AMD Polaris Thread: Radeon RX 480, RX 470 & RX 460 launching June 29th

[Midwayman]

This showed up briefly on newegg today.

 

[NomanA]

No it didn't. From last page:



Also, looks like WCCFTech benchmarks were fake:

But you posted those 1080 benchmarks just a page ago, whether they are fake or not, where 390x was behind those later 480 scores. Apparently they were good enough to post then, but not now.

The ridiculous thing about this is that you are posting reference benchmarks and updating the references actively, finding those that fit the narrative (480 is slower than 390x), over a difference of less than 1%.

Anyway, this is my last comment about it. Let's get back to topic.

 

[antihelten]

They dont overclock manually. But they can push a botton that says: "i want my cards 20% faster".

They can, but they won't, never underestimate how lazy your average consumer is.

Besides you don't actually expect that there will be a button like that do you?

The ridiculous thing about this is that you are posting reference benchmarks and updating the references actively, finding those that fit the narrative (480 is slower than 390x), over a difference of less than 1%.

Using a 390X that is coincidently overclocked by 1%.

 

[zinfamous]

I don't get what these fire strike and 3dmark scores provide. If you are buying a gpu to mine exclusively I see that. Or maybe you buy one just to stress and bench that's all then the scores make sense.

I buy mine to play games so real world benches make more sense. Still gotta wait until next week for any of that.

I gifted away my steam library and replaced all content with benchmark tools.

That's all that matters to a true PC gamer anyway. Are you a true PC gamer like myself? Running benchmarks is the best!

 

[Sweepr]

Using a 390X that is coincidently overclocked by 1%.

5 MHz gap between Sapphire / PowerColor and only 1 point difference in FSU GPU score, so it might well be above 2907 pts at 1050 MHz. I agree that's nitpicking at this point (margin of error %).

Still, I wonder where those WCCFTech results come from. If not 100% fake, maybe overclock results?

 

[krumme]

Price

wrong.
The price is a known entity and have been for a while. I didnt recall Gibbo saying the would move half the 1080 on day one one month ago when price was published.
"Hey great i will move 500 pcx day one"
Doa
It obviously tell us something about performance. And that he have 500 pcx

 

[antihelten]

5 MHz gap between Sapphire / PowerColor and only 1 point difference in FSU GPU score, so it might well be above 2907 pts at 1050 MHz. I agree that's nitpicking at this point (margin of error %).

So the Sapphire comes with a 0.48% overclock and is 0.72% faster and the PowerColor card comes with a 0.95% overclock and is 0.76% faster. Average of 0.72% overclock and 0.74% faster.

These scores are as close to identical as you can get.

Still, I wonder where those WCCFTech results come from. If not 100% fake, maybe overclock results?

Well, it is WCCFTech we're talking about here. I wonder how their "NDA ending on the 24th" prediction will pan out

wrong.
The price is a known entity and have been for a while. I didnt recall Gibbo saying the would move half the 1080 on day one one month ago when price was published.
"Hey great i will move 500 pcx day one"
Doa
It obviously tell us something about performance. And that he have 500 pcx

This doesn't really make any sense.

The 480 is priced at $200-230 whilst the 1080 was priced at $700. You really don't need anymore of an explanation than that for why the 480 will sell as much as Gibbo claimed, even it is only around 60% the performance of the 1080.

 

[krumme]

They can, but they won't, never underestimate how lazy your average consumer is.

Besides you don't actually expect that there will be a button like that do you?



Using a 390X that is coincidently overclocked by 1%.

I expect the button to be only "yes" and "no"
Then it doesnt matter if you are lazy

 

[Kenmitch]

If Gibbo expects to move 500 cards day one there is a reason for it and its not color or tdp. Its performance for £

That's a far stretch. Not many cards can do 4k even right now smoothly. But we can always dream and hope.

I figured it would automatically be seen as sarcasm.

Sorry for the confusion.

Not really why people even bother to hunt down and show results for 4k synthetic benchmarks.

I'd imagine even reviewers would skip the resolution unless testing out crossfire scaling and performance.

Should be a hot seller I'd imagine.

 

[antihelten]

I expect the button to be only "yes" and "no"
Then it doesnt matter if you are lazy

And you genuinely think that the existence of this secret button is why Gibbo is expecting to sell as many 480 cards as he is, and not simply because of the fact that the 480 is a decent performing card (60% of a 1080) at a very low price that sits right in the sweetspot ($200-230).

 

[crisium]

I only look at synthetic benchmarks for pre-release stuff like this. I immediately click next if I see them in a review, and will do so on the 29th as well.

TPU, Sweclockers, and Computerbase averages of performance are much better indicators of overall performance since they are made up from numerous actual games.

