AMD's RX 500 Series Specifications, Performance Leaked

So you take some slides from AMD marketing as fact? for example, checkout this detailed performance test of BF1 on youtube that shows a strix 970 is beating 480 but in AMD's slides 580(slightly overclocked 480) is 38% faster than 970! who do you trust more? your own eyes or AMD's marketing.
https://www.youtube.com/watch?v=LmiKzdbSVL0

RX580 is just a RX480 with a tiny overclock on core (about 5%) and no change on memory so we can expect less than 5% difference between these two cards. from many reviews on internet you can see any decent factory OCed 970 can compete well against 480. I bet a 970 at +1500 mhz can beat 480/580. but I agree, RX580 at 200$ will be very attractive for people on budget and 1080p screens.

 

I think the Guru3D numbers are very easy to trust. Wouldn't you agree?

http://www.guru3d.com/articles_pages/battlefield_1_pc_graphics_benchmark_review,6.html
http://www.guru3d.com/articles_pages/battlefield_1_pc_graphics_benchmark_review,7.html

 

That's the stock 970. on maxwell cards there is a big difference between reference and custom AIB variants unlike polarise and pascal. you can see in the slides,for benchmarks amd is using Strix 970 not reference. but did you notice how much they are closer in Guru3D's bechmarks. 7% difference in Guru3D bechmark vs 38% in AMD's.
Now check out Strix 970 vs 480. https://www.youtube.com/watch?v=LmiKzdbSVL0

 

Hmm okay! But it's also the stock RX480. Which OCs to ~1400 Mhz on air. If you don't forget to adjust the fan profile...
So just ask the people here who have a RX480 how high they clock.
Then you can put up to 100Mhz on top of that (not directly in clock speed gain but as a performance equivalent) and you have a 580.
Just look at the LN cooled RX480 benchmarks on the well known sites and you will get a feeling for RX580.

But as many here always repeat, and rightly so:

Wait for Benches

 

Ok let's calculate overclocking potential.
RX 480: (1400-1266)/1266=10.5%
GTX970: (1500-1150)/1150=30.4%

But my whole point is it's misleading to advertise RX580 as a good upgrade for GTX970 owners. it's laughable and ridiculous to say that.

 

As I said, this is probably the most driver-bottlenecked silicon on earth. With a driver release I can see it happening. It has happened at least twice before.

 

What is this driver you're talking about?

 

1) It's 4.5 GB more full bandwidth RAM.
2) Also the per clock performance increase is different for the two archs.
+100Mhz in Polaris means something else than +100Mhz in Maxwell.

https://goo.gl/02w8K6

&

http://www.guru3d.com/articles-pages/msi-geforce-gtx-970-gaming-review,26.html

As you can see/read, per clock scaling is much higher on the RX480 when you OC it.
That's the same logic why VRAM OC was more or less useless pre-Polaris but with Polaris it suddenly meant major gains.

You really can't ignore the effect of the different arches in that comparison.

 

Ok let compare actual performance gain from overclocking based on Guru3d results on 3d mark.

http://www.guru3d.com/articles_pages/gigabyte_radeon_rx_480_g1_gaming_review,36.html
Gigabyte RX480 G1 : (4801-4358)/4358=9.4

http://www.guru3d.com/articles_pages/gigabyte_geforce_gtx_970_g1_gaming_review,26.html
Gigabyte GTX970 G1: (11427-9568)/9568=19.4

19.4% on GTX 970 vs 9.4% on RX480. more than two times actual gain from OCing.

 

did you check what 480 gains from stock to OC in Firestrike?
did you check what 970 gains from stock to OC in Timespy?

. . .

what you are doing is like comparing two different game titles, one being DX12 one being DX11 ... and comparing two different card's gains each only on one of the two engines...

 

There was not Timespy when 970 was reviewed but Timespy is more GPU intensive compare to Firestrike so it's in favor of RX480 anyway.

 

That's why I asked you for these numbers instead! lol...

Because you compare two completely ... COMPLETELY ... different APIs and engines here. One on the first card, another on the other...

As I said... wait for Benches then you will see it anyway :nerd:
It's only a matter of weeks now. AiBs already put out the product photos.

 

I agree let's wait for the reviews but nothing can change the fact that RX580 is not an option for GTX970 owners who want to upgrade. in my opinion GTX 1070 and higher can claim that.

 

An rx480 with a slightly higher clockspeed an upgrade from a 970? Anyone who makes that upgrade will be highly disappointed. That's a terribly small leap in performance.

 

I agree, its a side grade then an upgrade. But it still would have some benefits over the 970.

Hell I would have 2 RX 480 8GBs right now if I did not feel they would be bandwith starved vs a R9 Fury above 2560x1080.

 

This thread is nuts, people arguing about marketing decisions and upgrade paths...

IMO, a person does what it wants with his hard earned money.
That said, there's smart decisions and not so smart decisions.
Making the most out of your money is a smart decision, but that doesn't mean you have to stick to an upgrading strategy.

My usual upgrade strategy is to wait and look for an Hardware part that performs 2x better than what I have and costs what I can afford (usually around 200€ mark).

If you're upgrading for anything less of x1.5 more performance, without other kind of benefits (freesync or similar, etc), you're not (in my opinion) taking the most out of your investment.

 

The main one was Catalyst 15.7, when the 300 series released. AMD was pushing so hard to make them feel "new", that they didn't allow the driver to be used with older Hawaii cards during the review period.

As for the other driver, I guess it was the same for my card also, along with a ton of incremental updates. On the efficiency front, I started with something like 600k draw calls under DX11, and I'm now sitting at ~1.4 million. It removed a ton of the bottleneck that my GPU had. When I got it, my GPU was considered the equivalent of a GTX 670, now it's recommended along with the GTX 780 (a card coming out to fight Hawaii that currently resides in the 290/390(x)).

 

in dx11 yes but not indx12:banana:

http://www.guru3d.com/articles-pages/gigabyte-aorus-gtx-1080-ti-xtreme-gaming-review,1.html

 

Okay I missed this post yesterday. I need to correct your revisionist history on this one here. The raw available performance of the Tahiti was higher than GK104 from the start same with Hawaii over GK110 yet AMD could not write a DX11 driver to leverage the power (this could be due to the pipeline differences). However this should not happen with Polaris over GP104 as GP104 has more raw power than Polaris.

Also when people were suggesting the 670 over the 7970 it was when the 7970 was $550 and the 670 was $400. In fact the 7970 did not surpass the 670 until right before the 780 was released.

 

GP104 is the GTX 1070 and the GTX 1080. So no, nu-Polaris won't be faster than GP104 and I never said that would happen. It might stand between the 1060 and the 1070 performance wise, but it isn't a GT104 competitor, it's competitors are GP106 and GP107 which seem to be the worst choice already. How is the 1070 and the 1080 brought into this conversation? Also I don't believe you understand the term "revisionist history". Unless you claim that what I said didn't happen.

I still have a 7970 from that era. My choices were either that or the GTX 670. They were both at ~350 euros at the time. The GTX 780 was released on May 2013, the 7970 has been faster than the GTX 670 since at least September 2012, so yeah, it was faster than the 670 almost a year before the GTX 780 was released. The gap now is hilarious, if not terrifying. My card is almost as fast in some titles, as a card released 18 months after it.