Overclock.net banner
15281 - 15293 of 15293 Posts

·
Registered
Joined
·
1,140 Posts
Yes this looks to be the way they are going on newer BIOS, to be honest with you the only reason I keep moving to a newer bios is in the hope that they will fix the fclk "issue"

Seeing that's never going to happen pretty much pointless using the newer bios seeing they seem primarily aimed at increasing stability rather than increasing performance.

On my lowly 5600x, at least on agesa 1.1.9.0 I can get all cores to boost to 5Ghz even with max cpu voltage restricted to 1.45v, if this artificial limit wasn't present I'm sure that the good cores could hold 5.1Ghz.

Moving to anything higher that restricts cpu boost overide frequency is pointless from a performance POV
I do have to add that with 1.2.0.5 my mouse issues are fixed with FCLK above 1900, so there is progress there for me. I can still produce USB disconnects though with y-cruncher test 17. I had these mouse issues on 1.2.0.3 even with the CPU set to AUTO. Seeing as with AUTO on 1.2.0.5 the voltage still goes up to 1.5v, I can't say it's a reduction in max voltage which has fixed the mouse issues. It has to be something else which has changed.

But the whole thing is a bit of a mess, if AMD wanted a 1.425v cap, why is it only introduced when EDC goes above 140? On Auto or PBO with EDC <=140 you can still see voltage go up to 1.5v. Effectively what this BIOS does right now is simply punish OCers with no real consistency. Which makes me think it's a bug. Maybe their new CPUs are going to have a lower voltage limit and as of now this beta BIOS doesn't have that properly implemented and it's conflicting with current CPUs?

Either you have a 1.425v cap or you don't. No point only bringing it in if the end user decides to go to 141+ EDC.
 

·
Overclock the World
Joined
·
3,372 Posts
Thankfully it also has resizable bar enabled as it's build around agesa 1.1.9.0.
Don't let yourself being marketing fooled
BAR mode & negative CO, exists since 1.1.0.0A
SMU 56.30 ~ maybe one of both even at 56.26
Marketing department will force you to update to 1.1.0.0C which had an ABL 1900 FCLK Lock
"Just so people can utilize BAR mode or negative CO"
Nonsense, lies and nonsense :)

Sadly i can't seem to find the post anymore from AMDRyzen Twitter
Only the 1.1.0.0D patch "with a new CO feature" is here, but maybe it remained into some more people's mind
But =<140 EDC -> 1.5VID, only >140 EDC -> 1.425VID. This feels… not right
Either you have a 1.425v cap or you don't. No point only bringing it in if the end user decides to go to 141+ EDC.
Yes it indeed is odd

But 1.4v limit existed since the launch day
Just it wasn't made obvious at all
I don't think 1425 is correct.
This is what hydra says, but this is already a voltage where you start to lose perf
(Likely not good copy of homework)
1413 or 1406mV was the peak
Eh let's say <1400 , as LLC droop exists

Please test how L3 scales
EDC limiter was well used on Matisse exactly to limit supplied voltage without ways going down on the freq strap
But on Vermeer, limiting EDC also limited L3 cache
Soo zero first stage throttlers, let the real voltage through, which you could limit then correctly with CO

Else , it was load dependent, if COs are actually correct,
Hence limited EDC will let the cpu take less voltage on high load
But then COs wont be correct for other applications

I'll check likely end of December, if Santa ASRock releases it till then
They need longer to verify it being "as rock stable" as possible 🤭
And if it fails verification, they won't release it at all without saying anything
Likely to not gain into marketing fire. Been doing half good half unliked things for AMD
Slowly balancing their AMD reputation I guess :)
* i wish them be slightly more rebel again, but that's me
 

·
Registered
Joined
·
211 Posts
2 PSUs, different brands, exactly the same? There's gotta be another variable here. Whether it's the motherboard, GPU or maybe my power source which I am going to check shortly. Would also suggest Corsair are cheating if they don't report the 12V voltage dip below 12v.
Curious if you have done any actual voltage testing with a meter (carefully of course).

Any crashes with the new PSU? I am no expert but I would think a slow dip with load might be ok up to a point, but quick fluctuations would cause the crashes. I had an old PSU that always read proper voltages but I was having random crash (black screen reboots). After testing everything I could, I broke down and got a new PSU. Voltages appeared almost the same under load but issues it stopped crashing.

