AMD's Tonga - R9 285 (Specs) and R9 285X (Partial Specs)

[parvadomus]

TechPowerUp power consumption is completely wrong and biased towards Nvidia. I cannot find any other review with such difference between it and 760 (60watts delta lol).

Here its 10watts less:
http://ht4u.net/reviews/2014/amds_tonga-gpu_-_radeon_r9_285_im_test/index47.php

20 more watts:
http://hexus.net/tech/reviews/graphics/74033-sapphire-radeon-r9-285-dual-x-oc-28nm-tonga/?page=14

19 more watts:
http://www.guru3d.com/articles_pages/amd_radeon_r9_285_review,5.html

44 less:
http://www.hardwareluxx.de/index.ph...eon-r9-285-mit-tonga-gpu-im-test.html?start=6

Between +20 and -20 watts depending on bench:
http://pc.watch.impress.co.jp/docs/topic/review/20140902_664769.html

 

[Silverforce11]

It doesn't matter because the power use is in relation to other AMD cards.

It's slower than the regular 7970 and uses more power. It's fail.

Fill rate of memory bandwidth is NOT an advantage, those are technical and in a real benchmark, you don't see it because its still slower than Tahiti. Unless all you do with your GPU is watch 3dmark tests in isolation.

2GB R285 variant for $225 and 4GB R285X for $275 and they would have locked down the mid-range completely. This card doesn't deserve to go for $250 when it cannot even compete against Tahiti AND has less vram.

 

[parvadomus]

It doesn't matter because the power use is in relation to other AMD cards.

It's slower than the regular 7970 and uses more power. It's fail.

Fill rate of memory bandwidth is NOT an advantage, those are technical and in a real benchmark, you don't see it because its still slower than Tahiti. Unless all you do with your GPU is watch 3dmark tests in isolation.

2GB R285 variant for $225 and 4GB R285X for $275 and they would have locked down the mid-range completely. This card doesn't deserve to go for $250 when it cannot even compete against Tahiti AND has less vram.

Not in the links I posted

 

[Keysplayr]

I kinda doubt it, it was fairly obvious by summer time mining was no longer where it was. AMD had plenty of time to appeal to different segments.

This card would have needed to be planned well before summertime dude.

 

[GodisanAtheist]

Do we know whether or not Tonga is some sort of pipe cleaner product produced at GF? I remember the HD4770 being a disappointment in many of the same metrics R9 285 is.

In all likelihood AMD needed a new product name going into the Christmas season and GCN 2.0/The-Next-Big-Thing was still too far out.

 

[ocre]

AMD seems to have an uncanny knack for always showing their products in the worst possible light.

If they had released the full Tonga chip (2048 SPs, 384-bit memory bus), it would have beaten Tahiti in most benchmarks and tied at worst in the others. With a price of $279, that would be a clear win - not worth upgrading for existing 280X/7970 owners, but at least a step forward in the state of the art. It would have made even more sense to pioneer GCN 1.2 technology on their APUs - the high-end Kaveri is really starved for memory bandwidth, and delta-based texture compression would be a big win here. But instead they choose about the worst possible slot in their lineup for the new card - one where it can't decisively beat the prior generation in cost, framerate, or power consumption. What were they thinking?

I agree with the other posters who said that this was really meant to be a 20nm card. Had they known from the start they'd be stuck on 28nm, I think they would have done a better job of optimizing the transistor budget. I'd be interested to see how the full-fledged 285X fares against its predecessor in the power usage arena.

Lets look back through time at brand new architectures that launched only as chips with disabled parts. Nvidia has a few to speak of. The gk110 originally and the gf100. The gk110 could have been mainly because of yields. It was a massive chip and although it didnt sip power, the consumption didnt appear out of control. But it still might have had something to do with power consumption, this seems less likely with the gk110 but we know for fact that power consumption played a huge role in why the gf100 was released cut down.

So far there has only been two main reasons to launch a brand new architecture only with cards that have disabled parts: Power consumption concerns and to increase yields. The gf100 launched with most likely both of those in play.

