R9 300 cards listed in new driver - R9 370 is a rebrand

[Subyman]

We don't even know if its a 100% rebrand or a re-tweak die to add in more modern features etc. AMD has publicly said its entire 2015 lineup are all FS compatible for instance. That alone would rule out a straight re-brand.

If these are rebrands, then they aren't necessarily new GPUs. So they could still launch new "cards" without free sync and not be breaking any promises.

 

[dacostafilipe]

Yes we are, look at 280 vs 290.

Yeah ... depends on how we look.

280 is 75.9% of 290. (-24.1)

290 is 133.6% of 280. (+33.6)

http://anandtech.com/bench/product/1068?vs=1332

 

[shady28]

Don't forget the rumored Tonga XT, and that the R9 285 uses a "Tonga Pro". That leaves room for a "Tonga" below and a "Tonga XT" above.

I don't think this line is dead, it may change code-names.

ie :
http://wccftech.com/amd-radeon-r9-2...features-2048-sps-384bit-bus-3-gb-gddr5-vram/

From a user standpoint the R9 285 itself didn't add much, but it trades blows with the 384 bit R9 280 while using a 256 bit bus. That was a move in the right direction. If they tweak power efficiency some more and get the Tonga Pro to work with a single 6-pin they'll have a winner.

Financially, I would think AMD would consolidate its lines and look to reduce the size of the memory bus (and hence, the cost of making the card) to raise margins.

Here's a guess if it works out that way. Naturally they may change the code name of the dies but I really think Tonga Pro R9 285 was a precursor to their next gen GPUs.

Fiji - R9 390 / X
Tonga XT - R9 380 / X
Tonga Pro - R9 370 / X
Tonga - R9 360 / X


It would be really stupid for them to abandon Tonga. They can bin that chip and fill out their entire 360 - 380 lineup, and increase margins for themselves and the AIBs all at the same time. They just need to make it a bit more efficient to get on an even footing with Nvidia, which may have been the point of the R9 285 in the first place.

 

[bystander36]

I'm aware of the details, just wanted to show what can be done within the same node. I only meant that AMD can come up with a similar improvement.

But then that is not a rebrand. Reusing the same node does not mean it is a rebrand.

 

[tviceman]

Don't forget the rumored Tonga XT, and that the R9 285 uses a "Tonga Pro". That leaves room for a "Tonga" below and a "Tonga XT" above.

I don't think this line is dead, it may change code-names.

ie :
http://wccftech.com/amd-radeon-r9-2...features-2048-sps-384bit-bus-3-gb-gddr5-vram/

From a user standpoint the R9 285 itself didn't add much, but it trades blows with the 384 bit R9 280 while using a 256 bit bus. That was a move in the right direction. If they tweak power efficiency some more and get the Tonga Pro to work with a single 6-pin they'll have a winner.

Financially, I would think AMD would consolidate its lines and look to reduce the size of the memory bus (and hence, the cost of making the card) to raise margins.

Here's a guess if it works out that way. Naturally they may change the code name of the dies but I really think Tonga Pro R9 285 was a precursor to their next gen GPUs.

Fiji - R9 390 / X
Tonga XT - R9 380 / X
Tonga Pro - R9 370 / X
Tonga - R9 360 / X


It would be really stupid for them to abandon Tonga. They can bin that chip and fill out their entire 360 - 380 lineup, and increase margins for themselves and the AIBs all at the same time. They just need to make it a bit more efficient to get on an even footing with Nvidia, which may have been the point of the R9 285 in the first place.

Sigh. The mythical full die Tonga rumors persist.

Anyways, the gap between Tonga and Fiji is going to be massive. Unless Fiji yields are craps, AND will take a huge hit on binning to fill that gap.

 

[MeldarthX]

Sigh. The mythical full die Tonga rumors persist.

Anyways, the gap between Tonga and Fiji is going to be massive. Unless Fiji yields are craps, AND will take a huge hit on binning to fill that gap.

*chuckles* actually it lives......not that we'd ever see it as all of them are going to Apple......

 

[shady28]

Sigh. The mythical full die Tonga rumors persist.

