AMD Radeon R9 Nano Nears Launch, 50% Higher Performance per Watt over Fury X

[GhostRyder]

I thought the 390 was a 290 with higher clocks. Is there something new with the architecture?

Nope, just a more mature and refined process. Most of what I have seen show 390 exceeding 1150mhz where as before 1100+ was hit or miss. Not much of an improvement, but enough it would seem to keep it relevant (For instance my 290X trio while each is slightly different get a little ridiculous beyond 1130).

 

[EarthDog]

I thought the 390 was a 290 with higher clocks. Is there something new with the architecture?

Nope. Just clocks.

 

[newtekie1]

Not exactly, the 390 can overclock a bit further than its predecessor 290 because of improvements on the silicon similar to the 390X's improvements.

I didn't say anything about the 290. Regardless of these "improvements" the 390 is still pushed almost to its limit with the stock clocks, which results in poor overclocking potential. The 970 is a much better overclocker, there is no arguing that.

And, IMO, the better overclocks are coming from the better PCB designs, not necessarily anything to do with the GPUs themselves.

 

[Aquinus]

I thought the 390 was a 290 with higher clocks. Is there something new with the architecture?

Just the extra VRAM, faster VRAM, and higher stock clocks. Nothing different with the GPU itself.

Sure, but the 390 isn't any better out of the box than the 970. At 1440p, their target market, they're basically dead even. And at this point the 970 is cheaper than the 390, so the question comes down to paying more for a card that is currently equal in performance, is worse in every other way, just in the hopes that the 8GB of memory becomes useful. And most of the experts(reviewers) have already said the 8GB won't be useful because the core just isn't powerful enough to utilize it. By the time you crank up the graphics to the point that the 8GB would be useful, the core is so bogged down with processing that you get no benefit anyway.

Will it? AMD's GPUs tend to be TMU heavy for the price you pay for them but less pixel pumping power which would explain lesser AA performance. The question is will an AMD GPU suffer as much as a nVidia one as more and bigger textures are used? I wonder if certain parts of the GPU are being left under-utilized because of how games run or what their demands actually are.

Either way, a little bit of patience and we'll see.

 

[GhostRyder]

I didn't say anything about the 290. Regardless of these "improvements" the 390 is still pushed almost to its limit with the stock clocks, which results in poor overclocking potential. The 970 is a much better overclocker, there is no arguing that.

And, IMO, the better overclocks are coming from the better PCB designs, not necessarily anything to do with the GPUs themselves.

Well I just meant there are improvements. But even so my point was that while the GTX 970 and others overclock to levels of 1400-1600, that is on the same levels as the R9 390 at just 1150-1200 according to benchmarks that I have been seeing. I am not arguing which is a better overclocker (As around 500 MHz versus up to 200mhz is obviously better), just that the numbers don't tell the whole story.

Also I do not think the better PCB designs are the only reason because we had non-reference 290X's and 290's that didn't yield automatic higher clock unfortunately.

 

[anubis44]

I like the idea, and the R9 NANO is probably my favorite from AMD's show a while ago.
But.

This is not the product AMD needs right now, at all.
We dont need more product that cost 100+$ above the GTX 970 and perform just slightly better. (slideshow shows 30.5FPS for R9 290X and 33FPS for R9 NANO, so... about 390X levels.)
We need a competition that will actually drive the market and get AMD out of its bankrupcy danger.

If AMD sells a shed-load of these NANOs, that will help their bottom line without doubt. It's looking like a pretty darn nice card, and I wouldn't be surprised if there weren't lots of custom cooled versions that might overclock very nicely. This, plus the greatly buffed Radeon performance in DX12, and AMD's upcoming Zen CPUs/APUs, and a nice uptick in PS4/XBox 1 sales will all contribute to AMD's financial success.

 

[cadaveca]

If AMD sells a shed-load of these NANOs, that will help their bottom line without doubt.

AMD fails to meet price/performance offered by NVidia at all price points in my local market. As such, no sales person in any PC store will recommend AMD unless it's specifically asked for. There's no features or anything at all - not even a free game - to entice the public to buy Nano. Fans of AMD will buy for sure, if priced right, but given that this is back-to-school season, and then holidays... I about a single SKU will really help AMD at all. They'd need to sell 500k units just to break even on this single SKU, IMHO, and I doubt they will have capacity for even half that.

But, I'd buy one, heck I'd buy a quartet of FuryX, if I could find them on the local market. I doubt that will happen soon since FuryX isn't even here yet.

