AMD RX 7000 series GPU Owners' Club

[Ferrum Master]

@VSG

Mate, why don't try making and article PTM7950 vs china knockoff PTM7950 and Kryosheet on some beefy GPU? At best scenario under waterblock. There aren't that many around having all those new kids. Could be click worthy for the site.

 

[Kabouter Plop]

Just make sure to let it settle in first 5 minutes of PTM7950 can be awful but that was not my experience cos first time it has to melt after it gets better and better, even after a week.

 

[Ferrum Master]

Just make sure to let it settle in first 5 minutes of PTM7950 can be awful but that was not my experience cos first time it has to melt after it gets better and better, even after a week.

I work in RMA buisness... I've been familiar with it for while already as Lenovo uses it in their laptops by default for quite a time already. But yeah... a proper test would not hurt...

 

[Kabouter Plop]

I work in RMA buisness... I've been familiar with it for while already as Lenovo uses it in their laptops by default for quite a time already. But yeah... a proper test would not hurt...

I wish there would be like 1 year run test or well run test until pumpout happens.
See how fast each paste pumps out on a gpu with lot of pumpout, my liquid devil 7900 XTX did mx-4 in 1-2 weeks
I suspect its because of uneven cooling surface but also uneven die surface contributes faster pumpout, as well as really high wattage from chip.
combination of all 3 probably makes it pumpout the fastest.
I do not think PTM7950 would pumpout but still be fun to test as well.

And the differences instead of comparing paste to paste you compare day 1 temps and record all thermal data such as ambient temp in case room temp or maybe do it controlled with waterblock, and keeping water idle temp consistent as possible thru the runtime, and record variances.
And then eventually after couple of days you start record higher temps then you had on day 1.

Since paste tested is always tested under best conditions but what it runs at after couple of days or weeks or months if never reviewed.
So for me MX-4 day 1 was 75c hotspot at i think 23-24c watertemp idle +4c under load on watertemp and after 1-2 weeks it went up to 92c hotspot
since high hotspot always happens due to bad thermal paste aplication or other issues, i think this would be good benchmark for how long thermal paste takes to pump out or degrade.

 

[Ferrum Master]

Since paste tested is always tested under best conditions but what it runs at after couple of days or weeks or months if never reviewed.

Well I have the "joy" seeing what it looks like if you see your repaired item as re-repair or broken again. After years or so...

As cheap de-facto I use MX-2 and not the newer ones... and it is pretty much okay... not best, but not worst, it has really great shelf life. But it has given me the most stable results for average users and devices. I experimented enough with my own devices, like with anything, to find something useful that would improve my professional job... but not yet.

Laptops are bitchy things... as those are bent, thrown otherwise just a humiliating objects with birth defects, then some ask our unclewebb to unlock power limits, so the paste can shit themself over 100C in that gimped thermal solution. So the experience varies if you open up things you have serviced after a while. For me it is different that I do few each day of those, so have formed my opinion on much larger data base. So... that's not like your online Reddit keyboard warrior making statements on your few tries getting into manhood.

So the pump out... I have seen it for any TIM including PTM7950 and liquid metal. Not sure how... you ragequit and toss your laptop while it is hot or something, but I have seen blank spots on both sides(die and heatsink) with no paste at all on any paste/LM used. For LM it could be that ASUS uses only it, it used the first gen Grizzly and it was bad in many regards, but I cannot tell what gen is what when opening up, as there are engineering change notices(ECN), to replace it as a mandatory thing if you have one at sight. For LM there are types... I have seen black spots like burnouts and voids... so... even when using LM it is an active metal and it soaks in somewhere... you have to keep an eye on it... it also spills and travels. Factory often does too much of it. There are variables.

So the pump out effect is really there, especially when great temperature changes occur(hotspots), also laptops are bending. twisting like a cloth as it ain't a completely rigid structure, so the observations vary just by user patterns even when looking on the same model.

I'd like to incline... my choice to experiment with Kryosheet on my 7900XTX for first time stemmed from these observations. I just wanted to put it on and forget... it looks like it works and the results are good.

