• 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 RX9070XT / RX9070 Owners Thread

Several things.
  • Check your motherboard BIOS is up to date, and you reset to defaults.
  • Do a true clean re-install of your OS e.g totally wipe your drive, and don't connect to the internet until you have installed your AMD chipset drivers (if Ryzen) and GPU drivers.
  • Check your PSU connection to your GPU/and PSU end, and make sure you are using separate 8-pin connectors, not two from the same cable.
  • Install one game only, that you are seeing the issue with then test that.
BIOS is up to date. PSU is a Corsair RM850x and connections have been checked. OS redo is something that will have to wait, although I could try a seperate drive and do it that way.
 
I seem to be having issues with my 9070XT Pulse.

Desktop stuff is fine. Across multiple games (DayZ, Cyberpunk, Generation Zero, Atomfall, etc) I will play for an amount of time, sometimes its immediate and sometimes I can be 30 minutes in the game, I will get a black screen and monitor loses its DisplayPort signal and switches to standby. My keyboard doesn't fully respond, but when I was in DayZ earlier I could still push-to-talk to another player and game was still running. I couldn't get to task manager or anything like that as there was no display.

I've tried DDU and AMD Cleaner to install latest and previous drivers. No change.

Have this in Event Log: Display driver amduw23g stopped responding and has successfully recovered.

2 minutes later this: Application RadeonSoftware has been blocked from accessing Graphics hardware.

Anyone else seeing this behaviour? Looking online I can see many similar situation on older gen cards.
As much as this sounds like something else. I wouldn't rule out power. I would be tempted to see how it behaves if you undervolt by say 50mv and lower the top MHz by say 100hz. If that works you need to rule out of it's the card or power supply
 
It's memory heat issues the Pulse suffers from under full load. This effects performance over long runs as the card throttles the memory.
Gamers Nexus did a teardown to find out why the hotspot and mem temps were so high. Basically just down to basic construction done to a budget. I believe there was over 30 degree differential between hotspot and GPU temps. They came to the conclusion the hotspot wasn't an issue but the mem temp was.
It's a very good card, probably the best budget one, but under full load the mem temps aren't great and benchmarks aren't long enough to show the impact of this.
It would be good to see the impact in a long gaming session.

TPU did record the average GPU clock speed for both Nitro and Pulse though. Nitro averaged 3239 and Pulse averaged 3078.

Gamers Nexus on Mem Temps of the Pulse: "Either way, it’s not in throttling territory and is still OK, but the memory temperature is a problem: It’s running at 90 degrees Celsius. The reason for this isn’t anything beyond Sapphire’s cooler design being relatively simple and its fan curve being relaxed as it is becomes a problem."

https://gamersnexus.net/gpus/tearing-down-sapphires-rx-9070-xt-pulse-thermals-fan-response-noise

OMG this has me thinking about the nitro again...it does look better, but RGb and nearly 200Mhz...I'm gonna have to change my order again!!
 
It's memory heat issues the Pulse suffers from under full load. This effects performance over long runs as the card throttles the memory.
Gamers Nexus did a teardown to find out why the hotspot and mem temps were so high. Basically just down to basic construction done to a budget. I believe there was over 30 degree differential between hotspot and GPU temps. They came to the conclusion the hotspot wasn't an issue but the mem temp was.
It's a very good card, probably the best budget one, but under full load the mem temps aren't great and benchmarks aren't long enough to show the impact of this.
It would be good to see the impact in a long gaming session.

TPU did record the average GPU clock speed for both Nitro and Pulse though. Nitro averaged 3239 and Pulse averaged 3078.

Gamers Nexus on Mem Temps of the Pulse: "Either way, it’s not in throttling territory and is still OK, but the memory temperature is a problem: It’s running at 90 degrees Celsius. The reason for this isn’t anything beyond Sapphire’s cooler design being relatively simple and its fan curve being relaxed as it is becomes a problem."

https://gamersnexus.net/gpus/tearing-down-sapphires-rx-9070-xt-pulse-thermals-fan-response-noise

I just changed the default fan curve on the Pulse and now the temps have come down quite a bit. The default fan curve is unnecessarily quiet to be fair.
 
It's memory heat issues the Pulse suffers from under full load. This effects performance over long runs as the card throttles the memory.
Gamers Nexus did a teardown to find out why the hotspot and mem temps were so high. Basically just down to basic construction done to a budget. I believe there was over 30 degree differential between hotspot and GPU temps. They came to the conclusion the hotspot wasn't an issue but the mem temp was.
It's a very good card, probably the best budget one, but under full load the mem temps aren't great and benchmarks aren't long enough to show the impact of this.
It would be good to see the impact in a long gaming session.

