Just Plain Mean
- May 18, 1997
We are not going to make the RVII embargo. I screwed some things up, and put Brent well behind schedule. We will a full review next week. Sorry.
Could you look into this? Low mounting pressure-> high temps might explain why so many outlets are seeing high noise/fans ramping super high.After completing all testing and photos, I reassembled my card and wondered if I could bring down Junction Temperature below the 110°C I've been seeing on the stock card. Changing the thermal paste didn't make much of a difference (barely 1-2°C better). My trick for such situations is increasing mounting pressure by adding little metal washers. Pictured above you see two of the washers: one to get a better look and a second one below one of the mounting screws, to indicate the location I placed it in. Of course you want to use a total of four washers — one for each screw.
With the washers installed, my Radeon VII Junction Temperature dropped by around 10°C to 100°C, which is a significant improvement.
Not really that surprising. Making big changes going from the 14nm node to the 7nm node would be more risk than making minor changes and using the die shrink to boost clock speeds for performance instead. And AMD hasn't really had the R&D budget to push bigger changes out before Navi. Vega 20 was always intended to be a stop gap between the original Vega release and Navi.This is a n interesting analysis, clock-for-clock showing how bandwidth-limited VEGA was guaranteed to be if you bumped the clocks with the same old memory bandwidth:
There's no rush.. The folks who are looking to buy this halo card already have it.
Hey guess what, the die shrink did nothing for power consumption, when all you did was up the clocks:
View attachment 140319
Once again it will be up to owners to manually under volt to fix the power consumption issues, assuming you got a golden sample.
But at least it's competitive with the RTX 2080. Nvidia left themselves wide-open this release.
The fan speeds seems to be directly tied to this "junction temp" sensor. Some reviews have seen 2900 rpm. That one sensor of 64 seems to be the main thing that ditactes fan speed regardless of the temp of the rest of the core. Then you have some pieces saying they used washers to remount the cooler which resulted in a lower junction temp of upto 10c and a slower fan speed.Really doubt it's as loud as 290X, blower card fan profiles tend to be way worse than dual/triple fan coolers.
Yes, fan speed is tied to Junction Temperature, not "CPU Temp." CPU temp sensor is a single sensor on the outside edge of the die. Junction Temperature is a series of sensors inside the GPU itself. AMD is monitoring these sensors and controlling clocks and power to specific areas of the architecture. You can monitor both values in AMD Wattman. You can see the delta below.The fan speeds seems to be directly tied to this "junction temp" sensor. Some reviews have seen 2900 rpm. That one sensor of 64 seems to be the main thing that ditactes fan speed regardless of the temp of the rest of the core. Then you have some pieces saying they used washers to remount the cooler which resulted in a lower junction temp of upto 10c and a slower fan speed.
No. I am not saying that it did not give him a bit better temps, but he would have to be dealing with an ambient or case temp that right on the edge of the fan curve to actually impact this.Reckon there's much to the story about using washers to better attach the heatsink for better acoustics?
Touched on this in the other thread. AMD reported a known issue with the black screen instability on X399 boards. So I put it on my personal X399 system and had this issue. This is very much related to the latest version of Win10. They got me over a fixed driver, but that did not fix it, but that driver along with a quick regedit fixed the blackscreen issue. So that is going to handled on the tech side, just need to get that driver, which was not a performance driver, pulled together and put out.TBH, with the many inconsistencies in the R VII performance, as well as driver bugs, stability and poor OC'ing reported by some reviewers, a delayed review could bring some fresh perspective? I'd like to read a much delayed R VII review, say a couple months. Have a feeling the card was rushed out too soon.
Those aren't chiplets. The extra chips around the die are the HBM2 modules. The height differences are due to manufacturing tolerances from the supplier.I was really hoping this card would be a win for AMD... 7nm, 16Gb of HBM2 Vram... It sounded pretty promising.
Maybe they can get a driver out that fixes the main issues by the end of next week. Then [H] will have to recheck some numbers
GamersNexus' card performed better with the stock thermal pad than with paste... Seems a bit like cutting corners with the silicon chiplets at differing heights... that shit needs sorted out before AMD really goes mainstream 'chiplet'.
We already delid CPU's.. what's next, resurfacing? ugh
Once we ran into the OCing problems, we moved to getting 4K included. I however do not think that heatsink and fan design is up to a lot of overclocking.I'm glad you opted to delay the review. The reviews released today were pretty much a mess. I really hope they will be able to get the OCing support figured out soon.