[Guru3d] Radeon RX 470 Benchmarks

[Ma_Deuce]

Going by the Firestrike database it scores similar to 280X 1150MHz with same 5960X. Am I off in not being that impressed in terms of a new node? Unless 470 is extremely cut down to 24 CU (1536 cores). Then it becomes technically impressive but also would be the end of the 2nd tier extra bang for the buck AMD has been providing.

It looks like it will be very good "bang for the buck" by the benchmarks. Im not quite sure what the number of CU has to do with that though... Either the card performs great for the $$$ or it doesn't. I'm looking forward to seeing some real gaming benches.

 

[Thala]

I want AMD to at least match the relative perf/area gains going to 14nm that Nvidia made going to 16. This is my whole point being disappointed. Nvidia was already ahead on 28nm. It appears to me that they may have gotten more ahead.

But if you decide to go wider instead of increasing frequency you cannot possibly match perf/area gains. However you have good chances of being ahead in perf/W assuming iso performance.

This has nothing to do with being ahead! That's my whole point. Personally i would always go wider instead of increasing frequency given a) application has sufficient parallelism b) die size is not killing my business opportunities.

 

[ocre]

The analysis posted is absurd. The entire reason the performance per watt metric exists is so to avoid the multitude of conclusions that can be reached when looking at stock vs oc vs temp vs die size across different architectures.

AMD is ahead of nVidia in perf / watt. Full stop. Pascal is showing 60-70% better perf / watt vs maxwell, which isn't bad. This would suggest they can create a 100 w actual tdp gtx 970 but what we're seeing with p10 is 100w gtx980+ performance and 70-80w gtx 970 perf in rx 470. P11 is likely to kick ass in notebooks as a consequence.

Shareholders should be happy. It remains to be seen just how good rx480 performs for enthusiasts when wattage is dialed up.

AMD has been slapped in mobile and mass market. While we're sitting around wondering whether rx480 is 390x +/-20% the real intention of p10/11 - efficiency gains - has already been a massive success. nVidia is for once categorically behind amd in efficiency.

So....
It's really really interesting that these kinds of things are being said:
Nvidia is categorically behind? AMD is ahead in perf per watt? Spoken in such absolutes!! Sure, I guess those kind of comments aren't getting criticized.

Anyway, we can't speak in such absolutes about polaris vs pascal when it comes to per per watt and then have an issue when others project....especially when the latter clearly states it to be projections based on limited info.

But, just because we happen to be posting in a thread that has the 470 in a chart with pascal....and the fact that amd provided some numbers, which are in the article in the op. You know, it was some numbers amd gave which makes you assume that Polaris will have some great perf per watt advantage in the first place. But, let's just follow this out here....
We have some percentages by amd which makes it easy to assume that Polaris will have a great effiency lead. And, at the same time, we also have some numbers from amd that this article in the op talks about. One is 110watt, which they compare to their older architecture. The other, some performance figures.

So, we have this chart and we have the 470 score. We have this 110 watt that amd provided and is used in their efficiency figures.

-we also have the 1080 in the chart, and we know what it's tdp is. We also can find pretty exact figures that have been tested....

So, based in both scores, using 110watt provide by amd and the known 1080 power consumption-----it's kind of not adding up to the 470 having a major perf per watt advantage. Not here, not even. It's actually, going by the info here, no advantage for Polaris.

Now mind you, this is going on limited info. But, that is all we have right now. And speaking specifically in reference to the data in the op...which the thread is about. Perhaps we shouldn't be talking in such absolutes when it comes to Polaris effiency advantage vs pascal.

At least, the numbers in the op benchmark doesn't show such things...that is, if we use the tdp amd provided for the performance

 

[JDG1980]

AMD marketing for Polaris has always emphasized perf/watt. Therefore, there is no reason to think they're going for the max feasible clocks. It's likely that they are instead picking the sweet spot, like they did with 1st generation GCN.

Back in 2012 when 28nm was still a new and immature process, the original 7870 (Pitcairn, similar die size to P10) could do over 1200 MHz on the reference card. Stock clock was only 1000 MHz. And the 7850 came clocked at only 860 MHz - it overclocked by about 33%.

 

[Vesku]

It looks like it will be very good "bang for the buck" by the benchmarks. Im not quite sure what the number of CU has to do with that though... Either the card performs great for the $$$ or it doesn't. I'm looking forward to seeing some real gaming benches.

280X was $180-200 so $150 for similar on a node change doesn't sound that great to me. On one hand it's good if 470 is a savagely cut 480 because that means 480 will be very nice at $199 on the other hand end of an era compared to 7950/280, 290, 390 vs next step up.

