Discussion: Polaris, AMD's 4th Gen GCN Architecture

Page 5 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
Alright, the R9 480 is launched: http://wccftech.com/amd-radeon-rx-480-polaris-10-launch/

Impressive by AMD, but performance seems identical to the rumored GTX 1060. I guess we'll have to wait and see what happens next. :)

Shoot, I missed the live stream. Oh well, I'll watch the recording tomorrow. :)
 
No, we have to wait for 480X and 490, the High score from the leaked 3dmark bench. At Same price /perf ratio, that would be around Fury perf so a worthy 1070 alternative.

At least I'm getting that, and not something that performs equally as my HD7850s in CF.
 
I think it's telling that AMD only mentions performance in the one game that relies heavily on the one feature of DX12 that AMD performs better at (Async Compute). No scaling for performance in other DX12 or any DX11 games compared to any competitors cards. Granted, they aren't going to show their weaknesses, but surely there has to be one other game where they can compete. Besides the performance in AoS and power efficiency, AMD hasn't really told us anything.

Only real benchmarks will tell us the whole story. At $199 I think it's going to perform somewhere between the 960 - 970. It's not going to hold a candle to the 1070 and depending on how badly nVidia neuters the 1060, not likely near the (yet unknown) 1060 either.

It's nice to see the process shrink, and the use of FinFETs, but from a purely performance point of view, I really think this release will be met with a lot of yawning.
 
RX480 is rumored to fall between GTX970 and GTX980 which is decent for a price point of 199$.

What remains unclear is if there is still a RX490 (with full silicon) rumored to be close to Fury non-X performance so for Nvidia between a GTX 980 and GTX 980Ti where GTX 1070 is above that (around Titan X) or if this is the RX480 but with 8Gb of GDRR5 iso 4Gb which seems unlikely unless it also runs at a higher frequency with full silicon unlocked at higher price point (guesstimated at 250$).

For 1080p users the GTX1070 is overkill while the RX480 seems underkill for next gen gaming titles if you want everything maxed out.
Is there an in between I wonder ...

 
I like what I'm reading here in the thread and in the articles, even being WTFROFLCOPTERBURGERTECH, they seem to provide reliable information this time, haha.

At $200 for the 4GB, with R9 390/970 performance levels, it's a new mainstream king until nVidia officially announces the 1060, which I believe will be very close to the rumored 480. I am putting it next to the 390, because reading the specs and being a tad positive, it will sit around there in performance; this is if they keep the ROP count similar. If they use 32 like Tonga then it won't touch neither the 390 nor 970.

Cheers!
 


Well, exactly. The leaked 3D bench DID show a more powerful Polaris 10 part, which might be the 480X or the 490. That would be a card I want. At least there seems to be space between the 2304 SPs of the new P10 and the 4000 something of the Fury X. A P10 card with 2800 SPs enabled could fill the gap and be a closer competitor for the 1070. If that part gets priced at 299 Id rather go for that then for the 480 at 199.


 
How is it around 390's/ 970's performance? 2 RX 480 gpu was faster than 1080 at 51% load in Aots. That makes me think that a single 480 should be more poweful than furyX/1080 in Aots unless both were not using same settings. Or 51% load data was sham. Lol!
 


This. There's too many ifs and one 480 will definitely not perform at Fury levels. But at 970 levels, they said so themselves.
 
What setting did they use on AoS? If gpu utilization is low then that's mean they can still increase in game setting unless the game setting already at max. Also some people take that because of that 50% utilization then a single 480 probably as fast as 1080. Ugh. If they were really that fast why not bench a single 480 vs single 1080 to begin with? Why need to use CF setup to tell the audience that your card in CF at 50% utilization or so is a bit faster than 1080? why not telling directly that a single card at 100% utilization is as fast as 1080?
 
Well, thats one of the things that is unclear. I believe we dont even know the resolution (1080 or 1440). I think there are some optimization issues which caused the GPU at 50 percent while still not scoring huge FPS numbers. Or there is indeed some foul play with CPU bottlenecking going on.

Of course a 480 isnt as fast as a 1080. They probably took a few weeks to configure a game and setup so that it seems its ridiculously fast just for creating the hype going on right now. And there is common sense of course. If the 480 were really that fast they wouldnt sell it for 199.

Conclusion: they are just creating fumes with this sheet, and we basically know nothing more than a little confirmation of the 3dmark leaks.
 


For some obscure bizarre reason, sometimes AMD cards scale over 100% in certain games. Maybe Ashes is such a game. The bottleneck for the performance in a single card situation was lessen or lifted while in XFire. This is obviously conjecture, but there is evidence of this happening.

In any case, the speculation for it being around the 970 is from AMD itself (no data provided, though). I just want to think it will be next to the 390 as well, since it would effectively move/shift the segments for this new generation of cards.

We need to keep a close eye on the information and speculation mill, since it's gonna get wild. READY THE SALT!

Cheers! 😛
 
Of course they would sell it at that cheap, their mission is to take back market share. The R9 390x a while back hit 100% performance increase in Ashes DX12. Their whole point was to show that gaming doesn't have to be expensive to play games at settings which are similar to the high end cards are capable of.

