without Eco Mode. Temps during this run went down to the range of 48-49C. Testing CBR20 ST briefly, I saw temps jumping around between 48C and 52C, with a consistent voltage of 1.38v reported by CPU-z.

In CBR20, Eco Mode + LLC Off + -.1v offset produced temps maybe .5C lower than without the offset. The final score was 6589. Ryzen Master reported once more that the benchmark hit the cap of 87W.

edit edit: just to sanity-check my results, I disabled Eco Mode and went back to my "normal" settings for default operation: LLC Off, -.1v voltage offset. Then I reran SuperPi 1.5 mod XS. Peak voltage was 1.392v (as observed with Eco Mode) and the best result I could return with this config was 9.055s - ever so slightly slower than in Eco Mode. Temps hovered between 44-46C. The CPU actually produced higher temps in Eco Mode in this ST application, and ST performance went up as well (by a tiny bit). CineBench R20 ST showed similar behavior . . . using my "usual" config, the ST bench only caused temps to oscillate between 49-50C (1.38v reported by CPU-z).
 
Last edited:

Seba

Golden Member
Sep 17, 2000
1,393
40
91
@Seba

Sorry about the Win10 activation problems. Not surprised by that given the board changes.t should definitely help your MT scores.
Monday I successfully activated Windows by phone, with a human operator.

I explained my situation (a second major hardware change, soon after the first one), then I gave the ID of my Windows installation (9 groups of 7 figures each) and received a response code (long string of numbers) which (after I entered it) activated Windows.

Same procedure failed Friday, when I tried to activate by phone, but with their answering machine (it was outside of working hours). I entered the ID, but was denied a response code by the Microsoft bot.

I think you can get some more boost performance if you try both the LLC and voltage offset tricks you tried with your last chip. It should definitely help your MT scores.
Possibly, but at least for now I plan to keep everything at stock (except I activated the A-XMP profile for the memory).
 

daperl

Member
Feb 15, 2016
63
18
51
Whoa, nice work. Thanks for doing that. It seems ECO-mode has expected results in MT tests but maybe ST tests divert power resources to one core creating a hot spot while still hitting the eco power target.
 

amrnuke

Senior member
Apr 24, 2019
263
209
76
After getting set up, just wanted to see if things look grossly correct from a benchmarking standpoint. This is on a Zen2 3600, a reference Radeon 5700, Gigabyte DS3H B450 mobo, 2x4GB 3200MT/s CL16 RAM kit, Samsung 500GB SATA 2.5" SSD.

GeekBench 5 CPU - 1235 single / 7061 multi
GeekBench 5 Compute/GPU - 59878
CB R20 - 475 single, 3510 multi

All stock, latest drivers, fresh Windows 10 install.

Anything look off? I'm still waiting on the RMA on 16GB RAM but doubt it would affect these results much.

Further, what other benchmarking would you guys recommend to ensure things are where they should be?
 

Seba

Golden Member
Sep 17, 2000
1,393
40
91
CB R20 - 475 single, 3510 multi
I have/had similar results in CineBench R20 with two different Ryzen 5 3600 CPUs (also at stock). One with a B450 board, the other with a X570 board. Both with DDR4-3200 CL16.
 
Feb 4, 2009
22,824
4,510
126
Monday I successfully activated Windows by phone, with a human operator.

I explained my situation (a second major hardware change, soon after the first one), then I gave the ID of my Windows installation (9 groups of 7 figures each) and received a response code (long string of numbers) which (after I entered it) activated Windows.

Same procedure failed Friday, when I tried to activate by phone, but with their answering machine (it was outside of working hours). I entered the ID, but was denied a response code by the Microsoft bot.


Possibly, but at least for now I plan to keep everything at stock (except I activated the A-XMP profile for the memory).
I had a similar problem reactivating windows, except the opposite. Called into MS and some foreign guy basically tough s motherboard change means you need new copy of windows.
fortunately before upgrading to Windows 10 I took a picture of my windows 8 key and was able to find it in OneDrive
That key worked.
 

