AMD RX 7000 series GPU Owners' Club

[3x0]

@Krit Here you go. AV1/HEVC both at 10mbit/s, Returnal Benchmark. Used AMD driver to record and MPC-HC to decode and save images. You're using WMP to play AV1 and VLC to play HEVC, might be some decoding differences between the players.

AV1


HEVC

 

[Krit]

@Krit Here you go. AV1/HEVC both at 10mbit/s, Returnal Benchmark. Used AMD driver to record and MPC-HC to decode and save images. You're using WMP to play AV1 and VLC to play HEVC, might be some decoding differences between the players.

AV1
View attachment 364654

HEVC
View attachment 364652

Thanks for answer! To me looks like you also HEVC looks slightly better. I was using OBS becouse it takes down less fps than amd software. Windows media player does no support HEVC thats way i was using VLC for it. When i was recording through amd software the difference between AV1 and HEVC to me it seemed smaller than in OBS.

 

[Athlonite]

There are a lot of RX 7000 owner's someone definitely had tested AV1 vs H.265 (HEVC) video quality at some point. Same bitrate, same resolution AV1 vs H.265 (HEVC) when recording video games in motion.

I need a confirmation which one has better video quality when recording pc games.

I asked this question in AMD Community (Radio Silence) the same here in Graphics Cards Thread no legit answers just circling around and talking about completely different things.

The H265 HEVC looks ever so slightly better but I bet the file is larger than the AV1 version and that's what it is about similar quality for a smaller file size which is what H265 HEVC did to H264 in the same way smaller file size for similar quality if all other setting are the same so say 10Mb 1080p60 each codec will give you better compression for a smaller file size at similar video picture quality

 

[Krit]

The H265 HEVC looks ever so slightly better but I bet the file is larger than the AV1 version and that's what it is about similar quality for a smaller file size which is what H265 HEVC did to H264 in the same way smaller file size for similar quality if all other setting are the same so say 10Mb 1080p60 each codec will give you better compression for a smaller file size at similar video picture quality

I checked bitrate every time the difference is not so big. Something like 10200kbps vs 10400kbps. Even when i manually set exactly the same bitrate for both HEVC will still beat up AV1 with to me is strange because by default AV1 should be better.

 

[wNotyarD]

I checked bitrate every time the difference is not so big. Something like 10200kbps vs 10400kbps. Even when i manually set exactly the same bitrate for both HEVC will still beat up AV1 with to me is strange because by default AV1 should be better.

If you use VLC to play the AV1 video as well, does it look as bad as on WMP?

 

[Athlonite]

I checked bitrate every time the difference is not so big. Something like 10200kbps vs 10400kbps. Even when i manually set exactly the same bitrate for both HEVC will still beat up AV1 with to me is strange because by default AV1 should be better.

What were the file sizes though were they roughly the same or was one smaller than the other, AV1 is supposed to be a faster encoder than say H265 at the same bit rate and it should produce a smaller file with much the same picture quality as the original
also please use the same program to play the videos comparing MPC-HC to WMP or VLC can result in vastly different image quality as each uses a different decoder

 

[Krit]

Tested again this time with AMD Software Recording and only VLC. This time HEVC file size by default was smaller so i manually settled higher bitrate to match AV1s bitrate. Look at bitrate image.

Right now image looks closer when i use VLC for both but still HEVC is clearly better than AV1 hypejob.

 

[wNotyarD]

Tested again this time with AMD Software Recording and only VLC. This time HEVC file size by default was smaller so i manually settled higher bitrate to match AV1s bitrate. Look at bitrate image.

Right now image looks closer when i use VLC for both but still HEVC is clearly better than AV1 hypejob.

Alright, we found the answer to the quality conundrum: it was the decoder on Windows Media Player.

However, to the little quality loss you have on AV1, how smaller its file is compared to the H265 one?
If it is quite smaller, I'd take the small quality hit any day.

 

[leonavis]

Does anybody know what height the pads have in Phantom Gaming 7800XT?

While the edge sensor is at 65°C, the junction goes over 95. I shouldn't have to replace pads and paste already, but it seems I'll have to.