It's obviously skewed, but based on what we saw, NVIDIA aren't even close to AMD in terms of DX12 performance. AMD are so ahead of the game with more cores in the CPU's etc, but there aren't enough content to support what they make.

I see a very bright future for AMD now that DX12 is slowly getting more attention, especially from the new Microsoft plan to port games to Windows 10.

And, where I live NVIDIA equivalents are 20% more expensive than AMD, so I can't even be bothered, I fanboy my wallet.
 
Well you dont get me convinced that AMD, because of their mission, suddenly lost any desire for making money and profit. It is nice that seem to have a good offering now, but they had to make amends after the R9 Rebrand Series. Cards like the R9 390 were way too expensive when they were introduced - and after. They let Nvidia get market share with the 970 so I dont feel very sorry for them now.
 


That's not true at all. Look at the launch price on Pcpartpicker and you'll see that 390's were sold cheaper than 970's, by about $20 on average. Due to the 390's NVIDIA lowered their prices, not the other way around.

That may be different to you but it was true where I live as well.

NVIDIA got more market share after half the people complained and always mentioned the R9 390's power consumption. It's funny that when someone posted help about the R9 390 power supply, most of the time were told to upgrade. But a GTX 980 Ti was apparently fine on a same wattage/same quality, a card which draws even more power. So much confusion between performance/watt and actual draw...

 
In regards to why they *can* sell the Polaris siblings cheaper have to do *a lot* with die size and market position. AMD needs 2 things badly: brand recognition back (that usually translate to mid-term higher sells) and actually SELL stuff. On this last point, you can have the best performing card, asking for 10x times the price of your nearest competitor and make it the most profitable ever product in your lineup, BUT... If you sell 0 cards, then that means squat. Remember profit margins go as far as you can actually sell the products. That is the sole reason they want to do a repeat of the HD4000 series. That generation brought the 2 points I am mentioning and AMD saw a mid term benefit from that.

Cheers!
 


Are two 480 in CFX would be a better choice than GTX 1070 or GTX 1080?
 


AMD is not doing anything other than marketing. They sell their products at a price that people are willing to pay.

People seem to forget that at launch the Fury X was the same price as a 980Ti and at launch a FX-62 was just as expensive as Intels top end.

If it performs at a certain price point, AMD will price it accordingly. They will not give you whatever nVidias top end is for less if they match or beat it. They are a company looking to profit not make friends with their consumers.

I personally ignore AotS due to it being only one game that uses that particular feature and there is more to DX12 than ASYNC Compute.



http://www.tomshardware.com/reviews/amd-radeon-r9-390x-r9-380-r7-370,4178-9.html

That is because a 980Ti reference uses the same power as a reference R9 290 while a 390X uses more power.

If you look at the system requirements the recommended is a bit higher on a 390X.
 
My point was that people overexaggerate. The R9 390 draw less than a GTX 980 Ti, I didn't mention R9 390x. Also, majority of benchmarks that are known, measure at the wall, which is inaccurate unless you need to calculate your power bill....

Peak power consumption while gaming is still below ~350W on either the R9 390 and the GTX 980 Ti, so why are people so obsessed about recommending a PSU that's far above that?

Only benefit if the PSU can't handle the load at above 50%, or it's to reduce the fan noise. But no logical arguments can be made about the R9 390's power consumption, which is why it's frustrating.
 
Here are your price-performance targets.

If you go by the Ashes of the Singularity benchmark shown at the AMD marketing presentation, then anything in the 50% range will be the equal of the RX 480 (x2 for near 100% Crossfire scaling). Since AotS is extremely AMD biased, then look in the 40% range for a more accurate view of overall performance averaged over several games (again assuming near 100% Crossfire scaling).

Currently the cards on Newegg that occupy the $200+ price range are the GTX 960 and R9-380, 380X, and 280. The best card of the bunch, the 380X can be purchased for $220, just under the price of a RX480 (8GB). On this chart, the 380X comes in right at the level of the 280X.

Conclusion: The RX480 will offer a 5-10% performance increase over the R9-380X for about the same price (or a little less for the 4GB).

perfrel_1920_1080.png

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1080/26.html
 
And if Ashes is not that biased, then the 50% range would be more fit. To that end, both the 970 and the 390 (non-X) are $300+ on average.

Since the truth is usually in between, then it is an improvement (by your own calculations) of roughly 15% for the same price range.

It is an interesting exercise, 17seconds.

Cheers!
 
390/970 performance for $200. yes please. i'll take that and not even second guess it later.

nvidia so far has priced each card a bit above the one it is replacing. that would put a 1060 at at least $250 or a lot more for the founder's edition. course they will have to drop that some if the 480 compares for less. win win for us :)
 

You would be basing that on the assumption that Ashes of the Singularity is a normal non-biased benchmark indicative of the majority of games. Thing is, the NDA lifts on June 29th, nearly a month away. Plenty of time to beat this into the ground before actual performance numbers drop.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom