AMD's RX 500 Series Reportedly Delayed

[Raevenlord]

We've previously covered how AMD's RX 500 series is to be a rebrand of the company's successful RX 400 series. Previous reports pegged the RX 500 series' launch on April 11th; now, it would seem that there has been a slight, one-week delay on the launch date, with it having been pushed back to April 18th. Apparently, this delay is looking to allow more time to "fine-tune the drivers".

The RX 500 series are purportedly straight rebrands from equivalent RX 400 series GPUs (RX 580 will be a rebrand of the RX 480, and so on down the ladder). The need for driver fine-tuning seems a little baffling considering these straight rebrands, but may have more to do with the reported Polaris 12 chips that are expected for launch than any other metric. Remember, RX 500 chips are expected to carry somewhat higher clock-speeds than their RX 400 originals, with some improved power/performance ratio being derived from improvements in foundry processes. But if the rebranding scheme holds up, don't expect these to bring in any meaningful changes towards these cards' performance. AMD is hoping Polaris tides them over through the mainstream market until it can introduce its Vega-based, high-performance GPUs, which are heralded to mark AMD's return to the high-performance consumer graphics segment in a while. Fingers crossed.




Source: Thanks @TheMailMan78

View at TechPowerUp Main Site

 

[Deleted member 50521]

Rebranding RX480, which tells you how little resources AMD can allocate to R&D new GPUs. AMD has successfully pulled off RyZen, doesn't look like they can pull off VEGA/RX500 though.

 

[Aenra]

Not sure i get this; assuming i understand what the article implies, why do this now? Who's gonna go for these when we're just a few months shy of Vega?
(including limited budget people, i wouldn't see them purschasing a new series, even if a rebrand, when cost is a factor)

edit: Am trying to exclude the obvious explanation. Not the best way to go about it, delays reminding of past delays except this time in between two major launches..

 

[Basard]

I'm sick of all these numbers. We should just give them names.... like the ones that came on the back of your AOL disk.

 

[newtekie1]

Not sure i get this; assuming i understand what the article implies, why do this now? Who's gonna go for these when we're just a few months shy of Vega?
(including limited budget people, i wouldn't see them purschasing a new series, even if a rebrand, when cost is a factor)

edit: Am trying to exclude the obvious explanation. Not the best way to go about it, delays reminding of past delays except this time in between two major launches..

Vega is just a top on GPU, it isn't displacing the current GPUs in any way. So if you are already considering a RX 480 and not a faster card, you'll buy the RX 580.

 

[cdawall]

I am curious if yields have finally improved enough that we will get the 95w rx480 we were promised.

 

[AntDeek]

Something doesn't make sense here to me, and if someone can explain it, great.

Why would they sell the RX 480 as an RX 580 instead of pushing the 480 more advertising wise? Seems identical to intel releasing Kaby Lake with no IPC improvements over Skylake

 

[Deleted member 50521]

2GHz Polaris core with 2304SP. There ya go hype train started again!

Seriously i am kinda sick of seeing rebrands after rebrands. First the entire 290X to 390X VS, now this shitty rebrands of RX480, which is not even that good to begin with.

 

[dorsetknob]

If they are Rebrands = old stock to shift before vega and as such price should be lower than the old stock price

Ps who says fat chance of that besides me ?????

 

[FordGT90Concept]

Remember, RX 480 was the first time in a very long time (ever?) that NVIDIA and AMD used different fabs for GPUs. It makes sense that TSMCs fab was refined where GloFo was not. I'm hoping for 1600 MHz Polaris but AMD is not going to let that information leak because that is practically the only difference between RX 480 and RX 580. I wouldn't hold my breath for it but I'd certainly hold my VGA purchase until it debuts.

Seriously i am kinda sick of seeing rebrands after rebrands. First the entire 290X to 390X VS, now this shitty rebrands of RX480, which is not even that good to begin with.

