Soldato
Is not compatible the Nitro & Reference block. Already got a new block for reference.
-
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 RX Vega 64 Owners Thread
- Thread starter Kaapstad
- Start date
Is not compatible the Nitro & Reference block. Already got a new block for reference.
Soldato
@Panos Here you go mate, those settings you wished for. Hope they get you up to 27.5KClick link for big one.
Cheers mate.
Soldato
I see.
I'd probably advertise it with 2 prices. One if they want the block fitted one without and just reinstall the original cooler if they don't want the block.
Soldato
@Panos Here you go mate, those settings you wished for. Hope they get you up to 27.5K
The reference took the settings without complaining.....
https://www.3dmark.com/3dm/30011584
Total 21997
Graphics 27390
Physics 16503
Combined 11126
In a warm room with the Predator on quiet mode...
Edit. Just realised broke the Graphics score....
Damn the 8600K needs a push..... Need to check RAM subtimings and see if I can do 5.3Ghz
Soldato
This is the best so far!The reference took the settings without complaining.....
https://www.3dmark.com/3dm/30011584
Total 21997
Graphics 27390
Physics 16503
Combined 11126
In a warm room with the Predator on quiet mode...
Edit. Just realised broke the Graphics score....
Damn the 8600K needs a push..... Need to check RAM subtimings and see if I can do 5.3Ghz
Associate
If only I could do 1180 on the HBM...The reference took the settings without complaining.....
https://www.3dmark.com/3dm/30011584
Total 21997
Graphics 27390
Physics 16503
Combined 11126
In a warm room with the Predator on quiet mode...
Edit. Just realised broke the Graphics score....
Damn the 8600K needs a push..... Need to check RAM subtimings and see if I can do 5.3Ghz
Soldato
Cheers.
Shame I do not have better CPU. But I believe can do 27500+ with some tweaking. Which card you have and what is the HBM voltage reported in HWInfo?
The reference took the settings without complaining.....
https://www.3dmark.com/3dm/30011584
Total 21997
Graphics 27390
Physics 16503
Combined 11126
In a warm room with the Predator on quiet mode...
Edit. Just realised broke the Graphics score....
Damn the 8600K needs a push..... Need to check RAM subtimings and see if I can do 5.3Ghz
Great result, I did think reference have better hbm2. What do you get in the extreme (next fs level up)
Soldato
I am going to run a full run on all benchmarks today
after tweaking the subtimings on my ram.
Soldato
Here are my thoughts having stripped 2 Vega 64s.
a) The stock paste is thick like cement (feels like Intel TIM but less crumbly), not covering fully the HBM or cores due to it's thicker than usual substance. (Same applies to Nvidia cards)
b) Usage of Kryonaut is a massive upgrade over the cement, but you need to spread it evenly and not use some blobs and pray it will work.
c) On AMD cards, liquid metal is worthless and useless. Only useful to Nvidia cards which are throttling from 30C.
d) On overclocking with Vega the maximum temp limits, as per @TonyTurbo78 post greatly affect performance. if you put them high, the card will draw more power and keep the clocks low, if you set them "just right" at what might seem low, it improves the performance and speed, while the power is kept low.
Seems it working like the PBO overclocking on Ryzen
a) The stock paste is thick like cement (feels like Intel TIM but less crumbly), not covering fully the HBM or cores due to it's thicker than usual substance. (Same applies to Nvidia cards)
b) Usage of Kryonaut is a massive upgrade over the cement, but you need to spread it evenly and not use some blobs and pray it will work.
c) On AMD cards, liquid metal is worthless and useless. Only useful to Nvidia cards which are throttling from 30C.
d) On overclocking with Vega the maximum temp limits, as per @TonyTurbo78 post greatly affect performance. if you put them high, the card will draw more power and keep the clocks low, if you set them "just right" at what might seem low, it improves the performance and speed, while the power is kept low.
Seems it working like the PBO overclocking on Ryzen
Panos the lc64 bios has a 70c max temp limit anyway, but i take it you are on the 220w+50 bios and not the 200w+50 bios switch position.
I agree with the stock tim I have a pic somewhere of my p64, the Tim had failed and caused massive temp spikes and lockups earlier in the year.
I agree with the stock tim I have a pic somewhere of my p64, the Tim had failed and caused massive temp spikes and lockups earlier in the year.
Soldato
The reference took the settings without complaining.....
https://www.3dmark.com/3dm/30011584
Total 21997
Graphics 27390
Physics 16503
Combined 11126
In a warm room with the Predator on quiet mode...
Edit. Just realised broke the Graphics score....
Damn the 8600K needs a push..... Need to check RAM subtimings and see if I can do 5.3Ghz
Brilliant mate, well done. I was hoping you could get past me. I did try 1200mhz on the HBM when I posted that big score, but I got artifacting during the run, 1180mhz I think is as much as I can push it.
The annoying thing is, I can bench at 1802mhz (1780mhz actual) but at 1050mhz on the HBM. Anything higher on the HBM at those ceiling core speed and it crashes. I found Vega doesn't like both core and memory at it's absolute screaming limit, although you can get away with one being there. Which is why I went with a higher HBM speeds and a very good (not incredible) core overclock works best for bench score. I think the benefit of having higher HBM speed for more bandwidth is better than jumping from 1750mhz to 1800mhz on the core.
The paste needs to be thick on the HBM as it isn't quite level with the GPU chip. They should really have thought about that when shaping the heatsink.
Soldato
Both my Vega cards (reference and Nitro) the HBM is leveled to the GPU. Only some initial release Powercolor & XFX Vega 64s had the different elevation.
Soldato
Thank you mate
Yeah, HBM speed is more critical than core speed to gain the performance.
I have seen it also when experimented with "Balanced Mode". If you select the mode and then go Custom, overclock the HBM to 1100, leave everything else intact.
The core speed it will go down around -5% (~78mhz) but the fps performance it will jump a hefty +10% going from 945 to 1100 the HBM overclock.
Soldato
@Panos Just looked at a compare. https://www.3dmark.com/compare/fs/16938414/fs/16741526#
It's quite interesting how your combined is 350 points higher yet you only have a 1fps increase in graphics test1. I wonder where that big leap comes from in the combined.
It's quite interesting how your combined is 350 points higher yet you only have a 1fps increase in graphics test1. I wonder where that big leap comes from in the combined.
Soldato
Actually if you see Combined score and Combine test are both +3.4%. And +1.70fps is 3.4%.
But FS scales nicely. What bugs me is TimeSpy. I know on Test 1 I will beat the GTX1080@2190 by miles, yet it will lose on Test 2 because it has Nvidia code path optimizations, which do not exist in DX12.
Soldato
Yeah I see that but my point is, I wonder where the boost in combined score came from? It can't be from the CPU being at 5.1ghz compared to my 5ghz as in the physics tests and score obviously the 8700k is higher. So what's upping that combined score in the test I wonder? (Giving you the extra 1.7fps in the combined test I meant)
Soldato
Ram & Ring speeds reduce latency.
My 3600C16 ram is OCed to 4000C17 while running 4.5Ghz Ring atm.
That is why I said might gain performance if tighten up the RAM sub timings, as they are pretty loose atm.
Soldato
Ah yeah, I don't touch RAM timings at all, I'm clueless at RAM tweaking. I set xmp 3600mhz and done lol.