See disabling parts just because is just not beneficial. The money is already spent for the silicon so disabling parts of chips for no reason would be strange. So which of the top two reason would Tonga be disabled for?
It isnt the most power efficient chip but surely its not totally out of control, like the gf100. So the other reason would be yield. Now being on 28nm for sooo long, you wouldnt think that yield would be a problem. Unless perhaps it was rushed into the 28nm node in a hurry, that might explain it.

Its really really strange when you see the claims of 5 billion transistors. This is insane. Could it really be true? Perhaps much more is cut down than suggested. I just find that hard to fathom. Its about 140% the gk104......

well, looking more at this Tonga has 116% the transistors of Tahiti. So its bigger but not so alarming after all. Maybe its needed for the bandwidth compression, encoding, and such?
I have been really doubting that the full tonga chip is 384bit. I just highly doubt this.

But then there is the final issue of the ram speed. Its a lot slower than i expected. So much so that there is no way its not starving the chip. And perhaps this is the final key to the mystery. Perhaps a full chip is so starved with such a limited bus and ram speed that AMD would rather put it off until they can get it resolved. Maybe they are having issues with memory controllers or their new bus. This is purely speculation on my part, 100%. I love talking about new tech and my brain starts firing in all directions.

But it is interesting when you consider that Nvidia only recently fixed their issues. In doing so it allowed them some insane ram speeds. launching a card with 6ghz ram when the gk104 launched was unheard of. It was impressive but now they have gone even further to 7ghz in the gtx770 and 780ti. I mean this really really helped Nvidia stretch out bandwidth on their small bus.

So the question is, why has not AMD launched a card utilizing the ultra high speed capabilities of Gddr5 modules? It is ironic considering AMDs close ties with Hynix and their suggested involvement in development of such things as HBM. There just seems to be no good reason they havent moved to faster ram speeds like nvidia. Unless they have some limitation such as their controllers. More bandwidth is always beneficial but at a certain point the benefits can become quite small. BUT....in the case of the r9 285, using such slow ram clocks is a major drawback. It makes very little sense to me and it seems likely there is some real reason....or that perhaps they are limited to this lower speed ram. For what ever reason......
But this too, it is 100% purely my speculation. Just some random thoughts on the chip

 

[AtenRa]

Except Barts was fundamentally a great chip. It was small and very power efficient. The only thing wrong with Barts at it's release was it's name. Tonga, on the other hand, is nearly the same size as Tahiti, is thus far slower, and isn't particularly more power efficient. What's the point? In fact, it's bigger than GK104, slower, and no more (if not less) power efficient. GK104 came out 32 months ago. What is not wrong with this scenario? A full-fledged GK104 is a $50 price drop and rebrand (GTX 960) away from making Tonga look like the worst release since the GTS 450.

Barts was slower than Cypress, cost the same as discounted HD5870/HD5850 cards at the time of Barts release and it was named a HD68xx series.
Yes it was smaller chip and it had lower power consumption than Cypress.

Tongas problem with the reviews is they all used custom cards and not the reference cards at the official settings. They also using differend drivers and different OSs, that makes the reviews to be all over the place in performance and power consumption measurements.
Also once again, drivers were not ready to fully exploit the new architecture, one two drivers from today the performance and performance per watt will be way different.

 

[3DVagabond]

Does anyone else think that this GPU was built for the anticipation (wrongly so) of continuing ridiculousness in the mining community? Cheaper to make than a 7950 but still pulling down two and a half bucks.

No. I have no idea why anyone would think that.

I dunno. It's just that this was so lackluster of a launch that I had to wonder what they were thinking. I mean, when Charlie Demerjian disses AMD, the apocolypse is nigh....

I can't believe you are actually channeling Charlie because he says something you like. He says anything even remotely bad about nVidia and he's a complete charlatan.

If you read him more you'd find that nobody is safe from his condemnation. He just really enjoys it when it's nVidia.

 

[Keysplayr]

No. I have no idea why anyone would think that.



I can't believe you are actually channeling Charlie because he says something you like. He says anything even remotely bad about nVidia and he's a complete charlatan.

If you read him more you'd find that nobody is safe from his condemnation. He just really enjoys it when it's nVidia.

Oh, well perhaps it's just me thinking outside the box then.