There's a meaningful performance difference between the cards though so I personally don't have a problem with it. 290X 4 GiB approximately compares to 390 8 GiB in performance. 390X is quite a ways a head of 290X.

 

[bogami]

I already have a beard down to the ground from waiting for this product and it seems that it soon will not be offered , in the best case in May ! If I rely on the current delays at the end of the year . And if this is only for the competition to GTX1080 ! I am not surprisedAMD has losses at such a high caliber personnel and development .

 

[FordGT90Concept]

I think RX 580 will land between 1070 and 1080. I think you're talking about RX Vega which definitely isn't coming until May/June at the earliest.

 

[dorsetknob]

I already have a beard down to the ground from waiting for this product

"" Its on its way on the way then you can GO COMPARE go compare "

'Go Compare' Guy gets his comeuppance

 

[crazyeyesreaper]

2GHz Polaris core with 2304SP. There ya go hype train started again!

Seriously i am kinda sick of seeing rebrands after rebrands. First the entire 290X to 390X VS, now this shitty rebrands of RX480, which is not even that good to begin with.

Rebrands are what happens when their is no new node that offers greater transistor density / die shrink etc. Why should AMD waste cash to make you happen redesigning a chip from the ground up for a process node they are already wringing out everything they can on with GCN.

Fact of the matter it used to be every gen got a new Node. now we get 2-3 gens per node as foundries fail to keep up with Moore's law.

AMD

HD 2900 = 80nm 420 mm2 Huge die bad performance
HD 3870 = 55nm 192 mm2 die shrink with changes cost effective performance
HD 4870 = 55nm 256 mm2 scaled up design with knowledge from 3k series.
HD 5870 = 40nm 334 mm2 new node design scaled up again good performance
HD 6970 = 40nm 389 mm2 same node new design scaled up again
HD 7970 = 28nm 352 mm2 new node similar die size exceptional performance
R9 290X = 28nm 438 mm2 same node design scaled up design is costly
R9 390x = 28nm 438 mm2 reused design due to cost effectiveness add tweaks
R9 FuryX = 28nm 596 mm2 new design extreme die size very costly bad performance
RX 480 = 14nm 232 mm2 back to small manageble die size profitable design.

Notice the pattern?

When their is no new node with which to shrink the design. AMD has to scale up the die size to be competitive. This is because Nvidia stripped their GPUs of computational abilities and focused on gaming. Where as AMD has a general usage design thats good at alot of things but not great. As such AMD GPUs are more expensive to produce ie larger die sizes vs Nvidia for the same performance. AMD in order to move forward with GCN needs new nodes so they can shrink the design and add more to it. With no new node AMD is stuck playing the waiting game. As such they tend to utilize rebrands with higher clock speeds and tweaked designs. As it allows the usage of older GPU dies and is more cost effective. But then again people with half a brain should know this already. AMD is worth 1/3 of what NVIDIA is let along needing to design CPUs and compete with Intel. They are also an asset light company meaning they have to pay others to produce anything they make. Which means they fight for fab time. All of which drives up costs and hurts there profitability in certain situations. Example in terms of a profitable GPU Fury series was a giant flop. Large Die / lower yields etc. for every Fury they made they could have produced nearly 2x R9 280x or 380x gpus which tended to sell well. They also had much higher yields per wafer meaning greater profitability.

RX 500 rebrand means AMD will likely reuse the chips they have with a few tweaks and higher clocks with similar power targets. Meanwhile Vega will be like Fury a seperate enthusiast line that targets upper crust of consumers. If it performs well AMD will make alot of money if not it will be another failure. However AMD has already placed most of their emphasis on the entry to mid range markets which have the largest sales volume and thus builds their market share. Market share that makes investors happy and by extention their stock price.

 

[OneCool]

RX480 should be rebranded as a RX570 and the RX580 be a different animal altogether.

AMDs release practices are just random BS these days!!!

I've heard the term "dropping the ball" but damn at this point you have to have the ball in your hands first... Really not the case here.

