Discussion in 'Videocards - AMD Radeon Drivers Section' started by Hilbert Hagedoorn, Feb 3, 2020.
Acknowledged, there will be some fix for game profiles that may remedy this issue. If not, it will be fixed in near future.
Shuddaup with this strawman.
I think your lack of knowledge in building a PC is obviously showing in this post. If you understood that my position was that you have a potatoe PC and I inquire about it, including the psu, then it should be noted that if you have a low end psu it would need to be replaced for a 5700 xt.
Furthermore, thanks! I will now recommend your strawman, a 750w gold certified psu, just to trigger you, although I would recommend better in reality.
They shouldn't buy just "a cheap psu" as you imply. They should consider a quality PSU which may cost more. Which is why I recommend a higher quality. It's clear you understood the reasoning for my post. But throwing in a strawman to create alternative narrative to imply it's ok to buy a cheap psu because you don't like it when I say otherwise is childish. It is flagrantly bad advice and that needs to be pointed out. Which is were your own knowledge on the subject is lacking.
Have to test two 750w psu one with single 12v rail and other with multi rails.
I just need my system arrive.
No BLSOD using a single PCIE cable (and with extensions as well). When I built the PC back in September, one of the many suggestions to fix the BLSOD was using 2 PCIE cables. Besides trying that, I went further. Connected the GPU to a dedicated EVGA G2 850W. It kept BLSOD left, right and centre.
Again, the BLSOD is a driver issue. As always, my fear is the BLSOD returning for me in future drivers since people are still experiencing it with current drivers. AMD might fix it for them but then break it for me.
Yes, because tearing your pc case apart to change pcie cables when you clearly say you didn't have a problem is something we all do randomly. Nice try though.
It's a pity we cannot downvote posts, cause some of them would be buried under ten kilometres of dirt already. And if someone "thinks" that something so low will trigger me then good luck. Just makes me shake the head in disbelief and laugh. Knowitall-copypastas everywhere that cannot read full posts with understanding or simply ignore some parts of them...
Well.. let's hope AMD will nail it down this time (or in the next after driver release).
It's quite clear you cannot defend your logical fallacy and when refuted this is all you can say. But do be a favor watch the video below.
Apparently, I'm not the only one not having issues...
Maybe I didn't explain myself clearly. I DID have a problem back in September with 19.8.x and 19.9.x drivers and also in December with 19.12.2/3 drivers. I had one of the worst experience of building a PC. I thought the GPU or the PSU or the motherboard was defective. Spent a lot of time trying things out and countless of hard resets until I went to the AMD forums. Then more waste of time trying out older drivers one by one until I got to the non BLSOD but still buggy 19.7.5 drivers. My point is that the BLSOD is not a hardware issue. It's a driver issue, else I would still be experiencing the problem with the current drivers.
The rippling and power spikes can be a concern but the Vega GPU had worse figures so if the PSU can handle that it shouldn't be a concern but I suppose the high wattage and fluctuations are only one issue out of several so kinda hard to say for sure.
EDIT: The stock model at least, custom designs drawing more power and having a higher amperage might be a different result though I don't think they can exceed Vega even for the heavier Navi cards which a few even have stock power draw limits up to 220w or higher. (Which could then be increased by the power limit slider but that goes for Vega too.)
So 200 something total under normal spiking up to 400 watt compared to roughly 300 spiking to 600 for Vega.
Unsure on the amps, Navi or the 5700's seem to have bios values ranging from 12 to 14 for this value with the Pulse being a bit of a outlier at 15 for the regular bios (14 for the silent one.) but then a few of the bigger Navi cards use a value of 20 here.
(Voltage wise it all seems to be around 1.2v GPU core and curiously 1.2v SOC for stock with most if not all custom designs down to 1.050v with seemingly no drawbacks.)
That is why a good quality PSU is important... but like I've said -- 650W bronze is good enough for 5700XT, even for a Nitro+. Depending on the PSU --- with good quality cables and a single 12V rail it doesn't really matter if connecting both 6+8 on one or two cables (again - cable thickness / quality matters and how much power it can transfer without heating too much). If more rails - it does make a difference as it will split power load between lines.
Myself owning AX1200i, which is just a tad bit of an overkill normally, but not that much if you consider a Vega 64 Liquid cooled edition + TRX40 & 3960X, 2x M.2 Samsung 970 SSDs in RAID, 8 RAM Modules totalling in 128GB, additional SATA controller and 12 HDDs
Micro-Stutter fixed in new W10 build (build no 2004) for the dual and multi-monitor users!
Yeah there's potentially two issues from what I know that can be a factor but current PSU's should be well equipped to handle both.