And "Channeling Charlie" ?? For real? That man has no business owning a computer as far as I'm concerned. And yes, I liked what he said. Is there a problem? I liked what he said because I know it bothers him deep down and is afraid Nvidia will do better. Don't you feel that way?
I don't read him at all, except when links are thrown at us in here or other forums. It's rare I ever click a semi-accurate link, but my curiousity was peaked when I read members comments here.

Hey, perhaps there is something in Tonga that we (by we I don't mean you also) and all reviewers are missing.

 

[3DVagabond]

The "fastest card" is a talking point and nothing more. No one in their right mind is going to buy that "fastest card" when dual card solutions are cheaper and just as fast. AMD having the best perf/$ is a matter of what the market dictates. Nvidia obviously commands a premium and nearly 2/3 consumers are willing to pay that premium. Nvidia and AMD have both reduced prices in response to each other. That is nothing new at all.

AMD is going to be in a world of hurt if Maxwell cards truly are coming next month and Nvidia scales up their TDP to the > 200 watt category. Maxwell will be faster, smaller, and more efficient than their direct competitors and Nvidia can price premium all day long while AMD simultaneously has to drop prices to stay competitive in frames per second only.

I understand. The only time the fastest card is worth a premium is when it's nVidia because simply being nVidia is worth a premium. Face it, people pay more for nVidia because they don't know any better. They pay more and get less performance, less VRAM, and pay extra for every little feature.

Personally, I don't care what 65% of the people buy. I simply pay less and get more. Makes me happy. And as long as there are enough people who do know better AMD will chug along. We've heard for a long time now that they are irrelevant, can't compete, doom, gloom, and despair, etc...

Hey, perhaps there is something in Tonga that we (by we I don't mean you also) and all reviewers are missing.

Not all reviewers. You might try clicking on some of the links posted that the reviews are positive about Tonga. For the record I'm not impressed with this card either. I just don't see it as a death toll for AMD as some do.

I think Tonga was due around the time of Hawaii (I remember reading rumors about a 256bit 7950 equivalent then), but due to mining they continued making Tahiti instead, which is better suited for mining. I think they were both originally planned for 20nm and that's why we had the lack of a proper cooling solution for Hawaii and Tonga isn't very efficient. nVidia seems to be coping better with diverting their chips to 28nm than AMD is. I can see a bump in the road here for AMD, but 20nm will be here and in the meantime they'll cope with perf/$, game bundles, their own features like Mantle and Freesync.

 

[Keysplayr]

I understand. The only time the fastest card is worth a premium is when it's nVidia because simply being nVidia is worth a premium. Face it, people pay more for nVidia because they don't know any better. They pay more and get less performance, less VRAM, and pay extra for every little feature.

Personally, I don't care what 65% of the people buy. I simply pay less and get more. Makes me happy. And as long as there are enough people who do know better AMD will chug along. We've heard for a long time now that they are irrelevant, can't compete, doom, gloom, and despair, etc...



Not all reviewers. You might try clicking on some of the links posted that the reviews are positive about Tonga. For the record I'm not impressed with this card either. I just don't see it as a death toll for AMD as some do.

I think Tonga was due around the time of Hawaii (I remember reading rumors about a 256bit 7950 equivalent then), but due to mining they continued making Tahiti instead, which is better suited for mining. I think they were both originally planned for 20nm and that's why we had the lack of a proper cooling solution for Hawaii and Tonga isn't very efficient. nVidia seems to be coping better with diverting their chips to 28nm than AMD is. I can see a bump in the road here for AMD, but 20nm will be here and in the meantime they'll cope with perf/$, game bundles, their own features like Mantle and Freesync.

Sorry. Should I have said "most" reviewers? The "positive" reviews don't make much sense as this card is unimpressive. And some people do see this as writing on the wall. So what. You shouldn't care about that. I personally think AMD is doing the best they can with what they have. That is all anyone can do.

 

[parvadomus]

Lets look back through time at brand new architectures that launched only as chips with disabled parts. Nvidia has a few to speak of. The gk110 originally and the gf100. The gk110 could have been mainly because of yields. It was a massive chip and although it didnt sip power, the consumption didnt appear out of control. But it still might have had something to do with power consumption, this seems less likely with the gk110 but we know for fact that power consumption played a huge role in why the gf100 was released cut down.