You using a UPS on your PC? You can monitor your incoming voltages that way. I would think a good PSU would handle small voltage changes ok, I always use a UPS on my gaming rig, modem etc. I've lost power and finished a round in game before. ;)

A bad DP cable can do unspeakable things, like blowing off the GPU and/or mainboard :eek:

More inclined to think it's the GPU or something else... What about the mainboard 12V sensor? Does it read like iCUE?
I had a DP cable causing my RGB to stay lit when powering of PC (coincidently after replacing my PSU). PSU tech support asked me to disconnect DP cable, I did and lights went out. Ordered a certified DP cable and was shocked when it actually fixed the issue. Cheap DP cables in fact can do strange things and I imagine cause damage.
 

·
Registered
Joined
·
1,140 Posts
Curious if you have done any actual voltage testing with a meter (carefully of course).

Any crashes with the new PSU? I am no expert but I would think a slow dip with load might be ok up to a point, but quick fluctuations would cause the crashes. I had an old PSU that always read proper voltages but I was having random crash (black screen reboots). After testing everything I could, I broke down and got a new PSU. Voltages appeared almost the same under load but issues it stopped crashing.

You using a UPS on your PC? You can monitor your incoming voltages that way. I would think a good PSU would handle small voltage changes ok, I always use a UPS on my gaming rig, modem etc. I've lost power and finished a round in game before. ;)


I had a DP cable causing my RGB to stay lit when powering of PC (coincidently after replacing my PSU). PSU tech support asked me to disconnect DP cable, I did and lights went out. Ordered a certified DP cable and was shocked when it actually fixed the issue. Cheap DP cables in fact can do strange things and I imagine cause damage.
Nah, just relied on software readings. One of the things I was interested about with this Corsair PSU is the hookup to my commander pro so that it shows voltages and other info in iCUE. But they just seem to read direct from the PSU. The issue here is the voltage being read at the GPU input end.

No crashes or issues. My crashes seem to have been fixed by lowering my core clock. Even on the Thor PSU this seemed to fix the kernel-power 41 reboots. But given I wasn't 100% convinced lowering core clock had simply fixed it all I still wanted another PSU. This one appears to dip even more than the Thor!

But no crashes. I left a SP game on for 5 hours in a row today on a busy screen just to keep the GPU holding a high boost/utilisation. No crash when I came back. Did some testing in 3DMark yesterday as well, no crashes.

As this 2080Ti is an RMA I think I'll just chalk it up to being a poor performer and where I have OCed it to is as good as it gets. Once I get a 3080 I'll be able to more seriously look at voltage.
 

·
Overclock the World
Joined
·
3,372 Posts
Soo Mindfactory (DE) continues to refuse working with ASUS
And @Alphacool refuses working with ASRock for their blocks
(Which at this point is just refusing sells by personal moral decision)

This brand "hate" makes me unhappy and unfortunately restrictive
Having to go out of my way to get ASUS boards , just so i can bughunt new AGESAs
And having to go out of my way, to deliver blocks from the other side of the world, just to cool a liked and actually great product

Resolve for both is sticking to MSI for leaked bioses and going DIY instead of sticking to alphacool
But this shouldn't be the resolve 😓
 

·
Old crazy guy
Joined
·
3,853 Posts
Any crashes with the new PSU? I am no expert but I would think a slow dip with load might be ok up to a point, but quick fluctuations would cause the crashes. I had an old PSU that always read proper voltages but I was having random crash (black screen reboots). After testing everything I could, I broke down and got a new PSU. Voltages appeared almost the same under load but issues it stopped crashing.
The major issue is usually transients; what you see as a power usage is always an average.

When the voltage readings are fine but the PSU fails under load is usually due to the OCP protection, the Over Current Protection.
For a brief moment there was a spike and the PSU couldn't hold it.

When the voltage readings are low on top you can have the UVP, the Under Voltage Protection, kicking in as well.
During the spike the voltage drops below 11.4V and the PSU will trip.

You need an oscilloscope to see that, a DMM is not enough.
The 2000 and 3000 series nVidia GPUs they suck an impressive amount of power to minimize micro stuttering when boosting.

Igor's Lab did the effort measure the transients with the 3090:

Colorfulness Purple Product Rectangle Violet


It's the reason why lately all the PSUs have extremely generous OCP tripping points.
Which is bad when the PSU is cheap as it ends up in flames like the Gigabyte unit tested by GN.

If you want to OC a big CPU and an nVidia GPU in the 3080/3090 class it's highly recommended to have at least 1000W of nominal wattage.
 

·
Iconoclast
Joined
·
31,663 Posts
if AMD wanted a 1.425v cap, why is it only introduced when EDC goes above 140?
Because voltage in isolation is mostly irrelevant. Capping EDC will allow higher peak voltages with the same reliability/longevity.

