Associate
They've done another update a week ago! Glad I checked.
https://www.amd.com/en/support/chipsets/amd-socket-am4/x470
Revision Number
2.01.15.2138
File Size
50.2 MB
Release Date
1/16/2020
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
The Radeon VII Owners Thread
- Thread starter Kaapstad
- Start date
https://www.amd.com/en/support/chipsets/amd-socket-am4/x470
Revision Number
2.01.15.2138
File Size
50.2 MB
Release Date
1/16/2020
Yes. I can keep it stable in 3dmark stress tests at around 105
@Jesaul That's what it should happen, even on water the JC limit should be preserved. However I'm facing a wall around 82C in the JC, anything above that would make timespy/firestrike crash. Strange, isn't? And it happens if the card is set to stock too if I lower pump/fans rpm to get higher temps, the test would crash around 82C+-.
Associate
Might be worth checking the HBM temp as that might be causing the issue and not junction temp. The internet has lit up in the past few days regarding issues with AMD drivers too. AdoredTV, and hardware unboxed have both touched on it. Core oc is still broken in the AMD drivers after 19.7.5.
Well, there are really so many factors that can influence. Like not enough voltage, power limits, too large frequency for current voltage, soc voltage that can influence... I've been experimenting with my card for 3 month
Well, there are really so many factors that can influence. Like not enough voltage, power limits, too large frequency for current voltage, soc voltage that can influence... I've been experimenting with my card for 3 month
Yes I do agree, but let's assume the card is stable, stock. I get same thing once I breach 82C+ if go above, crashes. I've also been testing with MPT for a long time, as wattman and others don't really function well. Think with me, if there's enough voltage OC should be stable, fine. Power limits were never a great deal and I could not see any much difference when changing for 0% to 25 or either 50%. SoC is something that I could not get any consensus, I bump around 5-10% to try to stabilize some overclock and got no result. But any way, in either scenarios, junction temperature should go up to 105C and then throttle, but my card does not.
Long time I don't use wattman to overclock, only MPT due the reason you mentioned. HBM temperature never go above 55C. I'm using HWiNFO which gives me same readings as GPUz.
Try this: at stock start increasing HBM freq. and test it for the full stress test. Try to reach 1130 MHZ(last stable for me). And check where it starts artifacting.
Maybe, maybe, your overclocking fails at memory.
Second: for SOC voltage, I used the evenmorepower reg file setting. So use the reg file and start experimenting with memory freqs. See if stability improves.
In short, memory overclocking can harm a lot.
@Jesaul I'll test it now as I'm at home, I don't think is hbm related, but I'll try as I did not think about this too.
Also is worth mentioning, I've just put 4 really thin washers in between the pcb and bracket and that limit of 82, 83C has increased to around 90C but is far away from the normal 105-110C. I've done the same tests (all 3dmark etc) and the temperature tolerance is higher now. What made me do this, is that I've found couple of screenshots from the time I was using it on air, clearly reaching 100C+ in the JT without any issue.
So, I might have some kind of bad contact?
This is how the paste is looking last time I tear it down to clear, looks well spread.
https://imgur.com/gallery/lzpwT5N
EDIT
Tested HBM and it goes all the up to 1200 without any issue, max temp 51C.
Also is worth mentioning, I've just put 4 really thin washers in between the pcb and bracket and that limit of 82, 83C has increased to around 90C but is far away from the normal 105-110C. I've done the same tests (all 3dmark etc) and the temperature tolerance is higher now. What made me do this, is that I've found couple of screenshots from the time I was using it on air, clearly reaching 100C+ in the JT without any issue.
So, I might have some kind of bad contact?
This is how the paste is looking last time I tear it down to clear, looks well spread.
https://imgur.com/gallery/lzpwT5N
EDIT
Tested HBM and it goes all the up to 1200 without any issue, max temp 51C.
Last edited:
Associate
After all the issue me and another forum member had with the alphacool blocks. When I installed my ek block I didn't hold back on the paste. I used the whole tube that came with the block. Manual spread, really quite a thick layer. I'm running 950mv 32c/60c junction. When I got 1000mv to hold for a few mins the other day to do 1900mhz that pushed the junction to 66c. (Alt +r in game, and just falls off).
@Chud_Jenny Hi, thanks for your input. I just did that couple of hours ago, gpu block off, cleaned well both bock and gpu die, and reapplied using way more than usual. Ideally I'd like to get same compound AMD uses, Hitachi HM03. But back to the issue, the problem is not how low I can get the temperature, this is totally fine. I run 1950/1100/1095mV and it barely goes around 65C. My issue is when I push the card, or when I decrease the speed of the pumps and fans making the JT increase to around 85C and then it fails, crashes, hangs, etc, it just does not go over and sustain let's say, 90, 95C, it crashes then. . Normally the cards should go around 105-110C in the JT before throttling. But mine is not, it crashes way before 100C. And it happens in any voltage, stock, mild or higher overclock. If I set stock and almost kill flow and fans, the card will go by 85C and will start to artifact.