So far there has only been two main reasons to launch a brand new architecture only with cards that have disabled parts: Power consumption concerns and to increase yields. The gf100 launched with most likely both of those in play.

See disabling parts just because is just not beneficial. The money is already spent for the silicon so disabling parts of chips for no reason would be strange. So which of the top two reason would Tonga be disabled for?
It isnt the most power efficient chip but surely its not totally out of control, like the gf100. So the other reason would be yield. Now being on 28nm for sooo long, you wouldnt think that yield would be a problem. Unless perhaps it was rushed into the 28nm node in a hurry, that might explain it.

Its really really strange when you see the claims of 5 billion transistors. This is insane. Could it really be true? Perhaps much more is cut down than suggested. I just find that hard to fathom. Its about 140% the gk104......

well, looking more at this Tonga has 116% the transistors of Tahiti. So its bigger but not so alarming after all. Maybe its needed for the bandwidth compression, encoding, and such?
I have been really doubting that the full tonga chip is 384bit. I just highly doubt this.

But then there is the final issue of the ram speed. Its a lot slower than i expected. So much so that there is no way its not starving the chip. And perhaps this is the final key to the mystery. Perhaps a full chip is so starved with such a limited bus and ram speed that AMD would rather put it off until they can get it resolved. Maybe they are having issues with memory controllers or their new bus. This is purely speculation on my part, 100%. I love talking about new tech and my brain starts firing in all directions.

But it is interesting when you consider that Nvidia only recently fixed their issues. In doing so it allowed them some insane ram speeds. launching a card with 6ghz ram when the gk104 launched was unheard of. It was impressive but now they have gone even further to 7ghz in the gtx770 and 780ti. I mean this really really helped Nvidia stretch out bandwidth on their small bus.

So the question is, why has not AMD launched a card utilizing the ultra high speed capabilities of Gddr5 modules? It is ironic considering AMDs close ties with Hynix and their suggested involvement in development of such things as HBM. There just seems to be no good reason they havent moved to faster ram speeds like nvidia. Unless they have some limitation such as their controllers. More bandwidth is always beneficial but at a certain point the benefits can become quite small. BUT....in the case of the r9 285, using such slow ram clocks is a major drawback. It makes very little sense to me and it seems likely there is some real reason....or that perhaps they are limited to this lower speed ram. For what ever reason......
But this too, it is 100% purely my speculation. Just some random thoughts on the chip

I think the bandwidth improvements allowed it to use slow memory chips. My theory about releasing the PRO version first is binning, and get the best possible chips to counter mid-range maxwell. The XT version may be a 384-bit 2048 shaders with performance about a GTX780. Or just 256 + 2048 and better than 770s.

 

[Gloomy]

I highly doubt this chip, or even a full bin, is catching up to a 290/780.

 

[sm625]

AMD is in trouble:

Notice how performance per watt really has not improved significantly. If this is all they have to compete against the bigger maxwells then their margins are going to take a serious hit. GTX900 series will likely be over 200% more power efficient.

 

[Wild Thing]

Some people here need to get a grip...

R285 is a modern, updated, mid range chip that replaces pretty soon the older gens still on shelves.
At $249 for a new card with warranty and plenty of new features it will do just fine to replace the jumble of older cards still on the market.
AMD is working on a full fat GCN 2.0 chip which may well be on 20nm and will go head to head with whatever NVidia puts up as its class leader...and probably do it cheaper to boot.
R285 is just the opening shot in what will be a very competitive high and ultra high end class of cards due in the next 6 months.
Tonga is just a grasshopper....the heavy hitting R390X is going to be the one to beat.

 

[Keysplayr]

Some people here need to get a grip...

Don't be so hard on yourself. I'm sure you'll come 'round eventually.
Some of us here are trying to understand the purpose of this card. I mean, reasons are given, but not very credible ones.

 

[Arachnotronic]

Meh, whatever. Can we get a high end card launch from somebody, *please*?

 

[Abwx]

AMD is in trouble:

Thery are not, this site measurement protocol is buggy, they dont know or dont want to isolate the real datas, let s look at hfr numbers :

http://www.hardware.fr/articles/916-7/consommation-efficacite-energetique.html

 

[boxleitnerb]

How is their protocol buggy? Just because it doesn't show what you want to see?
Different games and test scenes -> different results. Yes it can be that easy!