Kenmitch

Diamond Member
Oct 10, 1999
7,533
689
126
Fair enough. Let me try poking around a bit . . . one thing I can tell you is that setting Eco Mode with Ryzen Master makes the PC want to reboot itself. Hmmm! Odd.
The reboot flips the switches in the uEFI.

Did it effect your single core boost at all?

1571724832036.png

I was playing around with it on my 3700x last night and it's the 1st time I've seen my chip break the 4400MHz barrier....Actually I didn't see it, but it looks to have happened more than once.

I'll have to play around with it after work to see what it flipped in the uEFI. For some reason viewing the home page in Ryzen Master is showing max speed as 4600MHz. I'm leaning towards it activated PBO in the uEFI for some reason or another. Maybe it's possible to see what it changes and then tweak those settings to get not so ECO mode with higher single core boost speeds?
 

CHADBOGA

Golden Member
Mar 31, 2009
1,895
441
136
The reboot flips the switches in the uEFI.

Did it effect your single core boost at all?

View attachment 12249

I was playing around with it on my 3700x last night and it's the 1st time I've seen my chip break the 4400MHz barrier....Actually I didn't see it, but it looks to have happened more than once.

I'll have to play around with it after work to see what it flipped in the uEFI. For some reason viewing the home page in Ryzen Master is showing max speed as 4600MHz. I'm leaning towards it activated PBO in the uEFI for some reason or another. Maybe it's possible to see what it changes and then tweak those settings to get not so ECO mode with higher single core boost speeds?
Have you let @Kocicak know about this?
 
  • Haha
Reactions: Elfear

DrMrLordX

Lifer
Apr 27, 2000
13,507
2,949
136
The reboot flips the switches in the uEFI.
Yeah, I found the switch under AMD Overclocking. Funny how they hid that instead of putting it in the main Tweaker menu.

Did it effect your single core boost at all?
It seems to have. I got higher ST scores with Eco Mode activated than I did with it disabled. I also saw really high voltages (1.48v!) and temps in the mid-50s during ST benching which was bizarre. I didn't measure for actual peak clockspeed (oops).

I also want to add as an adendum, which I screwed up earlier by not mentioning: Eco Mode + negative voltage offset made SuperPi start to crash after a few runs. Restarting it made it work again. It was crashing from (apparently) . . . marginal voltage. Which meant, in ECO mode, the boost algorithm actually pushed itself into unstable clock territory with a -.1v offset. Fascinating.

I'm leaning towards it activated PBO in the uEFI for some reason or another.
I have an alternative hypothesis that (sadly) I can't necessarily support with evidence. But what if Eco Mode changes current restrictions on a per-core basis? One of the things it definitely does is change current limits, which you can see in MT mode. It may do this per core, so if I have n Watts committed to the operation of a single core but I have restricted current . . . that means jacking up voltage and (maybe) pushing the boost algorithm to seek higher clocks in tune with the higher voltage. Which might then drag current draw up - I think - and cause higher-than-expected clocks, power, and temps during ST operations. It's just a hypothesis, though. I couldn't/didn't measure current draw with and without Eco Mode in ST operations. Might have to do that later as time permits.

Have you let <redacted> know about this?
Zing! Seriously, I'd rather just let sleeping dogs lie.

Whoa, nice work. Thanks for doing that. It seems ECO-mode has expected results in MT tests but maybe ST tests divert power resources to one core creating a hot spot while still hitting the eco power target.
Glad to help. See above . . . it may be changing the way single cores reach their power targets.

edit: I know it's not absolutely related to the topic at hand, but this was a head-turner for me. I hit a new record in CBR20 running default mode:

CBR20default.png

This run used my "usual" settings. Just earlier today, I got a run as low as 7289 with the same settings. But the chip was feeling fast just now, so I reran it on a hunch and boom, it hit nearly 7500 in default mode. Matisse can really surprise you sometimes. Might be Win10 timing bug but I doubt it: SuperPi also pulled an abnormally quick time (9.02 vs the usual 9.05-9.06 or so). I'm kind of wondering if my Eco Mode performance results were generated similarly? Might need to get some corroboration from others to see if they get performance increases in Eco Mode (obviously Kenmitch is seeing the same effect; are others?).
 