If I have to, I'll open it up and measure myself... But I'd rather do it in one go, already knowing the heights.

 

[Krit]

However, to the little quality loss you have on AV1, how smaller its file is compared to the H265 one?
If it is quite smaller, I'd take the small quality hit any day.

Usually when recording AV1 files are about ~ 200-400kbps smaller than HEVC files but this time it was otherwise HEVC was about 400kbps smaller by default. There is no gain whatsoever from AV1 to me efficiency is worse.


Let's see now who is more demanding in terms of fps loss while recording....

Fallout 4 fps loss from recording OBS vs AMD Software, HEVC vs AV1

HEVC OBS @50000kbps, 2560x1080 had zero frame drop from 80fps to 80fps
AV1 OBS @50000kbps, 2560x1080 had about 1 frame drop from 80 to 79fps

AMD Software for recording is a clear disaster with massive drops.

HEVC @50000kbps, 2560x1080 had about ~ 7.5 frame drop from 80fps to ~ 72.5fps
AV1 @50000kbps, 2560x1080 had about ~ 7.5 frame drop from 80fps to ~ 72.5fps

 

[kapone32]

Does anybody know what height the pads have in Phantom Gaming 7800XT?

While the edge sensor is at 65°C, the junction goes over 95. I shouldn't have to replace pads and paste already, but it seems I'll have to.

If I have to, I'll open it up and measure myself... But I'd rather do it in one go, already knowing the heights.

Probably 1.5 mm

 

[Athlonite]

Usually when recording AV1 files are about ~ 200-400kbps smaller than HEVC files but this time it was otherwise HEVC was about 400kbps smaller by default. There is no gain whatsoever from AV1 to me efficiency is worse.


Let's see now who is more demanding in terms of fps loss while recording....

Fallout 4 fps loss from recording OBS vs AMD Software, HEVC vs AV1

HEVC OBS @50000kbps, 2560x1080 had zero frame drop from 80fps to 80fps
AV1 OBS @50000kbps, 2560x1080 had about 1 frame drop from 80 to 79fps

AMD Software for recording is a clear disaster with massive drops.

HEVC @50000kbps, 2560x1080 had about ~ 7.5 frame drop from 80fps to ~ 72.5fps
AV1 @50000kbps, 2560x1080 had about ~ 7.5 frame drop from 80fps to ~ 72.5fps

is OBS using your CPU or GPU to encode the video and why pick 50Mbps that's a pretty high bit rate what are you doing mastering a blu ray movie more than 15Mbps bitrate and you're into diminishing returns of video quality vs encoding speed vs file size

 

[Knight47]

Anyone here playing Final Fantasy 14? How stable is the driver for it?

 

[Space Lynx]

Anyone here playing Final Fantasy 14? How stable is the driver for it?

I have never had a single issue and I have hundreds of hours in FFXIV all on AMD rigs, and at least 200+ hours with my 7900 XT, again, never a single issue, fully maxed out at 1440p, I always enable Fullscreen mode to make sure freesync is getting applied, but yeah.

 

[leonavis]

I repasted it and used Thermal Pattys instead of Pads.

It's very strange. The original paste looked fresh, not dried out at all. The card was almost impossible to "open", card and cooling body were very tightly pressed together. After repasting, the temps for junction&edge are down massively, especially junction.

I suspect that the pads somehow hindered the GPU to entirely connect to the cooling body. With patties, that doesn't happen anymore, hence the insanely better temps. Kinda strange mistake for a card on that level. Oh well. I'm happy.

 

[tpa-pr]

This is an insane number to begin with. None or almost none 7900 XTX GPU can achieve such a frequency, let alone 7900 GRE with silicon lottery loser VRAM. 7900 GRE would've been a great value GPU if an average sample could achieve ~3000 MHz GPU / ~2750 MHz VRAM or higher. However, we're hard locked to ~2800 MHz GPU and silicon lottery locked to 2250...2550 MHz VRAM. I'd just buy a 7900 XT at this point tbf if NV GPUs were for some reason a no go. 7900 XT has plenty of VRAM bandwidth.

Late to the thread, but I have seen my 7900XTX boost up to 3.2Ghz. It's not for very long, but it does get up there, and that's at stock. That's the effective clock too, not the user-facing clock.