 

[Abwx]

At least, the numbers in the op benchmark doesn't show such things...that is, if we use the tdp amd provided for the performance

They used actual boards powers, so what are they..?..

 

[happy medium]

SO a good guess by reading this article...........
150$ 470 ~= 290 (200$ in 2015) with free games, clocked at 390 levels
229$ 480 ~= 290x (279$ in 2015) with free games, clocked at 390x levels

lower power and lower price, same 2013 performance.

WHats so great about these cards for the millions of people that bought 290's, 290x's ,gtx780's, 780ti's, gtx970's and gtx980's?

Seems good for the people who bought a 280x or 270x, not a bad upgrade.
Seems more like console graphics will be like in the new Xbox and Ps4 next year.

 

[Ma_Deuce]

280X was $180-200 so $150 for similar on a node change doesn't sound that great to me. On one hand it's good if 470 is a savagely cut 480 because that means 480 will be very nice at $199 on the other hand end of an era compared to 7950/280, 290, 390 vs next step up.

380X? was about $230 when it launched, correct? We are looking at 65% of the price and 7% more performance in fire strike. Probably much better perf/watt as well. I still think that's pretty solid.

 

[happy medium]

380X? was about $230 when it launched, correct? We are looking at 65% of the price and 7% more performance in fire strike. Probably much better perf/watt as well. I still think that's pretty solid.

yea but a single 6 pin powered 180$ 280x is ~ = to a single 6 pin powered 220$ 380x. vs a 150$ single 6 pin 470.
I'm guessing the 470 will be faster than a 280/380x?

 

[casiofx]

SO a good guess by reading this article...........
150$ 470 ~= 290 (200$ in 2015) with free games, clocked at 390 levels
229$ 480 ~= 290x (279$ in 2015) with free games, clocked at 390x levels

lower power and lower price, same 2013 performance.

WHats so great about these cards for the millions of people that bought 290's, 290x's ,gtx780's, 780ti's, gtx970's and gtx980's?

Seems good for the people who bought a 280x or 270x, not a bad upgrade.
Seems more like console graphics will be like in the new Xbox and Ps4 next year.

I think you had forgotten the original goal of Polaris: to increase market share.

What's the market share held by >$300 cards?
These <$229 cards are for the mass market, do those mass market people use GTX980s, 290X etc? No, they are probably still using some HD7770, GTX 450, 560 Ti, HD6850 etc. Which to them is a freaking big upgrade.

 

[happy medium]

I think you had forgotten the original goal of Polaris, to increase market share.

Mabe AMD knows something we don't know? You saying there will be no competition from Nvidia's 1060/1050?

You don't think Nvidia will show benchmarks of the lower end 1060/1050 in about 2 weeks?

No, they are probably still using some HD7770, GTX 450, 560 Ti, HD6850 etc. Which to them is a freaking big upgrade

SO last years 185$ 380/gtx960 was a big upgrade for them also.

 

[Vesku]

380X? was about $230 when it launched, correct? We are looking at 65% of the price and 7% more performance in fire strike. Probably much better perf/watt as well. I still think that's pretty solid.

280X 1150MHz + 5960X gets same Firestrike score.

 

[casiofx]

Mabe AMD knows something we don't know? You saying there will be no competition from Nvidia's 1060/1050?

You don't think Nvidia will show benchmarks of the lower end 1060/1050 in about 2 weeks?

Talk about it when it arrives, for now even 1070 & 1080 have supply issues.

Ironically, I found it funny when you criticized the technical and performance aspect of 470 & 480 when I saw your system specs. Seriously GTX960? That's like the worst maxwell ever in terms of price performance.

 

[darkswordsman17]

SO a good guess by reading this article...........
150$ 470 ~= 290 (200$ in 2015) with free games, clocked at 390 levels
229$ 480 ~= 290x (279$ in 2015) with free games, clocked at 390x levels

lower power and lower price, same 2013 performance.

WHats so great about these cards for the millions of people that bought 290's, 290x's ,gtx780's, 780ti's, gtx970's and gtx980's?

Seems good for the people who bought a 280x or 270x, not a bad upgrade.
Seems more like console graphics will be like in the new Xbox and Ps4 next year.

Seriously? Does this really need to be explained yet again? No, because you know exactly the answer to your question as you answered it yourself.

 

[happy medium]

Talk about it when it arrives, for now even 1070 & 1080 have supply issues.

Funny you should say that...........haven't seen 470 or 480's on the shelf. Talk to me about them then.

 

[Ma_Deuce]

280X 1150MHz + 5960X gets same Firestrike score.

OK, so what is the score of an OC 470?

 

[happy medium]

Seriously? Does this really need to be explained yet again? No, because you know exactly the answer to your question as you answered it yourself.

