AMD Radeon RX 590 Launches, Linux Support Presumably Okay

[nomadewolf]

Didn't RX480 launch in mid-2016 ?

It's been so long, i don't even remember!...

[L_A_G]

...

The problem with your 580-590-clock argument is that you're obviously basing your 580 maximum clock figures on what the launch cards could do. I specifically mentioned that when the 580 came out the 480s you could buy at the time clocked so much better than the launch ones that the primary difference between a new 580 and 480 was that the 580s had higher out-of-the-box clocks.

When the 580 came out people actually did compare a launch 480, then new 480 and a 580 to find that the new 480 was pretty close to the 580 in terms of how far you could push them. Thus when they're doing the same thing (new version of the same card with a slightly improved fabrication process) again, it's obvious that the clocking situation is going to be a repeat of last time.

Obviously your 480 won't compare to a 590 in terms of how far you can take it, but the point was about new 580s that clock higher than any 480 and 580s at launch.

Me? Well I'm perfectly satisfied with the 1070 Ti that I got for VR use this spring, but I have been thinking about getting an RX Vega 56 (for various reasons) if I can find one that's cheap enough.

[Jabberwocky]

Way to completely miss the point on comparing price to a second hand product in a flooded second hand market... Because nothing new compares in terms of pure value to a second hand product when the market is a buyer's market. It's the same reason why car reviews compare prices to other new cars, not second hand ones.

As for the 590 clocking higher than 580s, those comparisons are between out-of-the-box clocks and cards always have headroom over the out-of-the-box clocks. Overclock a 580 and you have something equivalent to an out-of-the-box 590. Additionally, the thing about production processes is that they mature over time, meaning that a new 580 will clock better than one manufactured closer to launch. We saw this in how the last 480s clocked pretty similarly to the first 580s.

I'm with audi.rs4 on this one. It looks like the stock 590 is clocked higher than what the average 480/580 can go. If the production process improves over time the gap will just increase. My 480 (sapphire nitro+ OC) did not have much headroom on core clock (sapphire did a good job). GPU-Z validated it as top 95% ASIC quality. My card's stock speed is higher than a reference 580, but there's no way I will be able to push it to a "out-of-the-box" clock of a 590.

I am still more than happy with my card. Power usage / temperature is important to me, so I won't even consider getting a 590. If I was an overclocker I would probably go for it "just because". What I am wondering is will this be the only GPU on GloFo 12nm? IIRC Nvidia Turing is TSMC 12nm and AMD must be looking at 7nm next.

If it wasn't for lack of freesync and power draw I would not have sold my Tahiti XT. I'm looking forward to 7nm and hopefully 16GB GDDR6 gaming card, not sure if 2019 would surprise me or that I will have to wait till 2020.

[accumulator]

Well the must surprising that Michael when tested all this hardware doesn't have any hang up with the kernel using amdgpu, and doing serveral benchmark. So far I just testing any benchmark make my system to gpu reset freezing all the machine.

Interesting.. I have the same issue. Also spontaneous reboots when idling?
Only after adding admgpu.dc=0 to the module options I haven't seen these annoying reboots anymore.

[L_A_G]

...

Way to completely miss the point on comparing price to a second hand product in a flooded second hand market... Because nothing new compares in terms of pure value to a second hand product when the market is a buyer's market. It's the same reason why car reviews compare prices to other new cars, not second hand ones.

As for the 590 clocking higher than 580s, those comparisons are between out-of-the-box clocks and cards always have headroom over the out-of-the-box clocks. Overclock a 580 and you have something equivalent to an out-of-the-box 590. Additionally, the thing about production processes is that they mature over time, meaning that a new 580 will clock better than one manufactured closer to launch. We saw this in how the last 480s clocked pretty similarly to the first 580s.

[tuxd3v]

https://techreport.com/r.x/2018_11_1...value-99th.png
Seems like good value for what average gamers want: 60fps+ 99% of the time at 1080p in the most popular titles at their highest settings. For my purposes, it looks like it's good value for getting more 6.9ms frames on the desktop than I do with my (standard clocked) WX 7100.

Yup,I always said that, a new revision of RX5xx makes sense..
It sits in a very good place..at nice performance, bellow 300$ and been more efficient than rx580.

[microcode]

https://techreport.com/r.x/2018_11_1...value-99th.png
Seems like good value for what average gamers want: 60fps+ 99% of the time at 1080p in the most popular titles at their highest settings. For my purposes, it looks like it's good value for getting more 6.9ms frames on the desktop than I do with my (standard clocked) WX 7100.

[dungeon]

AMD released the youtube video comparing it to Radeon RX480, link below.
https://www.youtube.com/watch?v=VRAPmwgMAbc

OK, up to 27% perf over original year 2016. Polaris RX 480 ... or on average probably half that

1/4 more perf is not nothing i guess

[Michael]

Any benchmarks for compute? Would be nice to see how it compares to nvidia offerings(especially pascal and turing) for tensorflow or similar.

Yes as said in the article, will have compute tests in the days to come after the gaming tests.

[polarathene]

I haven't regularly played any games in a decade+, only benchmark them.

Any benchmarks for compute? Would be nice to see how it compares to nvidia offerings(especially pascal and turing) for tensorflow or similar.