While me may ask for scientific experiment... the field experience is more important, you have to use a solution that fits all the checkboxes. No sure if we can ask VSG to run furmark, then put the device in backpack in some MTB mountains ride or bumpy roadtrip through gravel roads etc and then compare.

 

[Kabouter Plop]

Just what i needed more despair that i may need to repaste even PTM7950

 

[Ferrum Master]

Just what i needed more despair that i may need to repaste even PTM7950

I am sorry mate.

But in general experience it is marginally better than any conventional paste. But some still have the talent to somehow brake it. It is still a fairly new solution. Also our bumpy and uneven mixed die surface is something new.. with some cavities in the corners, like drains into sewers

We are beta testers... unpaid ones... that actually hurts.

 

[faye]

Despite all my recent ranting I can't justify the additional cost of moving to a 4090 for now, so I guess I'm a member of the club. Sapphire Pulse RX 7900 XTX. I'll get some pictures later.

 

[Nike_486DX]

Despite all my recent ranting I can't justify the additional cost of moving to a 4090 for now, so I guess I'm a member of the club. Sapphire Pulse RX 7900 XTX. I'll get some pictures later.

and dont forget the vhpwr power connector which is still a fire hazard. 3x 8pin is much safer.

 

[Ferrum Master]

Despite all my recent ranting I can't justify the additional cost of moving to a 4090 for now, so I guess I'm a member of the club. Sapphire Pulse RX 7900 XTX. I'll get some pictures later.

I've read your thread... it was one of those raising my eyebrows. But then I remembered during old days I had VIA and SIS boards and it was the same "fun"... I hope you can relate to the message I am trying to say.

I have spent like over the month on my RAM timings. None worked fine in stock, each frequency step had their own timing table for 24/7 stabilty, no they do not match if upping frequency as impendence and parasitics change. Same goes for CPU OC. Not even that... each BIOS has different settings and stability points.

Get used to testing... it takes heavy statistics... small steps, large tests. Just do one thing and do testing overnight. One step... then if OK do further. Each motherboard and CPU combo is really different these days. There are no universal 100% formulas.

 

[TheoneandonlyMrK]

So I have a 7600 In the building, on a pciex 3 board @ x8 rebar on?!?, didn't expect that.
It runs pretty good, it's a sapphire pulse, I like that it's a TRUE successor to the RX580 pulse, it's clearly different but the similarities are numerous.
Runs much, much better than expected, I am impressed given its reasonable cost I bought my rx580 for £219 so £249 is a steal.
Total build £600 , good IMHO.

 

[Speedyblupi]

Looks like I'm a member of the club now. I just installed my new XFX 7800 XT MERC.

 

[Kapone33]

Dual 480 thin rads is not enough you need dual Mo-ra3 with dual D5 pumps to cool a 7900 XTX ⱼₖ
All jokes aside i would seriously recommend PTM7950 if you going to repaste use this, nothing beats it, it will last a lifetime.

I am going to try that this weekend.

 

[thesmokingman]

I am going to try that this weekend.

You're gonna need quad D5 then. Already did the dual 1080 externals years ago cooling quad 580s and 290 lightnings.

 

[Kapone33]

You're gonna need quad D5 then. Already did the dual 1080 externals years ago cooling quad 580s and 290 lightnings.

I meant the PTM7950 I already have a sweet 420 rad for my GPU.

 

[jayjr1105]

@VSG

Mate, why don't try making and article PTM7950 vs china knockoff PTM7950 and Kryosheet on some beefy GPU? At best scenario under waterblock. There aren't that many around having all those new kids. Could be click worthy for the site.

This has already been done...

 

[Ferrum Master]

This has already been done...

Ahh... YT reviews... no thanks.

 

[lowrider_05]

I will only say, if you have a 7900XT/X and high Temps or Thermalpaste pump out problems, you need to get at least a PTM7950 or in my case a Kryosheet. Temps on the same Alphacool Waterblock (with 3 repastes) went from Core: ~60C Hotspot: ~89C to Core: ~45C and Hotspot: ~60C after a Port Royal Stresstest. This was such a great improvement with just this simple thing.

 

[Kabouter Plop]

I will only say, if you have a 7900XT/X and high Temps or Thermalpaste pump out problems, you need to get at least a PTM7950 or in my case a Kryosheet. Temps on the same Alphacool Waterblock (with 3 repastes) went from Core: ~60C Hotspot: ~89C to Core: 45C and Hotspot: 60C. This was such a great improvement with just this simple thing.