 

[caswow]

How is their protocol buggy? Just because it doesn't show what you want to see?
Different games and test scenes -> different results. Yes it can be that easy!

you can say this either way...

 

[Abwx]

How is their protocol buggy? Just because it doesn't show what you want to see?
Different games and test scenes -> different results. Yes it can be that easy!

Because the site i criticised measure the whole plateform, a bigger GPU will stress the CPU more than a little one, they are measuring the efficency of a whole plateform not of a GPU in isolation, for this you must isolate the GPU power comsumption and relate it to its fps and this is what hardware.fr did, they measure the power through the pci connector as well as through 12V rails that feed the cards, hence they can extract numbers that have a true technical meaning, their numbers will be the same whatever the plateform.

you can say this either way...

Actualy he cant and he s on the bad side for the reasons above,
with Hfr protocol you cant twist the numbers by changing plateforms or CPUs.

 

[el etro]

Barts was slower than Cypress, cost the same as discounted HD5870/HD5850 cards at the time of Barts release and it was named a HD68xx series.
Yes it was smaller chip and it had lower power consumption than Cypress.

Good times. Reception for barts was not the best, but one or two monts later both cards became top sellers.

 

[RussianSensation]

Don't be so hard on yourself. I'm sure you'll come 'round eventually.
Some of us here are trying to understand the purpose of this card. I mean, reasons are given, but not very credible ones.

I'll give you some.

1) Performance. Even if R9 285 isn't as good from a performance/$ as 280/280X, it's still better than the $250 760. Sure, some 760s are going for $230 but it'll be really quick before 285 matches the 760 in price and well 285 is faster overall from nearly every review online:

http://www.computerbase.de/2014-09/amd-radeon-r9-285-test-benchmarks/5/

More importantly, most of the time it's expected that NV/AMD card at a similar price level could be trading blows +/-15% in some games. However, in some games, 760 completely bombs against the 285. It'll take a card that's sometimes 10-15% slower at a similar price but no way when it's 30% and 40% slower in big games like this.

R9 280 had an MSRP of $279 but market price quickly dropped to $250. AMD priced 285 at $250 so they could clear 280 asap. Once 280's stock runs out, 285 will drop below $250.

2) Next gen UVD. How many times have both NV and AMD released a next gen card and claimed xyz UVD features but we find out later they don't work for years? This is AMD's way of using 1 SKU to launch its latest gen UVD without being embarrased of not having these advertised features work on a flagship $550 300 series card next year.

3) Strengthening track record of high transistor density - an engineering exercise for the future.

Tahiti = 11.8 million / mm2
Hawaii = 14.16 million / mm2
Tonga = 13.93 million / mm2

vs.

780TI = 12.66 million / mm2
750Ti = 12.64 million / mm2

Since AMD hasn't been able to compete with NV on die size, they have to cram more transistors into a smaller space to make up for this disadvantage. More experience under your belt hitting 14 million / mm2 gives more confidence that perhaps the next gen design you could shoot for 15-16 million / mm2.

4) Improving memory bandwidth efficiency by 40%.

7950 has 36% more memory bandwidth (240 vs. 176) but it's slower. While this benefit doesn't mean much for 280/280X user, it means AMD reduced the need for redundant memory bandwidth that plagued 280/280X/R9 290X series by utilizing the available bandwidth more effectively. This could mean a 384-bit bus on 390X to reduce power usage, OR less memory bandwidth bottlenecks of GDDR5 before GDDR6/HBM drops OR less need to use more expensive and power hungry GDDR5.

5) More than doubling the geometry performance on the same node.