 

[Casecutter]

While not the proper thread to discuss this... AMD should have never waited the release the 300 Series Re-Fresh cards and never all at the same time. I suppose there was the whole "working down inventory in the channel" was a major problem. While I think AMD was working hard to complete some driver "secret sauce", so perhaps it hinged on those two points.

Today seeing the "refresh" of the 300 Series as a pertinent adjustment, it distressing that they didn’t (juggled against selling down 290's) just released the 390; even if 290’s ($270) still had to have a place in the product stack. As all the refresh cards are furnishing the opposition a decent buzz in the market, getting them out early would’ve be advantageous. Even if it meant having the 200 Series below them, and the 300 Series above at the same point in time. In retrospect it doesn’t seem that problematic as not having something to maintain some usable PR on the front pages. Imagine AMD 390 in the market mid-April touting 8Gb and vying the 970, while the a Memory Allocation was somewhat still a sore subject. Then the 390X showing mid-May, finally the 380 like two week later.

A nice staggered release of new reviews leading up several week before the FuryX , then 3 weeks from that Fury, and now Nano. All that perhaps taking pressure from the drumming Fiji/HBM. AMD might have had to be less "talkative" on that subject, holding to "it's planned for a July release". AMD would’ve almost assuredly had more and favorable reviews (AMD needs to provide more reviewer samples), but no they choose "months" of silence and the mounting pressure to "do something". It all came apart at the seams as the "where’s Fiji/HBM" overtook their own PR with crazy speculation in the forums, all the while the "re-brand tag" really got unwanted traction. If AMD had a glut of 200 Series inventory in the channel, that was unhealthy, but the 6 months of their silence and the web's speculation was just as life-threatening.

 

[anubis44]

Yeah but, I bought my 390 with the intent that it would be good out of the box. You're right, I can't usually pull more than 1160Mhz on it without pumping a good bit of voltage and even then. The question was what were the alternatives. Paying a tiny bit extra for the 970 didn't make sense when the 390 does almost the same but gives you that 8GB of VRAM. Not to say that's useful yet but I've been occasionally touching that limit in Farcry 4.

The 390 has a lot of texturing capability versus the 970. On paper you would expect the 390 to do something vastly better than the 970 but we don't see that in a lot of cases. I suspect when we start using more memory for higher resolution textures that the 390 will suffer a lot less than a 970. This is all to be seen though. More demanding games are in order for us to see how that all goes over.

Your 390 with 8GB will get a nice boost in performance in DX12. GTX970 on the other hand will be stuck with the same performance in DX12 as DX11, and only 3.5GB of 'normal' memory.

 

[64K]

More news about the Nano

http://videocardz.com/57409/amd-radeon-r9-nano-confirmed-to-feature-4096-stream-cores

 

[moproblems99]

As such, no sales person in any PC store will recommend AMD unless it's specifically asked for.

Serious question, who actually goes to a store and buys PC products? At least near me, the only thing there is Best Buy, and well...its Best Buy.

 

[Aquinus]

More news about the Nano

http://videocardz.com/57409/amd-radeon-r9-nano-confirmed-to-feature-4096-stream-cores

The source's source seems of questionable integrity.

 

[xenocide]

Your 390 with 8GB will get a nice boost in performance in DX12. GTX970 on the other hand will be stuck with the same performance in DX12 as DX11, and only 3.5GB of 'normal' memory.

We can't really know that for sure. A single game that uses DX12 (which was originally optimized for Mantle) isn't a good way to definitively say the AMD cards will be better in DX12. I found it odd that in the Ashes benchmarks at all resolutions the AMD cards saw noticeable improvements going from DX11 to DX12, but the Nvidia cards almost overwhelmingly got worse in every instance or broke even. You would think by virtue of DX12 being a more efficient API they would at least break even under both,

 

[cadaveca]

Serious question, who actually goes to a store and buys PC products? At least near me, the only thing there is Best Buy, and well...its Best Buy.

We have several large retailers in Edmonton with large stock on most items. I will not mail order something when I can get it "price-matched" locally when building rigs for other locals. It just doesn't make any sense when it comes time to deal with returns and RMAs, since local shops are pretty good with quick turn around for RMA, and exchange or return is simple.

check out "memory express", they are just one of the shops I tend to visit on a near weekly basis. Since they price match any online retailer that is based on Canada, they've cornered the market for me since I get the price of other shops, but never pay shipping fees, even if just buying thermal paste (which is something I buy monthly).

 

[moproblems99]

We have several large retailers in Edmonton with large stock on most items. I will not mail order something when I can get it "price-matched" locally when building rigs for other locals. It just doesn't make any sense when it comes time to deal with returns and RMAs, since local shops are pretty good with quick turn around for RMA, and exchange or return is simple.

check out "memory express", they are just one of the shops I tend to visit on a near weekly basis. Since they price match any online retailer that is based on Canada, they've cornered the market for me since I get the price of other shops, but never pay shipping fees, even if just buying thermal paste (which is something I buy monthly).