First one is the very low power draw from the cards on idle which can cause the PSU to shut down though that would also be together with the motherboard and CPU having a low enough power draw which I think was mainly a issue on earlier power supplies from 2013 and before or lower quality models but those have other problems as well or can have compared to certified or known quality models.
(AMD also has a option in the bios for this on their motherboards increasing the idle power draw slightly to avoid this particular problem.)
Second one is the momentary spikes and shifting clock speeds and then the higher than expected transient power consumption which can be a issue but mainly on lower end modular power supplies and the common thing in current models seem to be just via circuitry and dividing it up a bit and for single rail mode well it should be able to handle this long as it's a 550 - 600w bit more perhaps and the efficiency while nice isn't as crucial you still have enough power to spare.
There's sites specializing in testing PSU's too and then you get 290X GPU's sitting at 380 - 400w consumption which if the PSU can take that then Navi shouldn't be an issue far as power consumption goes at least.
Still believe it mainly comes down to the software and something in the drivers but even with what AMD has fixed there's at least some titles or situations where a sort of shutdown can happen and the idle issue isn't entirely gone either even without browser or media playback hardware acceleration in effect.
(Well that and then there's also the AMD overlay and services which can be problematic but from my understanding mainly in other software not the desktop itself.)
EDIT: And OCP or over current protection and shutdown but once again the Vega GPU's and earlier could hit higher than Navi can so I wouldn't put the fault the PSU here.
GPU hardware wise though well that's a unknown that I can't say much about at all, clearly there's been a number of supposed issues but some of them I suspect are down to other errors.
The less than stellar "support" of PCI Express 4.0 on X400 series motherboards as one of these leading to AMD trying to disable it in newer bios code but that would require the user to be on those newer bios versions and for the manufacturer to still support said hardware for this.
(This has been kinda varied to say the least, some good some bad some terrible and sometimes just lengthy delays in support for older hardware.)
EDIT: Guess there's also the earlier PCI Express risers and modes for this and compatibility with 4.0 being a bit lacking until recently introducing another potential issue for those who have these in use for their graphics cards.
And whatever else could possibly apply, clean systems no immediate problems or compatibility issues and still hitting problems with Navi so yeah this problem is a pain to troubleshoot without just giving a top - 10 something suggestions on try it all and hope for the best which isn't exactly ideal.
EDIT: Let's see here...
AMD's various overlay services disabled/renamed and terminated.
Core driver only no control software or other extras.
Hardware acceleration in the browser or other software including media players and apps.
Power saving options like the PCI Express link state.
Fast boot or hibernation (Fast boot remains a Wattman issue still from what I know.)
Outright reverting to a earlier bios, clean install in a number of ways like DDU full removal.
Registry tweaking to disable certain settings like ULPS.
Ensuring features like various enhancements like enhanced sync, sharpening and such is off.
Display overrides with supplying additional power and/or disabling HDCP.
PCI Express 3.0 in the bios.
Using a single display.
Using a quality HDMI or DP cable.
Switching the PSU and/or PSU cables sometimes just using different connections on the PSU.
PCI Express riser or connector cable and compatibility here.
I'm sure I'm missing a few and it just keeps growing well hopefully the next driver or two will end this properly fixing the remaining problem unless it truly is something wonky at a hardware level.
(Which is going to look great I'm sure if the 'fix' is buying the improved RDNA with 2.0 and Navi20 well hopefully that won't be the outcome.)
EDIT: Oh yeah some suggest reformating and reinstalling the entire OS, I'd say that's a bit much but throw that in too I guess.
Or going for a NVIDIA card if we're adding that as well no surprise if nothing else works or not wanting to disable features or try how many different things that may nor may not do anything to improve stability or performance if lacking.
EDIT: And this is a text wall near unreadable mess already but eh I still like the hardware itself same as the initial impressions so hopefully the newer drivers will be what can finally resolve the software side of things leaving what I hope is a fairly small number of hardware defective GPU's that would need to be replaced though these should exhibit worse issues and also more consistently though I suppose that's not so easy for how complex this can all be.
Just an update. I gave 20.2.1 another chance last night. I was on 19.2.12 to get rid of the "no signal" black screen. I disabled ULPS after installing 20.2.1 and it seems to have fixed the black screen issue for now, will have to stay on this driver for further testing, fingers crossed.
I am on 5700XT
Disabling ULPS is one of the newest confirmed fixes, glad it helped you mate! Fingers crosses it is a perm-fix for ya!
What is interesting is that I can do 1.075-1.080mV and still get over 2000Mhz in game. Junction temps have been between 15c-20C lower by doing that and I can keep the fan speed around 50% and less. What's not clear is why there is such a discrepancy between 1.075mV to 1.20mV stability in games.