You got those temps with kryosheet ? what tbp your card hits at those temps ? if its at 410 that is really good still a really good improvement as good as mine, i only hit about 38-42c edge 60-70c hotspot tho but i can increase edge temp by few degree and drop hotspot some more if i enable instant replay recording via av1 which i do not cos AMD somehow cant add support for HDR tonemaping on recordings, but have no problem adding tonemaping for HDR on their overlays.

Steam btw got tonemaping for screenshots

 

[lowrider_05]

You got those temps with kryosheet ? what tbp your card hits at those temps ? if its at 410 that is really good still a really good improvement as good as mine, i only hit about 38-42c edge 60-70c hotspot tho but i can increase edge temp by few degree and drop hotspot some more if i enable instant replay recording via av1 which i do not cos AMD somehow cant add support for HDR tonemaping on recordings, but have no problem adding tonemaping for HDR on their overlays.

Steam btw got tonemaping for screenshots

This was @+15% Powerlimit so @463Watts with my Card, I can not say anything about Recording, because i never used it.

 

[Kabouter Plop]

You have better temps then me then are you sure you are also hitting 463w port royal has had some problems for me hitting full power so you have to not just look at the max but also current tbp during the test

 

[lowrider_05]

Yes, i had the Radeon Overlay running for the whole time. It does not always stay @ 463 watts but never got under 380 watts
btw, I have big Rad´s cooled with 4 200mm Fans and two D5 Pumps so my cooling capacity is very good and I use Thermal Paste on the VRAM this lets the Thermal Pad perfectly compress on the Die.

 

[Kabouter Plop]

Then you not hitting 463 consistently that will impact thermals

Btw i have dual mo-ra3 dual D5 so my cooling capacity is probably better then yours.
I do not use any undervolt either.

 

[lowrider_05]

Then you not hitting 463 consistently that will impact thermals

Btw i have dual mo-ra3 dual D5 so my cooling capacity is probably better then yours.
I do not use any undervolt either.

OK, i do Undervolt to 1090mv and besides Furmark etc no Benchmark/Game will always sit at max TBP but I don´t want to compete with your Setup/Temps, Just wanted to share how great the improvement was.

 

[Kabouter Plop]

OK, i do Undervolt to 1090mv and besides Furmark etc no Benchmark/Game will always sit at max TBP but I don´t want to compete with your Setup/Temps, Just wanted to share how great the improvement was.

Some games report max tbp but hit about 10c lower hotspot temp sometimes, only exception is ratchet clank rift apart can peek to the max not consistently or world of warcraft which can almost hit it consistently which is why i freaked out when i saw 92c hotspot the soonest.
Heck highest hotspot if ever recorded stock paste was 95-99c hotspot at +15% in a game bang on balls that reports 410w tbp or so but then only hits 62-63c hotspot at stock and 72c hotspot at +15%

Diablo IV for example also reports about 410w tbp but then only hits about 60-62c hotspot at 22-23c watertemp
Most my games are about 60-65 range actually
but if my watertemp drops to 20c i can technically hit max 65c hotspot or 66c technically 65,4 65,5 which rounds up under 1 minute run, what i hit max in 3D occt in 1 minute i never exceed in the games that can peek high hotspot.

So at 20c watertemp altho it has't gotten that cold reliably yet i probably hit about 55-65c hotspot but i dunno if PTM7950 will behave at colder temps it melts at 45c so i wonder how it behaves when it does not reach melting point as reliably yet, altho it never effected idle temps, usually 1c above watertemp

Flow rate atm 275,2 l/h so more then 5 l/m with only my gpu in dual mo-ra3 loop atm haven't added cpu yet cos if not had the time yet, im removing the internal loop cos that would just heat up my loop rather then decrease watertemp that will increase temps.

watertemp is at room temp or 1c above roomtemp while under load its +2c

edit: i just realized you have paste on vram is that even good idea that will dry out overtime as well or make no good contact since pads are usually least 1mm thick sometimes 1,5mm thick depending on cooler design.