It does punish OCers, but may preserve AMD's bottom line by reducing the number of returns.
 

·
Registered
Joined
·
1,140 Posts
Because voltage in isolation is mostly irrelevant. Capping EDC will allow higher peak voltages with the same reliability/longevity.

It does punish OCers, but may preserve AMD's bottom line by reducing the number of returns.
140 is incredibly low though, I was able to get good scaling in performance right up to 220. If high voltage is basically only used for micro-boosts then all AMD are doing is hurting potential ST performance just because you happen to have EDC above 140.

Doesn't really make any sense. There must be a more intelligent way to handle multicore voltage limits that aren't as clumsy as an outright hardcap depending on the EDC value.

As much as I want to say I'll trust the engineers, if midcycle you suddenly decide to take performance away from people who've been running settings for years that is NOT going to go down well. A completely new product introducing a new voltage limit is different from retroactively changing things on prior released hardware.
 

·
Registered
Joined
·
4,247 Posts
The major issue is usually transients; what you see as a power usage is always an average.

When the voltage readings are fine but the PSU fails under load is usually due to the OCP protection, the Over Current Protection.
For a brief moment there was a spike and the PSU couldn't hold it.

When the voltage readings are low on top you can have the UVP, the Under Voltage Protection, kicking in as well.
During the spike the voltage drops below 11.4V and the PSU will trip.

You need an oscilloscope to see that, a DMM is not enough.
The 2000 and 3000 series nVidia GPUs they suck an impressive amount of power to minimize micro stuttering when boosting.

Igor's Lab did the effort measure the transients with the 3090:

View attachment 2536271

It's the reason why lately all the PSUs have extremely generous OCP tripping points.
Which is bad when the PSU is cheap as it ends up in flames like the Gigabyte unit tested by GN.

If you want to OC a big CPU and an nVidia GPU in the 3080/3090 class it's highly recommended to have at least 1000W of nominal wattage.
Yeah well. My Focus + Gold 750w could not handle a 2080 Ti + 9900KS with a hard OC on both. Transients were way out of spec and it randomly tripped OCP in games / benchmarks.

Swapped it for a Prime Ultra Gold 1000w and that had no issues. It also runs my current 5900X + 3080 + full custom loop + 14 RGB fans and such just fine. No issues. Barely even has the fan spinning in Hybrid mode. And no OCP trips. I do wonder what would happen if I shunt the 3080 lol.. currently it has a 370w limit so..

Btw, my board, the ASUS B550-XE, will not POST at all with any bclk that isn't 100.00.. it just hangs on either "07" or a random different POST code.. weird..
 

·
Registered
Joined
·
1,140 Posts
Yeah well. My Focus + Gold 750w could not handle a 2080 Ti + 9900KS with a hard OC on both. Transients were way out of spec and it randomly tripped OCP in games / benchmarks.

Swapped it for a Prime Ultra Gold 1000w and that had no issues. It also runs my current 5900X + 3080 + full custom loop + 14 RGB fans and such just fine. No issues. Barely even has the fan spinning in Hybrid mode. And no OCP trips. I do wonder what would happen if I shunt the 3080 lol.. currently it has a 370w limit so..

Btw, my board, the ASUS B550-XE, will not POST at all with any bclk that isn't 100.00.. it just hangs on either "07" or a random different POST code.. weird..
If you can play New World with a heavy OC you're all good. Yes, New World managed to trip OCP for me on a 850w a couple of times. That reminds me, should try it now on 1000w.

Stopped playing it because it was crap but an OCP test will be a good idea!

As of now I was playing Outer Wilds and that managed to pull a full 380w through my 2080Ti just to run at 144FPS :ROFLMAO: PSU in was like 550w and this is WITHOUT PBO.
 

·
Registered
Joined
·
242 Posts
Because voltage in isolation is mostly irrelevant. Capping EDC will allow higher peak voltages with the same reliability/longevity.

It does punish OCers, but may preserve AMD's bottom line by reducing the number of returns.
But high EDC is beneficial for heavy multi core loads then voltage is far lower than 1.4, not light single core loads with high voltage. (Except for the L3 mega high edc behavior)

Even on 1203 (and before) it’s a stupid balancing there I get better single core performance by lowering limits from stock and better multi core performance by raising them.

Edit:
Well there are the strange all core light thermal load, high voltage high boost loads like mem testing - but are they a problem?

EDC 168
 
15281 - 15293 of 15293 Posts
Top