Ryzen Owners Zen Garden

[Chicken Patty]

Holy crap! I think that's the highest SOC power I've ever seen from a 1CCD CPU I'm used to seeing those numbers on 5900X

auto VSOC must be pretty high then, that certainly explains higher idle power and a bit higher idle temps. Core powers are probably because you're doing stuff with your PC, but SOC should not change much unless under heavy memory/memory controller stress

Yeah I'm pretty sure ASUS defaults are exaggerated. I might look into setting it manually. I am streaming NHL Playoffs now and have some stuff open so understandable, it's not really idle. But yeah VSOC is probably very high. I'll look into that.

 

[freeagent]

There is always a little room for a tune

 

[Chicken Patty]

There is always a little room for a tune

View attachment 344679

indeed. Default ASUS was 1.243v, I set it manually to 1.2v and stress testing now while I type this. So far so good.

So I wanted to see what kind of voltages I got during a stress test. Anything here you guys spot I should try and set manually that may be too high? This was taken after the run finished, clocks during the CB run clocks were 4.95 to 5.05 GHz. Towards the tail end they started dropping a bit as temp got up to 85ºc but my room is really warm now, 26ºc and fans under 1000 RPM fixed. If I get the room a bit cooler and the fans going it'll stay under 80ºc.

 

[tabascosauz]

indeed. Default ASUS was 1.243v, I set it manually to 1.2v and stress testing now while I type this. So far so good.

So I wanted to see what kind of voltages I got during a stress test. Anything here you guys spot I should try and set manually that may be too high? This was taken after the run finished, clocks during the CB run clocks were 4.95 to 5.05 GHz. Towards the tail end they started dropping a bit as temp got up to 85ºc but my room is really warm now, 26ºc and fans under 1000 RPM fixed. If I get the room a bit cooler and the fans going it'll stay under 80ºc.

View attachment 344682

1.25V actually isn't that bad for an auto rule. But it seems you unfortunately drew the short straw on your board and SOC power is just high. I've seen this before with the Impact and Unify-X, the latter being almost 3W higher SOC at all times at identical settings. But not to this degree. I'm not sure what the root cause is. At 1.2V I would still be around the 10-11W mark.

Things look fine. A little warm for -30 but it's fine. 86 is getting a little close to Tjmax. But I think your board is pushing a bit more package power than mine.

 

[Chicken Patty]

1.25V actually isn't that bad for an auto rule. But it seems you unfortunately drew the short straw on your board and SOC power is just high. I've seen this before with the Impact and Unify-X, the latter being almost 3W higher SOC at all times at identical settings. But not to this degree. I'm not sure what the root cause is. At 1.2V I would still be around the 10-11W mark.

Things look fine. A little warm for -30 but it's fine. 86 is getting a little close to Tjmax. But I think your board is pushing a bit more package power than mine.

This is after the change. it is what it is.

 

[tabascosauz]

There is always a little room for a tune

View attachment 344679

JEDEC?? 0.8V SOC? 5700G? lol

 

[Chicken Patty]

So did some diggin' and manually set some voltages in the BIOS. Bit lower board draw on idle and temps much lower. 26ºc ambient and not entirely idle CPU is low 40's.

 

[tabascosauz]

So did some diggin' and manually set some voltages in the BIOS. Bit lower board draw on idle and temps much lower. 26ºc ambient and not entirely idle CPU is low 40's.

View attachment 344688

Nice. Looks normal now. What did you change?

 

[freeagent]

JEDEC?? 0.8V SOC? 5700G? lol

.850 for 1600 1:1

A far cry from the speeds you guys are seeing

 

[Chicken Patty]

Nice. Looks normal now. What did you change?

Much better. I changed:

SOC: DEFAULT 1.243 to now 1.050v
VDDG: DEFAULT 1.100 TO now 0.850v
VDDP: DEFAULT was auto, don't recall how much but now set to 0.800v

Now getting idle temp in the high 30's low 40's while typing this and streaming hockey, still ambient temp of 25-26ºc depends if A/C is on or off at the time of the temp check.

Looks like we may have sorted out my issue for good. I'll trial it some more and report back, but much better now.

 

[tabascosauz]

.850 for 1600 1:1

A far cry from the speeds you guys are seeing

Running ddr4 that slow has to be some kind of felony for Mr. OC-man. GET ON IT!!

 

[AusWolf]

Much better. I changed:

SOC: DEFAULT 1.243 to now 1.050v
VDDG: DEFAULT 1.100 TO now 0.850v
VDDP: DEFAULT was auto, don't recall how much but now set to 0.800v

Now getting idle temp in the high 30's low 40's while typing this and streaming hockey, still ambient temp of 25-26ºc depends if A/C is on or off at the time of the temp check.

Looks like we may have sorted out my issue for good. I'll trial it some more and report back, but much better now.

View attachment 344691

1.05 VSOC might be a bit too low for 6000 MHz RAM, but as long as it's stable, happy days. Just keep an ear out for USB drop-off sounds.

