XFX Radeon RX 590 Fatboy 8 GB

[PerfectWave]

yeah that's the frequency, it looked strange to me too

I have rx480 and in multimonitor only the ram ramps to 2000 mhz the cpu clock stay at 300 mhz.

 

[phill]

I'm a little underwhelmed with the 590. If it was getting closer to the 1070 or at least faster than the 980 Ti for the same money, then it might be a more of a bargain than it currently is. I'm not sure why they have released it in a way as it's priced a little high for the performance (in my opinion) it gives..

The RX 580 should have been around the GTX 980 performance or slightly higher and this RX 590 should have been around at least 980 Ti performance and closer to the 1070 if it was trying to compete in my eyes.. You can get GTX 980's around the £175 to £200 mark then 980 Ti's around the £250 mark..

 

[AMX85]

RX-580 needs 64 ROPS update and faster memory speeds.

has memory bottleneck, check OC numbers, 2.5% OC on GPU, 12.5% on VRAM and 5.8% higher performance

greetings

 

[bug]

I'm a little underwhelmed with the 590. If it was getting closer to the 1070 or at least faster than the 980 Ti for the same money, then it might be a more of a bargain than it currently is. I'm not sure why they have released it in a way as it's priced a little high for the performance (in my opinion) it gives..

The RX 580 should have been around the GTX 980 performance or slightly higher and this RX 590 should have been around at least 980 Ti performance and closer to the 1070 if it was trying to compete in my eyes.. You can get GTX 980's around the £175 to £200 mark then 980 Ti's around the £250 mark..

Imho, it if offered the same performance as the 580 with reduced power draw and for a tad cheaper, it would have made more sense.

 

[lexluthermiester]

Based on the info available and the conversation here, one has to wonder if the 590 is an outright replacement for the 580?

 

[newtekie1]

I have rx480 and in multimonitor only the ram ramps to 2000 mhz the cpu clock stay at 300 mhz.

W1z's previous reviews of the RX 480 confirm that you are correct. Perhaps the high multi-monitor core clock and hence power consumption is just a driver bug.

Based on the info available and the conversation here, one has to wonder if the 590 is an outright replacement for the 580?

Pricing suggests it is not.

 

[theravenesia]

wish i've that vga card

 

[Jism]

Man, half internet is exploding over the fact that the RX590 uses more power then a 1070. Well duh... it's a 2nm refinement which is compared to a RX580 ~ 50 watts less (if you take undervolting into accounts with identical 1380mhz clocks from the RX580). It's never intended to be a better chip but due to the better proces, it allows for higher clocks up to 1700Mhz and perhaps even some more too, since not many reviewers tried watercooling or so.

At RX580 speeds your looking at a <150W card and once clocked like this sample it's doing 225W. It's already out of spec once you OC it. This card is no upgrade coming from a 480 or 1060 card at all. It's just a refinement proces which gives some advantages over a 480/580.

 

[Fluffmeister]

No doubt it would be more efficient at the original 480 or later 580 clocks, but it would still be less efficient than the 1060 and obviously no faster still... in short kinda defeating the point of it's existence.

It's a shame clocks are pushed at the expence of efficiency, and joe pub having to undervolt is a fix rather than a bonus on top of improved efficency savings, but hey ho, here we are.

 

[HD64G]

Polaris arch isn't as clock-high friendly as Vega and thus, to up its clock you need much higher consumption. To be able to get 980Ti performance on average and higher than that in most recent games for $275 with 3 new games though (close to $100 of added value), it isn't bad as a product at all. In fact it is placed where nVidia doesn't have an offering at all. Imho, that's a smart move from AMD considering their restraints until next year's Navi arrives. And 570-580 are in good prices now also getting 2 of the 3 games offered with 590.

 

[SoNic67]

People keep comparing apples with oranges. I cannot understand complains about a card that's $280 ( RX590, newegg), having the same performance as a card that's $360 (GTX1070), almost 30% higher.
I think this card has it's place on market.

 

[bug]

People keep comparing apples with oranges. I cannot understand complains about a card that's $280 ( RX590, newegg), having the same performance as a card that's $360 (GTX1070), almost 30% higher.
I think this card has it's place on market.

But I'm not complaining about that. I'm complaining both the 580 and the 1060 are within 10% of this, but cost $50 less (that's 20-25% less). Both having tamer power requirements. Sure, if you want to look strictly at performance and disregard everything else, this card is great.

 

[SoNic67]

AMD was always famous for it's high power requirements, why ruin that reputation?

 

[bug]

AMD was always famous for it's high power requirements, why ruin that reputation?

They've had better days.

 

[FordGT90Concept]

Pricing suggests it is not.