Just chiming in here from my experience with vega 64.
Vega 64 doesn't have to draw as many watts as even the stock settings. My card has a 260 watt limit stock, with my current undervolt it draws between 210-230 watts roughly under full load. I never noticed the card spiking in the voltage either. Performance wise the graphics clock sits around stock ratings and my memory clock is a steady 1100 mhz in game. I'm sure I can push the graphics clock up from where it is now, but I don't think it's worth the extra power draw for marginal gains.
As far as power supplies go, I have a platinum certified 750w evga power supply. I knew I would be overclocking all my components going into this build and wanted a power supply I knew I could rely on.
I disable hardware acceleration for all of my desktop applications and have no problems with black screens or anything else for that matter.
AMD does have to work on their drivers though, and I'm not looking for an upgrade right now but I would think twice before getting a new AMD card because of all these issues I've read about.
From an electrician's perspective on PSU:
1) Using a system that maxes out a PSU is not an ideal situation. The more you readily get closer to max output, you are increasing component stress, decreasing efficiency, which then leads to increased heat output. This will shorten the life of the PSU.
2) You can use a single cable for both GPU ports IF a) the rail can output the amount of power the card needs b) the cabling will not overheat from the current. If it's overly warm or too hot to touch, time to use 2 PSU connections/cables.
3) When I built my venerable i7 950 system, back in 2009?, I bought on sale a Cooler Master Silent Gold Pro 1200W ATX PSU for $189. (At the time I was experimenting with SLI, and the mobo could support QuadSLI) That system would hardly hit 450W, and thus my PSU was never stressed. It's also a single rail design @96A, so I never needed to worry about feeding the cards. I am still using that PSU, and the only thing I've had to do is change the fan twice. Odd thing is that it's a 135mm fan, and finding those is a PITA. So I drilled in new mount holes, and a 140mm fits just fine. On the first swap out, I sliced in a 3pin fan plug, making future fan replacements a snap. Currently using a Masscool Super Silent model, which solved the latest fan noise issue.
So, for $189 I've gotten 11, possibly 12 years of service. Totally worth it. (Also still using the Rev1 HAF-X case)
Agree on that. The PSU should always have like 20-30% room (this based on the PSU's efficiency -- bronze cert is a bare minimum, then we have a varying efficiency depending on the input voltage (110V vs 220-230V), then we have the total power counted from all the lines etc.).
Btw. hi fellow electrician/electronics fan ;-)
And actually the best efficiency is achieved around 50-55% of total load.
Some (usually lower quality) PSUs can even have a far worse efficiency when loaded under 30%.
Like with many PC components -- it is an art of balancing things out.
So let's count it shall we?
---GPU 5700XT Nitro+ in a sustained power-virus load (which will never happen and is a worst-case situation. Oh and this monster already clocks up to 2.0+GHz so if you considere a "normal" 5700XT + OC the result should be similar.
Power virus sustained situation - 310W -> https://www.techpowerup.com/review/sapphire-radeon-rx-5700-xt-nitro-special-edition/31.html
--CPU - let's take the high but not highest end because we want to cheap out on the PSU anyway! ;-) - 3900X
Again worst scenario which won't normally happen during gaming -- full load multi-core
210W -> https://www.techpowerup.com/review/amd-ryzen-9-3900x/18.html
(If you go with 3700X -- then shave 60-65W from it.)
--MoBo -- X570 and its power hungry chipset takes up to 15-20W
--Memory -- takes ca 5-10W
--SSD can eat up to 6-7W
--HDD -- let's call similar to SSD -- 7W
--Cooling -- let's count 5 fans in total, 3W each ---> 15W
--Some other peripherials -- 15W
--Some RGB fancy-shmancy stuff -- 10W
310+210+20+10+7+7+15+15+10 = 604W -=worst case=- with a power virus and full load on all cores.
And remember that some of the components don't eat up into our 12V power rail, so even if the PSU manufacturer counts the total power on all rails, it still should be enough.
Like I've said --- worst case with a power virus and a total multi-core workload running in parallel, all components active and under full load, SSD not idling, fans spinning at full power and RGB at full brightness. In a normal "gaming" workload it would be easily 50-70W less.
TL: DR --> It is safe to assume that with a 85% efficiency (bronze) a 650W PSU will be enough to drive a system described above with the best factory OC 5700XT in the most extreme situation and warranty stability.
Yes, it will reach it limits but even for this short time running under power virus and fullcore allin power load --- it should still warrant stability.
In a normal gaming situation there will be more headroom too. And the black screens during idling on desktop LoL.... have nothing to do with a 650W "bronze" PSU.
If everything goes well then new driver will be released next week.