 

[dogen1]

They solve nothing because we know nothing about those feature. Also if a game throws visible pixel sized polygons at the screen no culling accelerator would help, especially given the fact that even DX11 allows to cull patches straight in the shader already.

Well, I can't find it, but I did see a benchmark showing mvidia hardware culling many more triangles in a certain test. Maybe the "discard accelerator" is a leveling of the play field in this sense.

Actually, just found it here.

http://www.hardware.fr/articles/928-4/performances-theoriques-geometrie.html

 

[dark zero]

Crap!!! Rotated core?!!
There goes my hope of using my universal watercooling blocks.

Don't worry. That is only for reference cards

 

[thilanliyan]

Don't worry. That is only for reference cards

Do aftermarket cards ever use a non-reference core layout?
AFAIK aibs take the GPU package and put it on their cards, which means the GPU orientation would be the same...?

 

[MrTeal]

Do aftermarket cards ever use a non-reference core layout?
AFAIK aibs take the GPU package and put it on their cards, which means the GPU orientation would be the same...?

Well, they could rotate the package but I don't imagine many will do that. What universal block do you have though? Most should be large enough. Worst case, buy one of the 7970 copper shims from Aliexpress for $2 now and it'll be in by the time you're ready to use it.

 

[krumme]

And you genuinely think that the existence of this secret button is why Gibbo is expecting to sell as many 480 cards as he is, and not simply because of the fact that the 480 is a decent performing card (60% of a 1080) at a very low price that sits right in the sweetspot ($200-230).

Lol. No but i expect it to be so easy to oc even the lazy will do it and more or less all aib as this is desktop market and power usage is plenty low already.
Obviously the card is in the sweetspot for price and that will move it but not if the package is bad including lackluster performance. Its not a brand like nv.
Who knows perhaps its the overall package that is just spot on and there is plenty to sell? Or perhaps Gibbo could sell as many 1080 if he had them

 

[Stuka87]

Do aftermarket cards ever use a non-reference core layout?
AFAIK aibs take the GPU package and put it on their cards, which means the GPU orientation would be the same...?

Yes, GPU orientation would be identical to what is shown on the reference card. The AIB could technically rotate the whole package 45 degrees so that the chip itself is parallel to the board. But this would make running traces for memory a bit of a pain.

 

[Shivansps]

They dont overclock manually. But they can push a botton that says: "i want my cards 20% faster".
I am just saying there is a reason Gibbo expects to move half the 1080 ammount on day one. What else could it be?

Price, even if it was GTX970 level, at $229 for 8GB model with less power requeriments, is still a great value for the money, and this is something that is more close to the 980 than 970... thats why.

 

[Bacon1]

Tons of pictures of the 480 here:

http://imgur.com/a/Kpdx2

 

[Thala]

Nice. Looks like the confidence that we will see R9 390x level of performance instead of only R9 390 level of performance increases. (at stock clocks frequencies)

That means we don't only have perf/price winner here but possibly even perf/power in case of the 100W under gaming loads turn our to be true!

The ridiculous thing about this is that you are posting reference benchmarks and updating the references actively, finding those that fit the narrative (480 is slower than 390x), over a difference of less than 1%.

Guess Sweepr has just too much time at his disposal to query the whole net for the fastest R9 390X benchmarks available

 

[sirmo]

Tons of pictures of the 480 here:

http://imgur.com/a/Kpdx2

I've been checking out that PCB and it looks boss, very high quality as far as PCBs go.

 

[sze5003]

I've been checking out that PCB and it looks boss, very high quality as far as PCBs go.

Agreed, it looks sharp. I really wish we had some AIB designs to look at. I'm sure there will be a sapphire tri version, maybe MSI lightning.

 

[Sweepr]

Guess Sweepr has just too much time at his disposal to query the whole net for the fastest R9 390X benchmarks available

Actually it took only a minute, simply had to look for Geforce GTX 1080 reviews, plenty where that came from. You might want to read forum rules regarding user callouts.

 

[Thala]

Simply had to look for Geforce GTX 1080 reviews, plenty where that came from. You might want to read forum rules regarding user callouts.

Not sure what agenda you have. But searching the net for cases where an overclocked R390X is a hair faster than a stock RX 480 is ridiculous.
Apparently the scores you posted a day before were not good enough to fit your narrative.

 

[Sweepr]

Not sure what agenda you have. But searching the net for cases where an overclocked R390X is a hair faster than a stock RX 480 is ridiculous.

A simple Google search would save you some embarassment. 1055 MHz for the Tri-X, so only 5 MHz above AMD's 'stock' 1050 MHz. Not my fault that these specific Radeon RX 480 results can't match certain VGAs with updated drivers. Now please keep your personal attacks and OT out of my thread.