Intro pricing. They want to give AIBs a chance to move all of their RX 580s before pricing cutting the RX 590 to where the RX 580 is today. This is RX 480 all over again (RX 580 debuted for high price then replaced it). RX 480, RX 580, and RX 590 are all effectively the same card with the only difference of note to consumers is higher clockspeeds.

 

[SoNic67]

Yeah, why are we surprised by that pricing game? It's just normal business practice.
Also, they have small fixes for the bugs - like the issue of dual monitor with different resolutions that affected 480 power draw is not present in 580.

 

[HD64G]

Some interesting findings on Polaris 30 vs 20 efficiency and undervolting tests. 12nm give much more efficiency if tuned properly and AMD continues to sell heavily overvoltaged chips for some reason.

https://www.tomshw.de/2018/11/22/ef...n-erkenntnissen-beim-untervolting-igorslab/2/

 

[hat]

If only I had a decent board in my secondary rig. I'd shove my 1070s over there for mining, and get one of these cards (or one close to it) to support team red...

 

[bug]

Some interesting findings on Polaris 30 vs 20 efficiency and undervolting tests. 12nm give much more efficiency if tuned properly and AMD continues to sell heavily overvoltaged chips for some reason.

https://www.tomshw.de/2018/11/22/ef...n-erkenntnissen-beim-untervolting-igorslab/2/

Those graphs are dead-wrong. The last graph says there's almost a 30% power saving for 590 at 1500MHz. And yet, if you scroll up, you will find that 30% difference in the core voltage graph. The power draw graph shows something like a 15-20% improvement (hard to tell since it's not labeled right).

Anyway, what do you think is going on here?
a. AMD doesn't know their own products
b. The cards probably need those voltages to be stable outside of a few (unspecified length) benchmarks?

 

[hat]

Why should anyone have to undervolt their graphics card? Flashy graphs that say "look, it's not that bad! (when undervolted)" don't mean squat to me.

 

[HD64G]

Those graphs are dead-wrong. The last graph says there's almost a 30% power saving for 590 at 1500MHz. And yet, if you scroll up, you will find that 30% difference in the core voltage graph. The power draw graph shows something like a 15-20% improvement (hard to tell since it's not labeled right).

Anyway, what do you think is going on here?
a. AMD doesn't know their own products
b. The cards probably need those voltages to be stable outside of a few (unspecified length) benchmarks?

Apart from the graphs not being perfect there are 2 major points when thinking of those results.

1st point is that Polaris 30 has much better power consumption for the same clocks than Polaris 20.
2nd point is that AMD tends to set their GPU's voltage much higher than they could. Why they do that? That's a nice topic for tech journalists to seek for answers.

 

[bug]

Apart from the graphs not being perfect there are 2 major points when thinking of those results.

Wrong != not perfect.

1st point is that Polaris 30 has much better power consumption for the same clocks than Polaris 20.

Smaller production nodes are more efficient than bigger ones. Congrats on discovering warm water!

2nd point is that AMD tends to set their GPU's voltage much higher than they could. Why they do that? That's a nice topic for tech journalists to seek for answers.

Assumptions... As I have already told you, voltage levels are set to ensure stable operation. But you choose not to see that.

Seriously, if you have nothing to add, not posting is an option.

 

[W1zzard]

2nd point is that AMD tends to set their GPU's voltage much higher than they could. Why they do that? That's a nice topic for tech journalists to seek for answers.

Yields

 

[HD64G]

Wrong != not perfect.


Smaller production nodes are more efficient than bigger ones. Congrats on discovering warm water!

Assumptions... As I have already told you, voltage levels are set to ensure stable operation. But you choose not to see that.

Seriously, if you have nothing to add, not posting is an option.

The irony about 12nm helping efficiency of Polaris 30 vs 20 is just showing your ignorance. 12nm isn't so much different as you think-suppose.

As for the graphs in my post, there are faults indeed for the % of power saving in stock vs uv testing. But the curves' difference of voltage/clock scaling show much about how much more efficient the new chip can become if tuned by hand. For instance when using 150W, the 580 can run @1225MHz and the 590 @+1400Mhz while at 200W they can run @1370 and 1580MHz respectively.

Vega GPUs increase of performance through undervolting is a fact. My RX580 is also stable 0.1V below the default core voltage.And my FX8350 works perfectly fine for months now at 1.3V where the motherboards default-auto core voltage is 1.4V. Another proof of AMD's overvolting their chips above their needs.

Yields

I hear that @W1zzard and might be the best shot indeed. Imho, the biggest reason behind all that is that the software (drivers-OS) isn't so advanced to always keep up with the flactuations happening in the cpu-gpu's clock/voltage combos-steps, especially now that the GPUs have more complex turbo features built into them with many sensors sending data to decide what is the max limit in any specific moment. As a result, binning chips has become more time consuming than before and they don't have so much time-money to spend in it. So, AMD plays that on the safe side and one has to manually find the chip's lowest stable voltage.