Last edited:
Associate
A very helpful figure came out of Steve at Gamersnexus. For ever -10c in temp = 4% power saving. What i'm trying to say is are you running out of power? I run my card 1800@949mv 1140mhz HBM standard power. My default core mv is 1063 so quite low. 1900@1000mv is the sweet spot but uses a fair amount more power for 100mhz. The 7nm process is very good until you take it out of its efficiency curve. 100mhz costs around 30w extra in my setup. 1800mhz uses 160-170w due to the low leakage with temps 32/60 when my loop warms up.
What's your default voltage?
Associate
Ok, been away for a month or so with a new job and unable to mess about with OC'ing and gaming but back now, and updated to 20.1.3 this morning.
Not only is the new UI awful
but OC is still broke for me. Can load in a profile, say 2000mhz core @ 1112mvand 1150 HBM and still get the same old issue where in game it isn't applied. Not only that but the clocks are actually dropping off....@Jesaul are you saying then that you can get your OC to stick using the latest drivers by using the MorePowerTool? I haven't come across it before, or used Power play tables at all. A quick google seems to allude to the fact that it does the same thing?
Keen to OC my vii, but don't want to blow it up!!
Yes, it does the same thing.
In short, you dfu and clean install new drivers. After that apply evenmorepower registry. Restart. Load MorePowerTool. It will have values from registry tables inside. Modify them, save to file, write to SPT. And reset.
In the driver you only set power(like 99%) and HBM freqs.
Associate
Ok, will give it a go. Any good guides about for MPT?
Associate
Ok, so downloaded MPT and applied MorepowerVii_50. I can see in the tables where you would enter max clock freq, but not where you would enter the Mv? Only see options to increase the max voltage ranges, not applied voltage? Do you still have to do that in Wattman?
I did try changing the clocks and voltage in wattman but they didn't apply. Just hobbled the clock speed as before.
No, forget wattman once you start using MPT, even cause wattman does not work. It might if you only change gpu freq and voltage, the momento you touch mem clock it goes messy, and the other way around as well.
The tab Features you don't need to use it would be handy if you could rely on wattman. Same for overdrive limits. Fan depends on what you want and if you have you card on air, if is on water don't bother changing anything there.
You need to focus on the power and voltage and frequency tabs.
What MPT does, it basically manually input the value you want in the registry, alternatively you can use the premade .reg files we find around forums, but I'd stick with MPT.
So open it and load you bios (previously dumped from your card using gpuz), go to frequency tab and set there what you would set in wattman - gfx maximum and memory dpm2 is what I change here, that's it. gfx minimum, soc, all the other memory DPM I leave as it is. In the power and voltage tab set maxium voltage gfx, and power limit gpu if you believe this change anything (I don't think it works but....).. And that's it.
If you used any regfile before loading the bios is not necessary, you will be able to change the values in that way.
You dont need wattman once you are using MPT, I leave wattman on auto and don't even open it, as it might mess the registry entries. All the other options on power and voltage and frequency tab depends on what you want.
these are my settings, for example,
That's a good question, power. From the wall both cpu and gpu under load it draws around 350W, my psu is a rm805x, so there's enough headroom.
My stock voltage is 1080mv, if I check from wattman voltage curve, but, if I leave the card totally stock and run any load, the voltage displayed on the OSD is 1118V, weird.
One thing I mentioned before about this is, even if I set all stock, and turn both pump and fans down, I'll face the same problems once I start breaching 85C in the JT. So, default power consumption, and same issue, so I assume I'm not running out of power.
I've done a check on those sites where they calculate the whole consumption, I included every bit of the rig, even oc and voltages, and tops up around 450W.
What I really wanted to do, as a sane check, is to put the original cooler back. But that it too much, I'm not willing to do that. I rather use it with fans on 80% to do not get around 85C in the JT, or just run it at 1950/1095mV.
Associate
Ah, I think I get it. You just set the max values for core, mem and core mv and that's that.
Just seemed to work differently according to the MPT how to guide, extending the possible ranges in wattman. I'll give it a go and see though.
So set wattman back to default
You can, indeed, to extend the wattman ranges changing the values on the overdrive tab, but, if wattman (the newer ones) does not work there's no sense in doing anything in that tab
most drivers that function well with wattman are 19.7.5 (or 11) and below
Last edited:
Associate
You can, indeed, to extend the wattman ranges changing the values on the overdrive tab, but, if wattman (the newer ones) does not work there's no sense in doing anything in that tab
most drivers that function well with wattman are 19.7.5 (or 11) and below
Thank you. I'll give it a go today
