Cards run a voltage/frequency curve. Minimum voltage on the "old" 6900XT is 800mV @ 2600MHz (approximately). The "voltage" slider only adjusts the "curve", does not influence the max or min voltages.
I believe this issue running in All AMD GPU from previous rDNA 1 ( 5600,5700/xt) to new rDNA 2 ( 6700,6800,6900/xt ) this cause stuttering in games black screen low performance you see people reporting for 2 years it is a hardwares chip issue not driver . I believe Amd try real hard to mask it but couldn't fully fixed it . Maybe until new technology for next gen .
If that is a hardware level issue many users will be upset. Get these issues when the Card isn't utilized in full when certain games runs into a CPU limit. It clocks all over the place low as 40MHz going into sleep mode for a moment.
@@KleinMeme driver can't fix it most people will not notice as we see in the video it happens extremely fast in every frame . I had the latest Intel CPU still happening
I observed back in November when i was trying to get top Fire Strike score hat gt1 would always get higher fps in window mode. Also setting min freq to 100MHz less than ma gave me some more points.
Another set of voodoo tricks to learn, I guess. It'd be interesting to check how the score changes with regards to v/f and if moving sliders around really affect anything (it looks like the GPU overclocks all by itself with minimal input from the user)
Why there are those idle low voltages at the beginning with "stock" configs? Do they affect real world performances? There is any real improvements in keeping the min clock higher? Like minimum % FPS or 0.1%fps drop etc..?
9:30 I'm not sure why you are surprised the driver software already tells you that. Essentially it has an already Written voltage/Frequency curve, which you can lower at certain stages but you can't shoot over. This also explains why it lowers voltage when it reduces clocks (for what ever reason rdna2 at stock jumps up and down in clocks) and why it levels put when you increase minimum clock.
15 minutes of content + 2 minutes of outro, is that a record? I hardly had time to dunk my Penguin! LOL!!! Super interesting indeed. What is the betting that our man succumbs to undoing the mods for comparison content sooner rather than later?
Just wondering about something here: I have an Asus TUF OC RX 6800 (it's what i could get hold of for a semi-reasonable price) and I have some observations: 1) It has really limited oc headroom: have problems getting above 2400, and the voltage is perma pegged at 920 mV for some reason. 2) I replaced my PSU with a new one and used PCIE power cables with in-line caps 3) Coil whine reduced significantly. It was also easier to push the clock frequency. 4) I seem to get better performance with 100 Mhz delta between max and min clock frequency in certain "frequency holes" or combinations. Is the degree and frequency of voltage regulation that you see here combined with poor input filtering likely to create the problem in 4) ?
Random question. I'm a.newbie in the cpu world. Trying to overclock some c die gskill trident rgb 3600mhz cl18. Just read c die sucks. Anyway to tighten it up?
Noticed this weird behavior myself. Not really a voltage setting more an offset for a predetermined voltage to frequency curve that reacts to what clocks you set.
@@cosmic_gate476 when you tweak it using MMP it will force core clock to go into safe mode anytime you exceed 2150mhz even by 1mhz. It's a hard limit in the vbios and as far as i know only XTXH special die has vbios that is unlocked and you can't flash that since it doesn't work and there is no vbios editor yet because the vbios is signed/encrypted
If the "max frequency" slider increase the voltage, what would happen if you decrease the max voltage and increase the max frequency in order to have a higher frequency with a lower/same voltage ?
The excat thing one would expect to happen. If you reduce the voltage, you'll get the same frequency at a lower voltage. The voltage slider acts as an offset, so 1200mV only means 1200mV at the stock max boost. If you reduce the max boost, voltage will scale down with it as it's set to 1200mV at 2600MHz for example, it will do 1100mV if you run it at 2500MHz. Now if you manually lowered the voltage slider to 1100mV, this is now the voltage the card will use at 2600MHz, so you can increase the max boost to like 2700MHz and it will rise to 1200mV again. My 6800XT has a stock boost of 2400MHz, when I set the voltage to 1050mV, it's doing that. But I've also increased the max boost to 2550Mhz, resulting in a voltage around 1120mV.
well you just confirmed a behaviour which many already saw. bringing up the min clock boosts stability because it lessens those extremely low drops because of the power management. max clock sets the point of the voltage curve from where you can set your voltage offset. your voltage offset is set with the wattman voltage slider and its calculated -1150= your offset. it was really annoying for the first time until someone explained it to me it was an offset and not an actual voltage set tool, like before RDNA(2)..... driver sets max voltage for 6800xt is 1150mv, 6700xt has 1200mv. but the voltage curve set from the max clock can exeed those values, so your offset might not even be set because 1300mv - 50mv is still over the max 1150mv...
I noticed something very odd with my 6800xt ref. I would run superposition in windowed mode and play with the OC and tried to see changes. The mhz suggestions did almost nothing, however as I lowered the voltage from 1150 to something less, the framerate would increase, but the mhz reported by HWInfo was always the same, both the GPU Clock and Effective clock stayed locked in at say 2600mhz, but my framerate could change by 10% or more. I think either its lying about the clock rate, or its disabling cores to keep within the 293w target. If thats the case, you might have to be careful adding too many volts, the power limit seems to override it and you might end up with less performance from higher volts.
@@manicdan481 2800mhz might be unrealistic, on my water cooled reference 6900xt 2650mhz is about the highest I could get with all powerplay mods stable. pulled 380 watt in stress test
@@ActuallyHardcoreOverclocking Isn't that related also to the issue users have while also gaming and getting stuttering in some games who don't utilize the Card in total because of that aggressive regulation? It behaves super weird.
@@KleinMeme I do have stutter in all games, but not in the benchmarks! In games it freezes like 1ms and runs again, I tried everything to fix/control this but couldn’t figure out unfortunately
@@ActuallyHardcoreOverclocking I had a bad experience with 6900 xt because of this unstable clocks, micro stutters in games and when those happened in games, the core clock, memory clock, power draw and gpu load where deep to 0 and back to top. (Everything else is/was ok in my pc)
@@ActuallyHardcoreOverclocking I know lol. I'm running a 6900xt sapphire MBA on water. The 1175mv is holding it back a little now, hit a brick wall. I've just changed to a B550 unify, my 8pack C14 (4x8GB) isn't very happy on it. Posting has been fun. Hard limit at 3600mhz on 4 sticks after that for memory controller drops to half clocks for fun! So time for a new 2 stick kit of ram. Half thinking of geting "Crucial Ballistix MAX BLM2K16G40C18U4B 4000 MHz" £220.99 til midnight.
I’m building my first pc but I want to get all the components in the correct place. Do you have a video that explains all that or that you would recommend? I already bought all my parts.