I envy you, that is a really cool getup.

 

[EarthDog]

Lol dave, buy a baster's worth and stop the frequent trips.

 

[anubis44]

So the Performance/Watt is not about FpS, but the efficiency of the bandwidth?


While yes 680 could be considered Nvidia's mid-range, one could consider Tahiti was the mid-range. At the time AMD could see Nvidia couldn't absolutely start flooding the Full-Keplers into the gaming market for many months... while yes Titan showed @ $1,000 Feb 2013... AMD saw no reason to substantiate what was a semi-quasi Professional offering. Had they had 7990 prior to Titan they could've at least had something but it took till April at that point it had little value.

The true GTX Gaming version showed as GTX 780 June 2013; AMD had Hawaii in the market Oct 2013, but yes the 290X was 5 months behind. That's was where AMD really started faltering, not have the money or foresight to get moving with big die's like Hawaii and Fiji has been their Achilles Heel.

AMD has plenty of money to evolve their cores, just like nVidia does. The fact that nVidia gives them completely different names like 'Kepler' and 'Maxwell', and AMD uses revision numbers like GCN 1.0 and GCN 1.2 doesn't mean AMD's cores aren't changing. Since GCN 1.0, Radeon cores have been optimized for DX12. It's not AMD who's behind, it's nVidia. nVidia has pushed out DX11 junk as recently as Maxwell, and tons of people have lapped it up. Instead of seeing that nVidia shoved out Maxwell last September because they were scared sh!tless DX12 benchmarks would expose their weakness in that API, tons of nVidiots went out and bought $400+ Maxwells like they were going out of style only based on DX11 benchmarks. Now they're going to regret their decision shortly when they see a $250 R9 290 pushing pixels like a $1,000 Titan in DX12.

 

[moproblems99]

AMD has plenty of money to evolve their cores, just like nVidia does. The fact that nVidia gives them completely different names like 'Kepler' and 'Maxwell', and AMD uses revision numbers like GCN 1.0 and GCN 1.2 doesn't mean AMD's cores aren't changing. Since GCN 1.0, Radeon cores have been optimized for DX12. It's not AMD who's behind, it's nVidia. nVidia has pushed out DX11 junk as recently as Maxwell, and tons of people have lapped it up. Instead of seeing that nVidia shoved out Maxwell last September because they were scared sh!tless DX12 benchmarks would expose their weakness in that API, tons of nVidiots went out and bought $400+ Maxwells like they were going out of style only based on DX11 benchmarks. Now they're going to regret their decision shortly when they see a $250 R9 290 pushing pixels like a $1,000 Titan in DX12.

How long did it take DX11 to become mainstream? I think people have plenty of time to enjoy their current purchases.

 

[geon2k2]

We can't really know that for sure. A single game that uses DX12 (which was originally optimized for Mantle) isn't a good way to definitively say the AMD cards will be better in DX12. I found it odd that in the Ashes benchmarks at all resolutions the AMD cards saw noticeable improvements going from DX11 to DX12, but the Nvidia cards almost overwhelmingly got worse in every instance or broke even. You would think by virtue of DX12 being a more efficient API they would at least break even under both,

Well there are some reports on the web that AMD has more lanes to communicate with the CPU and that it will for sure see better performance improvements with the DX12. Even the game in which AMD cards sucked really bad on DX11, Project Cars, seems to benefit a lot from DX12.

Read more here:
http://hardforum.com/showthread.php?t=1861353&page=4

There was another article on a tech site not on a forum, however I cannot find it now ....

 

[arbiter]

AMD has plenty of money to evolve their cores, just like nVidia does. The fact that nVidia gives them completely different names like 'Kepler' and 'Maxwell', and AMD uses revision numbers like GCN 1.0 and GCN 1.2 doesn't mean AMD's cores aren't changing. Since GCN 1.0, Radeon cores have been optimized for DX12. It's not AMD who's behind, it's nVidia. nVidia has pushed out DX11 junk as recently as Maxwell, and tons of people have lapped it up. Instead of seeing that nVidia shoved out Maxwell last September because they were scared sh!tless DX12 benchmarks would expose their weakness in that API, tons of nVidiots went out and bought $400+ Maxwells like they were going out of style only based on DX11 benchmarks. Now they're going to regret their decision shortly when they see a $250 R9 290 pushing pixels like a $1,000 Titan in DX12.