 

[AusWolf]

Late to the thread, but I have seen my 7900XTX boost up to 3.2Ghz. It's not for very long, but it does get up there, and that's at stock. That's the effective clock too, not the user-facing clock.

It would be nice if lower tier SKUs, like the 7800 XT or 7600 could do the same. Unfortunately, they're artificially power limited. I've always felt there's more potential in those cards that AMD isn't using to the fullest.

 

[Macro Device]

I have seen my 7900XTX boost up to 3.2Ghz

Reading comprehension being a luxury once again. I was regarding VRAM clocks, not core clocks. Having 3 GHz on the former is highly unlikely (not even sure if possible), having 3 GHz on the latter is even boring.

It would be nice if lower tier SKUs, like the 7800 XT or 7600 could do the same.

They don't need it. They need price reduction. Performance per watt is already beyond all repair, we don't need to butcher it even harder by extending their overclocking.

 

[MrSleezebag1964]

Navi31 Owners - have you had long term problems with junction temps? Through a strange turn of events, I've actually got my 3rd 7900XTX about a month ago. I'm at 100% "failure rate" for maxing out junction temps (110°, jet turbine fans, etc.) after ~9 months of daily use. Why did I buy three XTX's, you may ask. Fair question, I didn't really. A brief review:

1. The OG Reference (RIP) - Bought December 2022 from AMD official store. I was relieved to "not have the high junction temp" problem with the launch batch. Typical gaming would see ~89° junction temp. High junction developed ~10 months later. The card also had some instability and artifacting - thankfully AMD honored the RMA and replaced it. (TBH I think the card was fine. But RMA window closing!)
2. The Encore - This was replacement Reference model from 1st RMA. Began daily use in Nov 2023. Again, ran just fine (~89°) for at least 6 months and left service in October '24 and by then it would regularly get 105°+.
3. The Wildcard - (Sapphire Pulse 7900XTX). TLR buddy had Cable Mod brand cables fail and destroy the connectors on this card. Cablemod wrote him a check for after-tax price, no questions asked. (Without needing to send melted card!). He had YouTube guy NW Repair fix power connectors, cost him like ~$200 or so with shipping and everything. He bought the Sapphire Nitro + (jealous) with his blank check. Sold to me for $450. Couldn't pass it up! (Especially knowing no chance to buy a 24GB AMD card soon). Interestingly, after a month or so of running this card, it hits 110° junction temp! It absolutely did NOT for first week or so.

Anyways, with all that out of the way - I have to wonder: Could I just have incredibly bad luck, or does every 7900 XT/X have this happen sooner or later?

Note - aside from gaming ~20 hours or less/week, I have been tinkering with many AI image-gen models. They are very intense, short duration workloads (90 sec or less). Perhaps the rapid heat/cool cycles are worse for the factory paste than long gaming sessions with more steady workloads.

All that to say, planning to replace paste with PTM 7950, and figured I should replace all thermal pads since coolers will be off. I can't seem to find dimensions for the pads anywhere! TechPowerUp's teardown has great pictures of both model's pads, but not sure what thickness to order.

 

[remekra]

Navi31 Owners - have you had long term problems with junction temps? Through a strange turn of events, I've actually got my 3rd 7900XTX about a month ago. I'm at 100% "failure rate" for maxing out junction temps (110°, jet turbine fans, etc.) after ~9 months of daily use. Why did I buy three XTX's, you may ask. Fair question, I didn't really. A brief review:

1. The OG Reference (RIP) - Bought December 2022 from AMD official store. I was relieved to "not have the high junction temp" problem with the launch batch. Typical gaming would see ~89° junction temp. High junction developed ~10 months later. The card also had some instability and artifacting - thankfully AMD honored the RMA and replaced it. (TBH I think the card was fine. But RMA window closing!)
2. The Encore - This was replacement Reference model from 1st RMA. Began daily use in Nov 2023. Again, ran just fine (~89°) for at least 6 months and left service in October '24 and by then it would regularly get 105°+.
3. The Wildcard - (Sapphire Pulse 7900XTX). TLR buddy had Cable Mod brand cables fail and destroy the connectors on this card. Cablemod wrote him a check for after-tax price, no questions asked. (Without needing to send melted card!). He had YouTube guy NW Repair fix power connectors, cost him like ~$200 or so with shipping and everything. He bought the Sapphire Nitro + (jealous) with his blank check. Sold to me for $450. Couldn't pass it up! (Especially knowing no chance to buy a 24GB AMD card soon). Interestingly, after a month or so of running this card, it hits 110° junction temp! It absolutely did NOT for first week or so.