Alright, so 285 doesn't produce the magical 2x performance/watt of Kepler but none-the-less AMD more or less doubles Tahiti's geometry performance and trumps R9 290X by 70%! Since tessellation has been a weak spot for AMD, this is a big breakthrough without even discussing what GCN 2.0 will bring over GCN 1.2. OK so we know GCN 1.2 won't match Maxwell since it just catches up to Kepler but with AMD finally catching up to NV in tessellation, it forces NV to innovate and double tessellation with Maxwell. You know that magical phrase - competition forces innovation. :thumbsup:

6) Architectural timeframe context. Perhaps AMD cannot afford or doesn't want to take a risk of designing a brand new architecture like GCN for 3-5 years and then have it fail (Bulldozer anyone). First of all, it's very expensive and AMD doesn't have the financial resources of Intel/NV. Second of all, it's too risky. What they keep doing is improving GCN step-by-step as GCN is solid enough to last them another 2-3 years. They doubled tessellation performance per mm2, they increased memory bandwidth efficiency by 40%, and they did this in about a year since GCN 1.1 launched. Not bad consider Maxwell has been in development for 3-4 years! They have another 6-12 months to refine GCN 1.2 even more to improve performance/Stream processor and perhaps wait long enough for 20nm to close the performance/watt gap. Since NV is repeating bifurcation of its next generation (GK104-> GK110 and now surely GM204 / GM210), maybe AMD thinks it's better to have incremental changes to the architecture

There are even more points:

- new pre-scaler and upgraded display controllers
- upgraded instruction set for compute
http://videocardz.com/51462/amd-officially-announces-radeon-r9-285-never-settle-space-edition

^^^ These minor improvements don't translate to more performance in games but they allow AMD to test and slowly implement new features that deal with 4K and 8K compatibility, as well as focus on improving GPGPU performance for its HSA initiative and FirePro line.

So overall, for gaming only, 285 doesn't set the world on fire and fails hard on performance/watt but AMD has done a lot of other things here that allows them to focus more on performance/watt and other gaming aspects with R9 300 series having at least addressed memory bandwidth and tessellation issues with GCN 1.2.

Thing is though, between $200 R9 280, $250 R9 285, $260-280 R9 280X and $325 R9 290, $450 R9 290X, AMD has the entire $200-500 segment locked in for anyone but the most hardcore NV loyalist. And that means NV will have to respond with aggressive improvement in performance, performance/watt and performance/$$ to keep their market share.

AMD's desktop GPU competitiveness is nothing like its CPU competitiveness. That's a win-win for us gamers since NV can't stand still.

---
BTW, you never ripped 650Ti a new one and that card was an absolutel turd at $149. 285 at $249 is actually a better card than the 760 because it offers a more consistent gaming experience and comes with free games!

AMD is in trouble:

Only if you believe R9 2xx series will compete with Maxwell for the next 2 years. Fact - 285 has been designed and positioned as a 280 replacement, to compete with 760. If AMD set out a target to compete with a GTX960, they would have waited to launch it next year.

Let's see what NV does with pricing since rumours point to 970/980 being a 670/680/770 successor (aka faster than 780, maybe slightly faster than 780Ti but not a GK110 replacement). With R9 290 selling for $325-350 and R9 290X for $440-460, AMD has wiggle room to drop prices while NV will likely price 970/980 at $399/$499 to keep their 50%+ profit margins and recoup expensive R&D on a new architecture. I already said a long time ago that I expect Maxwell to take the next round since it's a brand new architecture going against AMD's improved 3-4 year old GCN. But since NV went all price premiums happy, it allows AMD to exist in lower pricing bands. Not everyone wants to pay $500-700 for a single GPU.

 

[el etro]

look at this(From hardware.fr):

3) Strengthening track record of high transistor density - an engineering exercise for the future.

Tahiti = 11.8 million / mm2
Hawaii = 14.16 million / mm2
Tonga = 13.93 million / mm2

vs.

780TI = 12.66 million / mm2
750Ti = 12.64 million / mm2

Since AMD hasn't been able to compete with NV on die size, they have to cram more transistors into a smaller space to make up for this disadvantage. More experience under your belt hitting 14 million / mm2 gives more confidence that perhaps the next gen design you could shoot for 15-16 million / mm2.

This reflects badly(and directly) on power consumption, is one of the things that made Hawaii and Tonga so inefficient.

 

[Keysplayr]

BTW, you never ripped 650Ti a new one and that card was an absolutel turd at $149. 285 at $249 is actually a better card than the 760 because it offers a more consistent gaming experience and comes with free games!

Probably because it wasn't called a 665Ti. See what I did there?

You are a pure bang for buck guy. That is it.