TPU did record the average GPU clock speed for both Nitro and Pulse though. Nitro averaged 3239 and Pulse averaged 3078.

Gamers Nexus on Mem Temps of the Pulse: "Either way, it’s not in throttling territory and is still OK, but the memory temperature is a problem: It’s running at 90 degrees Celsius. The reason for this isn’t anything beyond Sapphire’s cooler design being relatively simple and its fan curve being relaxed as it is becomes a problem."

https://gamersnexus.net/gpus/tearing-down-sapphires-rx-9070-xt-pulse-thermals-fan-response-noise

The reason the Nitro has higher core clocks is because it is using 10% more power as standard. Those extra core clocks are worth about 3% extra performance. 10% more power for 3% more performance is not my idea of a remotely good trade off. When I set a higher PL, the avg core clocks go up. It’s how it all works and why the Nitro is more expensive. It’s always been the way with the OC cards that have higher power requirements.

I’m sorry but this is all made up nonsense about throttling etc. even your own link tell you throttling isn’t an issue. Mine gets to about 88c under very long gaming sessions at stock. With my more efficient settings it hits about 84c. I can set a more aggressive fan settings and it will max out at 80c.

Show me where you found evidence that a 9070 XT pulse is throttling due to VRAM temperatures please?

Because I can assure you it most certainly does not throttle because of VRAM temperature, even if I push it to +10% PL and let VRAM sit at 90c or even 92c for longer periods.

Techpowerup show a VRAM temperature difference of 3c between a Nitro + and a Pulse at normalised noise levels. 77c Nitro + and 80c Pulse

 
Last edited:
The reason the Nitro has higher core clocks is because it is using 10% more power as standard. Those extra core clocks are worth about 3% extra performance. 10% more power for 3% more performance is not my idea of a remotely good trade off. When I set a higher PL, the avg core clocks go up. It’s how it all works and why the Nitro is more expensive. It’s always been the way with the OC cards that have higher power requirements.

I’m sorry but this is all made up nonsense about throttling etc. even your own link tell you throttling isn’t an issue. Mine gets to about 88c under very long gaming sessions at stock. With my more efficient settings it hits about 84c. I can set a more aggressive fan settings and it will max out at 80c.

Show me where you found evidence that a 9070 XT pulse is throttling due to VRAM temperatures please?

Because I can assure you it most certainly does not throttle because of VRAM temperature, even if I push it to +10% PL and let VRAM sit at 90c or even 92c for longer periods.

Techpowerup show a VRAM temperature difference of 3c between a Nitro + and a Pulse at normalised noise levels. 77c Nitro + and 80c Pulse

You obviously didn't read the article. The first part of the qoute is related to the high hot spot temp not VRAM. It then goes on to state VRAM at 90 is a problem. I didn't make anything up.
You then go on to state VRAM temps at a 250w load which has no bearing on full load temps as that's 80-100w off their full power limit.
 
I just changed the default fan curve on the Pulse and now the temps have come down quite a bit. The default fan curve is unnecessarily quiet to be fair.
This is what they suggest in the article.
It says the stock fan is too slow to react due to how it uses the GPU core temp to respond to and doesn't care about VRAM temps. They also say it can be increased with no audible difference due to how quiet it starts off at.
 
Last edited:
You obviously didn't read the article. The first part of the qoute is related to the high hot spot temp not VRAM. It then goes on to state VRAM at 90 is a problem. I didn't make anything up.
You then go on to state VRAM temps at a 250w load which has no bearing on full load temps as that's 80-100w off their full power limit.

I did read the article and I’m telling you as someone who owns a 9070 XT Pulse, that the VRAM temperature is NOT a problem. I also gave you multiple scenarios of how the temperature behaves. Stock, max PL overclock and with a more aggressive fan curve.

I don’t know how else to tell you my experience is the Pulse VRAM temperature (or hot spot) actually DO NOT cause any throttling.

I mean according to Techpowerup, the Nitro + has worse VRAM temperatures than the Pulse. The reality is these temperatures seem to be a design feature of the 9079 XT. Mid 80s and even higher are normal.