Anyways, with all that out of the way - I have to wonder: Could I just have incredibly bad luck, or does every 7900 XT/X have this happen sooner or later?

Note - aside from gaming ~20 hours or less/week, I have been tinkering with many AI image-gen models. They are very intense, short duration workloads (90 sec or less). Perhaps the rapid heat/cool cycles are worse for the factory paste than long gaming sessions with more steady workloads.

All that to say, planning to replace paste with PTM 7950, and figured I should replace all thermal pads since coolers will be off. I can't seem to find dimensions for the pads anywhere! TechPowerUp's teardown has great pictures of both model's pads, but not sure what thickness to order.

Had the same happen on Gigabyte model, RMAed it and changed to Nitro+, no problems for like a year now.

 

[Zach_01]

Navi31 Owners - have you had long term problems with junction temps? Through a strange turn of events, I've actually got my 3rd 7900XTX about a month ago. I'm at 100% "failure rate" for maxing out junction temps (110°, jet turbine fans, etc.) after ~9 months of daily use. Why did I buy three XTX's, you may ask. Fair question, I didn't really. A brief review:

1. The OG Reference (RIP) - Bought December 2022 from AMD official store. I was relieved to "not have the high junction temp" problem with the launch batch. Typical gaming would see ~89° junction temp. High junction developed ~10 months later. The card also had some instability and artifacting - thankfully AMD honored the RMA and replaced it. (TBH I think the card was fine. But RMA window closing!)
2. The Encore - This was replacement Reference model from 1st RMA. Began daily use in Nov 2023. Again, ran just fine (~89°) for at least 6 months and left service in October '24 and by then it would regularly get 105°+.
3. The Wildcard - (Sapphire Pulse 7900XTX). TLR buddy had Cable Mod brand cables fail and destroy the connectors on this card. Cablemod wrote him a check for after-tax price, no questions asked. (Without needing to send melted card!). He had YouTube guy NW Repair fix power connectors, cost him like ~$200 or so with shipping and everything. He bought the Sapphire Nitro + (jealous) with his blank check. Sold to me for $450. Couldn't pass it up! (Especially knowing no chance to buy a 24GB AMD card soon). Interestingly, after a month or so of running this card, it hits 110° junction temp! It absolutely did NOT for first week or so.

Anyways, with all that out of the way - I have to wonder: Could I just have incredibly bad luck, or does every 7900 XT/X have this happen sooner or later?

Note - aside from gaming ~20 hours or less/week, I have been tinkering with many AI image-gen models. They are very intense, short duration workloads (90 sec or less). Perhaps the rapid heat/cool cycles are worse for the factory paste than long gaming sessions with more steady workloads.

All that to say, planning to replace paste with PTM 7950, and figured I should replace all thermal pads since coolers will be off. I can't seem to find dimensions for the pads anywhere! TechPowerUp's teardown has great pictures of both model's pads, but not sure what thickness to order.

Sapphire Nitro+ 7900XTX
Stock for 11 months and so far so good

 

[Ferrum Master]

ll that to say, planning to replace paste with PTM 7950, and figured I should replace all thermal pads since coolers will be off. I can't seem to find dimensions for the pads anywhere! TechPowerUp's teardown has great pictures of both model's pads, but not sure what thickness to order.

200$ for changing stupid plugs? I need to revise my pricing model, there are way too rich people around.

It was 1.5mm if I recall correctly.

I use the card for more than a year without issues. Well, people say putting card on water is too much trouble... Reading this crap does not look like it. The stock coolers were and still are crap for most cards, that have power target past 300W, for smaller power envelope it does not matter