Last edited:
  • Like
Reactions: lightmanek

Kenmitch

Diamond Member
Oct 10, 1999
7,533
689
126
I have an alternative hypothesis that (sadly) I can't necessarily support with evidence. But what if Eco Mode changes current restrictions on a per-core basis? One of the things it definitely does is change current limits, which you can see in MT mode. It may do this per core, so if I have n Watts committed to the operation of a single core but I have restricted current . . . that means jacking up voltage and (maybe) pushing the boost algorithm to seek higher clocks in tune with the higher voltage. Which might then drag current draw up - I think - and cause higher-than-expected clocks, power, and temps during ST operations. It's just a hypothesis, though. I couldn't/didn't measure current draw with and without Eco Mode in ST operations. Might have to do that later as time permits.
Had a short day at work today as it was really my day off anyways. I took a look in the uEFI to see if I could figure out what was up with the little extra boost I saw. It looks like ECO mode made some changes to the PBO settings. I have played around in there before without any success increasing my maximum boosts. PBO was disable in the uEFI before I used Ryzen Master to enable ECO mode. Some of the setting may have been leftovers form me? I know for a fact I didn't mess around with the power limits as I left them on auto when I tried. I guess that's why on the home page of Ryzen Master it shows my max core speed as 4600MHz.

1571774766554.png

Now I'm wondering if I can just play around with the limits and get semi-eco mode with higher single core boosts. I remember reading somewhere about diverting some of the power limit to the cores by lowering another voltage. Do you have any idea which one(s)? I was thinking it's SOC, but my MB has NB SOC and a Chipset SOC voltage option.
 

DrMrLordX

Lifer
Apr 27, 2000
13,507
2,949
136
Had a short day at work today as it was really my day off anyways. I took a look in the uEFI to see if I could figure out what was up with the little extra boost I saw. It looks like ECO mode made some changes to the PBO settings. I have played around in there before without any success increasing my maximum boosts. PBO was disable in the uEFI before I used Ryzen Master to enable ECO mode. Some of the setting may have been leftovers form me? I know for a fact I didn't mess around with the power limits as I left them on auto when I tried. I guess that's why on the home page of Ryzen Master it shows my max core speed as 4600MHz.

Now I'm wondering if I can just play around with the limits and get semi-eco mode with higher single core boosts. I remember reading somewhere about diverting some of the power limit to the cores by lowering another voltage. Do you have any idea which one(s)? I was thinking it's SOC, but my MB has NB SOC and a Chipset SOC voltage option.
Hmm! Interesting. NB SoC would probably be the one to lower. That's going to restrict memOC though.
 

Kenmitch

Diamond Member
Oct 10, 1999
7,533
689
126
Hmm! Interesting. NB SoC would probably be the one to lower. That's going to restrict memOC though.
That's the one I thought it was. I tried lowering it to 1.05v's last night and it didn't really do much for all core boosting in ECO mode. Maybe got me another 25-50MHz here and there.

This AM I was in the tweak the RX 5700 mode so was playing around with overclocks on it. I was still in ECO mode with the lower NB SoC voltage and it caused my system to reboot. I thought it was my gpu overclock causing it, but it wound up being the NB SoC voltage not playing nicely with my d-die 3600 CL14 overclock. Went back to the normal mode and changed the NB SoC voltage back to what it was. Still tweaking the 5700, but I've got it consistently boosting to 1850MHz so far....Broke the 9k barrier in Time Spy finally!
 

ASK THE COMMUNITY