 
Last edited:
At end of the day a £65 difference between a base card at msrp and the top card by the same aib isn't a lot....just look at msi and asus for the 5070ti/5080/5090...you can be paying £500 more to go to their top cards from their msrp cards, even with the msrp cards being well over msrp
to me, the nitro was worth it...metal shroud rather than plastic, beefier cooler, I like the hidden cable system, makes my build look clean and tidy, and i like the rgb strip...compliments my build. you can also undervolt it like crazy too...seeing as there's no special binning of gpu die's, performance between them all will be marginal if you play around with everything yourself, but the nitro is a top performer..just like with nvidia gpu's where there's only a few fps differnce between msrp and top tier cards
so if you have a build where you're already stretching your budget to get the 9070xt, your pc case sits under a desk out the way, or you can't see the gpu as yopu don't have glass side panel or it's dark with no rgb etc, save the cash and get the pulse (on my sons build I've got a vertical gpu bracket...think as seeing the 3 fans rather than the front, I'd go with the pulse on that build)...but if you have a case where everythings on show and you want the clean look and rgb, just pay up..it looks good, no cables is a winner, is silent (well mine is), and (my two previous gpu purchases have been nvidia fe cards, so msrp), I have no buyers remorse paying up for the nitro over the pulse...if i was offered the choice now, I'd do the same...£200 more..pulse it would be though
it's literally just personal choice and only you can decide if £65 is worth it
 
Reinstalled the drivers and getting slightly better results in Time Spy:

Overall:27285
GPU: 31725
CPU: 15217

I was hoping to be up aroung 30000 overall with this setup though.

its just the cpu holding you back now(on the score) the GPU is where is should be now atleast
the cpu score is just becuase people like me have 32 cores, in game your system would kill mine...
 
Last edited:
its just the cpu holding you back now(on the score) the GPU is where is should be now atleast
the cpu score is just becuase people like me have 32 cores, in game your system would kill mine...
@martos123 Exactly this. It's working perfectly fine. 31725 is a good GPU score for this card, I wouldn't worry about your CPU score in 3dmark.
 
Memory isn't throttling at 90C, That is absurd. If it was, then all the cards are on the edge of throttling. 3C in favour of the Nitro for memory, is nothing.

Also GN stopped being a serious channel when they started to exaggerate and dramatise everything.
 
Last edited:
Have you changed VRAM to fast timings and set to 2700/2760 to see if you can eak more out of the power limit with UV option?
May have an adverse effect on the GPU core frequency due to increased power usage, but worth a try if not done already.
100mV just wasn't an option with Steel Nomad, didn't like it, in fact any time my core hit 3400+ I crashed, I settled on -75mv, but had to limit to +8% for pushing as +10% took the core too high again.

Happy enough with my result at -75mv and +8% (Puts me in the top 60th percentile), though will be running at either -10% and -18% mostly.


NHi4N6H.png
 
Last edited:
100mV just wasn't an option with Steel Nomad, didn't like it, in fact any time my core hit 3400+ I crashed, I settled on -75mv, but had to limit to +8% for pushing as +10% took the core too high again.

Happy enough with my result at -75mv and +8% (Puts me in the top 60th percentile), though will be running at either -10% and -18% mostly.


NHi4N6H.png
This is awesome stuff.
This boost crash is the same as what I found with my card in Steel Nomad, anything above 3400 and it would crash shortly after hitting that speed. Reducing PL, as you did, or giving a negative GPU frequency offset was the way to get round this, but this eats into some of the gains.

The UV definitely gives the best performance improvements and matching stock with a -18% power limit is very good! That must run pretty cool.

Do you know of any tools that let you play with the Voltage / Frequency curve rather than just set an overall offset? I did this on my 3090 FE to get the most out of it as at certain frequencies you could have a large UV and others the UV needed to be less, basically fine tuning the UV across the frequency spectrum.
 
Last edited:
gddr6 throttle temp is 105c with my memory at 2800mhz it only hits high 80's so i dont see there been a problem
Yeah, with a 15 degree overhead, it's not even close. I wonder why Gamers Nexus think it's an issue, they don't explain. I didn't know what the throttle temp was so assumed, very incorrectly, it was around 90 degrees with them saying it was a problem.
 
Yeah, with a 15 degree overhead, it's not even close. I wonder why Gamers Nexus think it's an issue, they don't explain. I didn't know what the throttle temp was so assumed, very incorrectly, it was around 90 degrees with them saying it was a problem.

you also have to take into account fan speed, bad thermal contact and many other factors.
with an heavy OC my ram dosnt hit the temps they got running at stock???

we all know that one lad that had to RMA a GPU for running at 100c out of the box. if they had tested 10 and got the same results then thats more conserning but still the temp is not realy a problem temp
 
Last edited:
Back
Top Bottom