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

[jimmysmitty]

It seems to me as though he was inferring that only one game was used to get that result, hence why I asked for a link. Now if when the card is released and the results show the 1060 to be faster across the board we will know he was talking out of his arse.

No need to get so defensive on an opinion. You really love nVidia, don't you? 😛

In any case, AMD used Ashes and they were right for that game. I would imagine when nVidia releases more slides we'll know how the cards will stack, since the 480 has been reviewed very thoroughly by now. IMO, there is merit to his theory.

Cheers!

AMD has a history of cherry picking benchmarks to prove their performance. Remember the FX 8150 launch slides from AMD? Where they compared only things they won in against Intel CPUs they beat for those parts?

That is why marketing slides are useless. Always wait for official benchmarks from Toms/Anand etc.

I see an custom boards coming with 175W TDP:

https://www.overclockers.co.uk/sapphire-radeon-rx-480-nitro-oc-8192mb-gddr5-pci-express-graphics-card-gx-37b-sp.html

It will be interesting to see how the GTX1060 compares. Hopefully we will have reviews on the paper launch day.

I note that nVidia proponents are already pointing out "their" card is more energy eficient (120W to 150W for the AMD, or even 175W for the above card). Although the proportional difference looks bad for AMD, assuming the leaked numbers for the 1060 hold true, it's not as relevant at this node compared to last generation. The absolute power differential between a R9 390 and a GTX970 was something like 130 Watts, which is a lot of heat to get rid of. But at 14/16nm the absolute differential is only 30 Watts, which is not much of a cooling or noise concern.

Well so far TSMCs 16nm is showing to be quite a bit more power efficient than GloFlos 14nm. A much smalle chip from nVidia could use quite a bit less power if done right.

 

[salgado18]

I see an custom boards coming with 175W TDP:

https://www.overclockers.co.uk/sapphire-radeon-rx-480-nitro-oc-8192mb-gddr5-pci-express-graphics-card-gx-37b-sp.html

It has an 8-pin connector, I wonder how is the power distribution.

By the way, are there non-reference boards with blower coolers? AMD's new cooler seems good, but I don't think it's good enough.

 

[Turb0Yoda]

I see an custom boards coming with 175W TDP:

https://www.overclockers.co.uk/sapphire-radeon-rx-480-nitro-oc-8192mb-gddr5-pci-express-graphics-card-gx-37b-sp.html

It has an 8-pin connector, I wonder how is the power distribution.

By the way, are there non-reference boards with blower coolers? AMD's new cooler seems good, but I don't think it's good enough.

Not at the moment that I personally know of. There should be soon though.

 

[TehPenguin]

https://www.youtube.com/watch?v=E3OVwMIpqsQ


Found a VR review for the 480 with some benchmarks at the end vs 970&980 at stock clocks. So no stock 480 vs superclocked 970.

Not super detailed but it shows that the 480 is actually slightly faster than the 970 when both at stock speeds.

 

[jimmysmitty]

https://www.youtube.com/watch?v=E3OVwMIpqsQ


Found a VR review for the 480 with some benchmarks at the end vs 970&980 at stock clocks. So no stock 480 vs superclocked 970.

Not super detailed but it shows that the 480 is actually slightly faster than the 970 when both at stock speeds.

Thats not something anyone is questioning. The only thing I see is that the majority of currently sold and owned 970s are overclocked well beyond the stock 970s.

I watched a video recently of a guy who modded a GTX 970 to use a AiO water cooler. was great for cooling but the max clock he could get was 1350MHz on the core and 2200MHz on the VRAM before it became unstable in the games.

It makes me wonder how high the AiBs will be able to get it clocked if someone with an after market water cooling setup didn't get too high. Maybe they will with superior VRMs and cooling for those.

 

[Martell1977]

It will be very interesting to see what AiB's can do with better cooling. ASUS had a advertisment on their website for the upcoming 480 @ 1328mhz core. Which is interesting as that is the highest OC I've seen online that people have been getting before the card becomes unstable. Hopefully the cards will still have some headroom for people to OC. With the patch to fix the power draw and everyone expecting a performance drop from it (I'm not so sure, if they can just get the card to draw from the 6-pin instead), more room to OC should go a long way, either way.

 

[Robert Cook]

Just saying Mousemonkey, apparently that bar graph was a reddit satire of what Nvidia is planning to say. at least that is what I have heard.

I just posted what I saw on the foreign site linked to by the other guy, if you have a problem with it then take it up with him.

I have no issues with you posting it, I simply figured I would point it out as it seems even videocardz fell for it at one point. I believed it to be real at first too. sorry for coming across as condescending in my last post.

 

[17seconds]

Just saying Mousemonkey, apparently that bar graph was a reddit satire of what Nvidia is planning to say. at least that is what I have heard.

I just posted what I saw on the foreign site linked to by the other guy, if you have a problem with it then take it up with him.

I have no issues with you posting it, I simply figured I would point it out as it seems even videocardz fell for it at one point. I believed it to be real at first too. sorry for coming across as condescending in my last post.

Since that graph is being plastered all over the web as being Nvidia originated, can you provide some evidence that it's a reddit prank?