 

[Chicken Patty]

1.05 VSOC might be a bit too low for 6000 MHz RAM, but as long as it's stable, happy days. Just keep an ear out for USB drop-off sounds.

did a brief stress test, did some CB runs so far so good. But when I head off to bed in a few I'll leave it testing over night to be sure.

 

[AusWolf]

did a brief stress test, did some CB runs so far so good. But when I head off to bed in a few I'll leave it testing over night to be sure.

If there's an issue, then it's probably not during a CPU stress test when you'll notice it, but in a game, or some other lightly threaded work.

 

[Chicken Patty]

If there's an issue, then it's probably not during a CPU stress test when you'll notice it, but in a game, or some other lightly threaded work.

firing up some games now to test, will let you all know how it goes.

Well, just gamed for about 3 hours and so far so good. Guess I'll continue to keep an eye on it for the next few days and see.

 

[Athlonite]

This is after the change. it is what it is.

View attachment 344683

well those are still higher than mine, mine and I've not done any tuning the only differences here are our mobo's yours Crosshair Hero X670E mines a Strix X670E -E Gaming WIFI
and it's idling at ~28c

 

[Chicken Patty]

well those are still higher than mine, mine and I've not done any tuning the only differences here are our mobo's yours Crosshair Hero X670E mines a Strix X670E -E Gaming WIFI
and it's idling at ~28c

View attachment 344843

yeah this board seems to draw a lot of power regardless. I can try going lower, maybe at some other time. What are your ambient temps like?

 

[Klemc]

So, about my Asus, it's B650E-E for 7800x3d, what do you think of the numbers ?:

Ryzen Owners Zen Garden

so maybe bz's info about 2033FCLK for 3000UCLK was outdated (and whoever else was claiming in here that certain hidden Fabric-UCLK ratios exist and should be followed)? Or it's just AIDA being AIDA and ycruncher will reveal that 2033 does perform better. Who knows. Seems to be too big a...

www.techpowerup.com

 

[TheLostSwede]

Just updated to the F24c UEFI release after having been on the "old" F13 release for a long time now.
Tweaked things a bit while I was at it. Not too shabby.

 

[Keullo-e]

Updated my BIOS because why not.

I've always liked to keep them up to date anyway even if there's no actual benefits, no matter the platform.

 

[TheLostSwede]

Updated my BIOS because why not.

View attachment 345331

I've always liked to keep them up to date anyway even if there's no actual benefits, no matter the platform.

Well, there have been som weird UEFI security issues that have been patched, so it might not be for nothing this time around.
Specifically this:

LogoFAIL Vulnerability Affects Almost Every PC Running Windows and Linux

Binarly's research team has discovered a collection of security vulnerabilities known as "LogoFAIL", which affects image parsing components within the UEFI firmware of a wide array of devices. These vulnerabilities are especially concerning because they are embedded within the reference code...

www.techpowerup.com

So, about my Asus, it's B650E-E for 7800x3d, what do you think of the numbers ?:

Ryzen Owners Zen Garden

so maybe bz's info about 2033FCLK for 3000UCLK was outdated (and whoever else was claiming in here that certain hidden Fabric-UCLK ratios exist and should be followed)? Or it's just AIDA being AIDA and ycruncher will reveal that 2033 does perform better. Who knows. Seems to be too big a...

www.techpowerup.com

You have me beat in CPU-Z by some margin, so it's looking pretty good to me.
I would try and tweak the memory timings though, they're very relaxed, not that it matters to CPU-Z.

 

[Keullo-e]

Well, there have been som weird UEFI security issues that have been patched, so it might not be for nothing this time around.
Specifically this:

LogoFAIL Vulnerability Affects Almost Every PC Running Windows and Linux

Binarly's research team has discovered a collection of security vulnerabilities known as "LogoFAIL", which affects image parsing components within the UEFI firmware of a wide array of devices. These vulnerabilities are especially concerning because they are embedded within the reference code...

www.techpowerup.com

Interesting, I've missed that. Feels like there's more of these vulnerabilities these days than let's say like a decade ago. Or at least they're reported much often.

 

[AusWolf]

Interesting, I've missed that. Feels like there's more of these vulnerabilities these days than let's say like a decade ago. Or at least they're reported much often.

As tech gets more and more complex, I think it's only natural to assume that hacks and vulnerabilities follow along.

 

[Shou Miko]

AM5 Epyc Zen4 cpu's even with 3D v-cache is spotted:

AMD EPYC 4004 Zen4 "AM5" CPUs with 3D V-Cache have been leaked - VideoCardz.com

EPYC 4004 CPUs spotted on sale in the USA A seller from California has briefly listed AMD’s yet unreleased EPYC 4004 series of CPUs for the AM5 socket. AMD is about to launch AM5-based EPYC server processors, and there is now evidence that this is indeed happening. As we reported 3 days ago...

videocardz.com

 

[TheLostSwede]

Interesting, I've missed that. Feels like there's more of these vulnerabilities these days than let's say like a decade ago. Or at least they're reported much often.

I would say that there are also more people trying to (and succeeding in) find(ing) them.
More companies also have bug bounties, making it worthwhile to try and find these things.