Yea, explain to me how paying 230$ for next year console graphics are exciting in a PC enthusiast forum?

@ casiofx, my 170$ overclocked 960 4gb is a good buy for a place holder..

 

[Ma_Deuce]

WHats so great about these cards for the millions of people that bought 290's, 290x's ,gtx780's, 780ti's, gtx970's and gtx980's?

This is just a guess, but I bet that the 490 and 490x will be marketed more toward that crowd...

 

[casiofx]

Funny you should say that...........haven't seen 470 or 480's on the shelf. Talk to me about them then.

Yes lets talk about them, i wonder which one will show up on the shelf first? Custom 1070 and 1080 at msrp or rx460,470 and 480.

 

[Vesku]

OK, so what is the score of an OC 470?

I hope it's a guaranteed 20+% otherwise the 2nd tier card as sleeping giant AMD tradition is dead.

 

[darkswordsman17]

Yea, explain to me how paying 230$ for next year console graphics are exciting in a PC enthusiast forum?

@ casiofx, my 170$ overclocked 960 4gb is a good buy for a place holder..

Um, first that's not even what you asked. You first asked:

WHats so great about these cards for the millions of people that bought 290's, 290x's ,gtx780's, 780ti's, gtx970's and gtx980's?

Which has been repeatedly explained why that is just plain stupid to even be asking. But you obviously know this because you then follow it up by saying:

Seems good for the people who bought a 280x or 270x, not a bad upgrade.

Literally everyone but a select few that happen to be constantly parroting a few of the same nonsensical arguments, has understood that this is exactly meant to replace those lower cards and is not meant for the people that own those cards you specifically mentioned (unless you greatly value the few features it has over them, namely lower power consumption and updated video features like newer HDMI/DP support). This is not an enthusiast card. Complaining that it matches previous enthusiast level cards is...well at this point you can't even pretend to be this dense or confused, it should be considered straight up trolling.

Yeah, why on earth would people be excited about having a ~$200 video card capable of being put into cheap and/or small computers that will offer similar performance of likely to be $399+ consoles, several months before those come out? Gee, this is really hard to figure out.

 

[iiiankiii]

SO a good guess by reading this article...........
150$ 470 ~= 290 (200$ in 2015) with free games, clocked at 390 levels
229$ 480 ~= 290x (279$ in 2015) with free games, clocked at 390x levels

lower power and lower price, same 2013 performance.

WHats so great about these cards for the millions of people that bought 290's, 290x's ,gtx780's, 780ti's, gtx970's and gtx980's?

Seems good for the people who bought a 280x or 270x, not a bad upgrade.
Seems more like console graphics will be like in the new Xbox and Ps4 next year.

But yet, you bought a freaking GTX960 (a card that is TWO TIMES SLOWER than a $200 R9 290) for $170 in 2015. LOL. Comedic Gold.

 

[beginner99]

It appears that AMD has not matched the die size gains from the new node. P10 vs Hawaii and GP104 vs GM200 demonstrate this clearly as the ratios are similar, actually likely slightly more favorable to AMD yet they gain far less. The 25% cut 1070 still draws even (slightly beats both stock) with the fully enabled GM200. The only 11% (guessing….) cut 470 seems to draw even with the 9% cut 290? Ouch. If AMD matched Nvidia’s gains, a 11% cut P10 should be easily faster than fully enabled Hawaii.

Problem is you can't compare it that why because one is a big die and one a small die and certain parts of the chips have a fixed sized especially video encode/decode functionality. So that part will always take a increasingly larger part the smaller the chip gets.

Second point you don't take into account is that 980Ti reference (eg. GM200) was rather low clocked and had a huge OC headroom where the 1070 (and 1080) have a much smaller headroom. AMD almost certainly did the exact opposite. Hawaii is basically factory OCed and running at frequency way above optimal performance/watt. That's why 290 reference got a terrible reputation. AMD certainly fixed this so we will now get a performance/watt optimized reference part and probably much faster factory OCed AIB card probably with 8-pin connectors and much worse performance/watt but it is these cards that need to be used for your comparison to remain fair. Or you need to use custom 980 Ti models which are easily 20% faster than reference and faster than the 1070 and use more power and reference completely invalidating your idea.

 

[casiofx]

But yet, you bought a freaking GTX960 (a card that is TWO TIMES SLOWER than a $200 R9 290) for $170 in 2015. LOL. Comedic Gold.

It would mean RX470 is around two times faster than GTX960 right? He should be happy instead (/sarcasm)

 

[AtenRa]

These cards are made for DX-12 and Vulkan, comparing them against last gen cards only in 3D Mark FireStrike is stupid.