 

[chaosmassive]

2GHz Polaris core with 2304SP. There ya go hype train started again!

Seriously i am kinda sick of seeing rebrands after rebrands. First the entire 290X to 390X VS, now this shitty rebrands of RX480, which is not even that good to begin with.

beat this

HD 7850 > R7 265 > R7 370

 

[newtekie1]

I think RX 580 will land between 1070 and 1080. I think you're talking about RX Vega which definitely isn't coming until May/June at the earliest.

I don't think, even with the speed bump, we are going to see it make up anywhere near that amount of ground.

 

[Steevo]

beat this

HD 7850 > R7 265 > R7 370

Nvidia GF108 it lived about four lives, many others as well.

AMD did the same with older 9xxx series, they lived on many generations, as did some of the X1600 series as 5xxx models.

 

[FordGT90Concept]

I don't think, even with the speed bump, we are going to see it make up anywhere near that amount of ground.

Yeah...probably a little more optimistic on my behalf. ~1070 is probably more realistic. The only reason why Polaris didn't kick ass and take names was because of it's pathetic ~1300 MHz clockspeed. Reference GTX 1070 runs at 1683 MHz. A Polaris 1600 MHz clockspeed would theroretically lead to a 23% boost in performance. That's GTX 1070 territory, no?

 

[crazyeyesreaper]

Ah but see thats the fallacy Clock speed doesnt matter as much as arch.

In this case Nvidia went 16nm AMD went 14nm

Its apples to oranges due to the foundries being different etc. Suffice to say AMD needed 14nm to hit its power target. While 16nm via TSMC would likely have allowed higher clock but also used more power. Its likely AMD could get a 480 rebrand CLOSE to a 1070 however it would require loosening the power target and pushing higher clocks which even then getting higher than 1400 mhz without insane power consumption is unlikely in a large number of GPUs. 1600 Mhz + is unrealistic.

That said even with overclocks a 480 will never reach 1070 level.

Currently a stock 480 only offers 65% of the performance of a 1070.

Considering GPU clock speed scaling results. AMD even best case 100 mhz OC gives about 8% return.

So 1266 mhz = 65% 1366 =73% 1466 = 81% 1566 = 89% 1666 = 97%.

even if clock speed scaling was linear AMDs 480 would need a clock speed near 1700 Mhz to be competitive. Thats not gonna happen.

 

[Kal-EL]

Kudos to CEO Lisa T. SU. Finally a CEO that understands that pre-maturely releasing crappy drivers at product launch kills product value.

 

[newtekie1]

Yeah...probably a little more optimistic on my behalf. ~1070 is probably more realistic. The only reason why Polaris didn't kick ass and take names was because of it's pathetic ~1300 MHz clockspeed. Reference GTX 1070 runs at 1683 MHz. A Polaris 1600 MHz clockspeed would theroretically lead to a 23% boost in performance. That's GTX 1070 territory, no?

Not really, it'd need about a 40% speed improvement to hit 1070 territory.

Its apples to oranges due to the foundries being different etc. Suffice to say AMD needed 14nm to hit its power target. While 16nm via TSMC would likely have allowed higher clock but also used more power.

This isn't entirely accurate. It might sound odd, but GloFo's 14nm FinFET that AMD used is actually less power efficient than TSMC's 16nm FinFET that nVidia used. We know this because we have a rare case where both foundries have produced the exact same processor using these processes and the 16nm FinFET came out more power efficient.

 

[OneCool]

I think RX 580 will land between 1070 and 1080. I think you're talking about RX Vega which definitely isn't coming until May/June at the earliest.

Honestly that's where I put Vega. Between 1070 & 1080. This RX580 will be below the 1070..Just my opinion though.

 

[FordGT90Concept]

Vega has almost twice as many compute units as RX 480. Vega is ~GTX 1080 at minimum. Since RX 580 really has no reasonable chance of hitting GTX 1070, there will probably be a cut down Vega to answer to 1070.

 

[Relayer]

Had AMD announced a launch date?