Wow just Wow i don't even know where to start on tearing up everything you just said as complete stupidity.

 

[anubis44]

The anti-AMD lunacy in this thread is breath-taking.

News flash. AMD is the one who has been giving more value for money than nVidia. All GCN Radeons (but especially GCN 1.1 and newer) will tend to be MUCH faster in DX12 than they were in DX11, since GCN was designed to for compute performance fed by multi-core CPUs running multi-threaded code. nVidia GPUs, including the current Maxwell, are designed to be fed by a single-core in serial. In other words, all current nVidia GPUs are rendered obsolete by DX12, if you'll excuse the pun.

Don't belivee me? Read this explanation by a retired AMD GPU engineer. This guy REALLY knows what he's talking about, he's no fanboy, he's an actual, bonna fide expert, his posts (go ahead and read through the thread and be prepared to be blown away) should silence the idiotic nVidia fanboyism on here:

http://www.overclock.net/t/1569897/...singularity-dx12-benchmarks/400#post_24321843

 

[64K]

@anubis44 The question I have is how long before there are enough DX12 games for it to matter. When I play a game and I enjoy it then I will likely play it again after a couple of years. All those DX9 and DX11 games in my library are still a factor for me to consider. If more and more DX12 games start showing up that run better on an AMD card then it will be time to go red but that could conceivably take a couple of years.

Both AMD and Nvidia are deceitful at times so I would rather wait until some DX12 games drop before deciding.

 

[moproblems99]

Also, I have no hate for AMD. They were just too late to the dance this time.

 

[cadaveca]

The anti-AMD lunacy in this thread is breath-taking.

News flash. AMD is the one who has been giving more value for money than nVidia. All GCN Radeons (but especially GCN 1.1 and newer) will tend to be MUCH faster in DX12 than they were in DX11, since GCN was designed to for compute performance fed by multi-core CPUs running multi-threaded code. nVidia GPUs, including the current Maxwell, are designed to be fed by a single-core in serial. In other words, all current nVidia GPUs are rendered obsolete by DX12, if you'll excuse the pun.

Don't belivee me? Read this explanation by a retired AMD GPU engineer. This guy REALLY knows what he's talking about, he's no fanboy, he's an actual, bonna fide expert, his posts (go ahead and read through the thread and be prepared to be blown away) should silence the idiotic nVidia fanboyism on here:

http://www.overclock.net/t/1569897/...singularity-dx12-benchmarks/400#post_24321843

That's all nice, but I don't use any DX 12 apps at this time, so there is zero value to me from such until I buy a game, and then it'll be that one app. It's hard for me to be excited about something I will only use in the rare occasion. Do also keep ion mind I used to run around calling myself AMD's #1 fanboy. I cannot at this juncture, because their offerings and false marketing and lies left a bitter taste in my mouth. Nevermind that they have failed to keep high-end SKUs in stock locally.

I made many AMD GPU purchases on tech they touted as new and all that, but not once did they ever actually deliver on it until many many months later, if at all (Crossfire (broken cursor), Eyefinity (First was Crossfire running Crysis on three dell 30-inch screens, never did work in native res although they had live demos running at LANs and when I was there, they wouldn't let me see the back of the monitors or PC) + (Frame time problems that I complained about for years before they actually acknowledged, and only did when they had ZERO choice), CTM (never made it to prime-time, precursor to DX12), Mantle (New CTM, backed by better code, still not used by more than 10 or so games right now).

It's not that I favor AMD, it is that they fail to deliver. While for me personally and my own purchases, NVidia HAS delivered. NVidia touts features...and they work! What a novel idea!

Until AMD actually PROVES that what they hype before a lunch is actually a true statement, it'll be hard for most anyone to take them seriously.

That same post about ashes was also made here on TPU, but yet you link to another site... Interesting.

 

[EarthDog]

@anubis44 The question I have is how long before there are enough DX12 games for it to matter. When I play a game and I enjoy it then I will likely play it again after a couple of years. All those DX9 and DX11 games in my library are still a factor for me to consider. If more and more DX12 games start showing up that run better on an AMD card then it will be time to go red but that could conceivably take a couple of years.

Both AMD and Nvidia are deceitful at times so I would rather wait until some DX12 games drop before deciding.

Its worth it to QFT and give thanks...

While I understand what our AMD fellow is getting at, this is the WHOLE POINT. I could buy AMD now, have a DX12 AAA title or a few come out in the next 2 years and have better performance in those. OR I can buy a different card now that has better performance in the VAAAAAAAAAAAAAAST majority of games. Then, in two years when more DX12 titles are available in the market, then look at who's top dog and make a purchase then.