 

[Suzuki@1473780084@1473780163]

What graph?

 

[BurgerandChips66]

Very carefully , I was born blind in my left eye , and loss vision in my right due to my type 2 Diabetes. I use 2 different pairs of glasses , but they don't help very much .Sometimes it takes me a longer time to respond because of the vision issues.

I have no issue with your bold type. It helps me read the posts more clearly and I need all the help I can get!

I have astigmatisms in both eyes and I'm short sighted as a result (-6 and -5.5) as well as having had detached retinas in both eyes. After a couple of operations I can now see 'ok' but it's not great tbh. Still, things could always be worse...

 

[17seconds]

What graph?

These ones. They show a +15% performance advantage for the GTX 1060. So if it's a prank it's a pretty weak one, as those numbers are probably very likely.

So to be so self-assured while giving a moderator a hard time about it really needs some evidence. These are being promoted everywhere as originating with Nvidia.

Nvidia GeForce GTX 1060 presentation slides Leak

 

[turkey3_scratch]

But does it really matter if it is 15% faster? It will probably cost at least 25% more, possibly giving it worse price/performance. But we will see. It is also hard to take some graph online very seriously, especially a graph made by Nvidia. Just like how some graph made by AMD cannot be trusted too well.

Though in this case, I believe that graph will turn out to be pretty accurate. But in the end, it doesn't even matter since cards are priced according to their performance.

 

[TJ Hooker]

These ones. They show a +15% performance advantage for the GTX 1060. So if it's a prank it's a pretty weak one, as those numbers are probably very likely.

No idea if it's supposed to be a joke, but if it is I imagine the joke would be based on the potentially misleading horizontal axis, not the 15% performance claim. I.e. representing a 15% performance increase (which isn't anything crazy) in such a way that, visually, it looks like it's 75% better and stating the 1060 is "MUCH FASTER". I could see that as mocking the way Nvidia (and AMD) massage the data to make their products look more impressive.
*Shrug* But I don't know.

Edit: Ah, it looks like there is a satirical graph from reddit that's similar the the graph we've been talking about:
https://i.redd.it/6vl9niydfn6x.png

 

[BurgerandChips66]

Perhaps they missed out the 'Not' from in front of 'Much Faster' by mistake...

 

[Rogue Leader]

I am sure there is a good chance the 1060 will be faster than the RX480, however that picture is very bogus.

 

[Suzuki@1473780084@1473780163]

Everyone manipulates graphs in their favor to intentionally mislead, it's a grey line.

If that graph is from NVIDIA, then people shouldn't say it's not or that it is, because they can remove it and start denying it. People are acting as like they're about to go to court for this.

Is it really that difficult to make a personal prediction, wait until the card gets released, and compare your prediction to the results, then discuss?

 

[BurgerandChips66]

AIB Partner RX 480 News - ASUS Strix & Sapphire Nitro

https://youtu.be/KKibEXAL_3M

 

[Robert Cook]

These ones. They show a +15% performance advantage for the GTX 1060. So if it's a prank it's a pretty weak one, as those numbers are probably very likely.

No idea if it's supposed to be a joke, but if it is I imagine the joke would be based on the potentially misleading horizontal axis, not the 15% performance claim. I.e. representing a 15% performance increase (which isn't anything crazy) in such a way that, visually, it looks like it's 75% better and stating the 1060 is "MUCH FASTER". I could see that as mocking the way Nvidia (and AMD) massage the data to make their products look more impressive.
*Shrug* But I don't know.

Edit: Ah, it looks like there is a satirical graph from reddit that's similar the the graph we've been talking about:
https://i.redd.it/6vl9niydfn6x.png

yeah I guess Reddit just took the scale and played with it more, my mistake, I guess that really is the real nvidia graph.

 

[TehPenguin]

A different question: how is it with AMD cards? Is the reference the cheapest model or have there been aftermarket cards(with better cooling) priced the same?

 

[Suzuki@1473780084@1473780163]

Aftermarket are more expensive

 

[turkey3_scratch]

The 300 Series didn't even have a reference card.

 

[Suzuki@1473780084@1473780163]

Any proof that pcie was the problem yet? I doubt that personally, because I havent found a source that is accurate to electricity, like reading comedy.

 

[renz496]

@turkey3_scratch

When everyone offering their own cooling solution with 390/390X XFX made unexpected move by selling some of their 390/390X using 290s reference cooler.

Now we know how RX480 (reference) fare i want to know about RX470 and 460. Where did you guys think 470 and 460 sit compared to current cards?

 

[BurgerandChips66]

I'd expect the RX 470 will perform slap-bang in the middle of the an R9 380 and R9 390. So like an overclocked R9 380X.

Speaking of overclocking, for the reference card there might be a bit more headroom than with the RX 480. So custom RX 470's might overclock rather well and nearly match an R9 390 (though that's the best case scenario).

The RX 460 looks like it will perform like an R9 370 +10% extra and with better overclocking potential.

Both cards will handle 1080p but only at sub-ultra settings for 60fps.

 

[TehPenguin]

https://www.youtube.com/watch?v=3csIt7PEQzk

apparently the power draw issue will get a patch sometime today