Anyways, the gap between Tonga and Fiji is going to be massive. Unless Fiji yields are craps, AND will take a huge hit on binning to fill that gap.

http://wccftech.com/amd-tonga-xt-r9-m295x-benchmark-surfaces-impressive-results/

 

[Enigmoid]

. But I still don't think this is going to be a competitive product in 2015, not when it lacks modern features like FreeSync, Crossfire XDMA, and 4K video decoding.

None of those are very relevant for the target market.

Freesync is costing quite a bit at least for the next year or so. It still requires a monitor upgrade.

Who recommends crossfiring midrange cards like the 270X? Nobody.

4k monitors and $150 video cards do not go hand in hand.

Its a bit of a loss but the number of people looking for those features is pretty small.

IMO the furthest thing keeping AMD from respinning cards is that die size is going to grow if they increase performance. Maxwell increased perf/mm^2 but by much less than perf/W (980 is 60% faster than the 770 but 33% larger -> 1.6/1.33 = 20% better perf/mm^2). Adding XDMA, Freesync support, trueaudio (what happened to that BTW?) bloats the die even mode making the die more expensive. Might just as well sell the next larger old die for a lower price than end up paying masks and such and end up selling at the same profit.

Unless the new features results in a net influx of money there is no point in re-spinning.

 

[Enigmoid]

http://wccftech.com/amd-tonga-xt-r9-m295x-benchmark-surfaces-impressive-results/

Its already in the imac; 2048 core chip, 256 bit bus. And it overheats like mad (we are talking 105 degrees overheating)

It faster than the 780m but thats already an old chip, the 880m had been out for quite some time.

Is the M295X faster than the 780m? Yes.

Is it faster than the 880m? Yes.

Is it faster than the 970m? No.

It uses a lot more power than the 970m though.

 

[Cloudfire777]

You guys are a little behind in the mobile world.
M295X aka Tonga are already used by Dell in the Alienware 15. The only OEM using it, well except Apple. The rest use GTX 980M or 970M.
Thats the full Tonga, aka 2048 shaders.
R9 270X is probably a Tonga. Not sure if its Tonga (R9 285 rebrand - 1792 shaders) or Tonga XT (M295X "rebrand" - 2048 shaders). But who cares, sucks either way

The mobile M295X is a 125W chip. Desktop version with higher voltage, who knows. But its over 190W obviously.

It uses a lot more power than the 970m though.

Yes correct. 970M 75W-80W, M295X 125W.
Plus the 970M is 20% faster. And cost the same. Which is why OEMs don`t use the M295X due to the mess 880M became to be (hot, throttling due to power etc. It was a 125W chip too).

This is why I said earlier in this thread that AMD really really really needs a more efficient chip to even have a chance in mobile. But since all the high end/midrange desktop 300 chips is looking like rebrands, AMD will continue to bleed money and be almost non existant in the mobile space. Its horrible and they are perfectly aware of the situation :/

 

[shady28]

Its already in the imac; 2048 core chip, 256 bit bus. And it overheats like mad (we are talking 105 degrees overheating)

It faster than the 780m but thats already an old chip, the 880m had been out for quite some time.

Is the M295X faster than the 780m? Yes.

Is it faster than the 880m? Yes.

Is it faster than the 970m? No.

It uses a lot more power than the 970m though.

I know all of this, the link points to a 5 month old article about the m295X "Full Tonga XT" 2048 SP chip. It's pretty comparable to the 970M in performance.

My point is that this GCN 1.3 chip fits perfectly for the R9 380 / 380X, particularly if they utilize the full 384 bit bus that it is capable of. It would likely raise that line to compete with the GTX 970.

The Tonga Pro that currently resides in the R9 285 is perfect for the R9 270 / 270X. This would be a direct competitor to the GTX 960.

I don't think AMD is going to abandon this brand new architecture, which if you look at the rumor mill that is what people are saying (there's no Tonga in the future according to most of these posts).

Possibly they will tweak it to make it more power efficient and call it something new, but it will still be Tonga.

It would be incredibly, mind-numbingly, stupid of them to tack on Freesync and such on Pitcairn and call it something new when they have been ramping Tonga up over the last 6 months.

 

[therealnickdanger]

Allow me to play devil's advocate.

Freesync is costing quite a bit at least for the next year or so. It still requires a monitor upgrade.

While more expensive than average non-dynamic refresh displays, the upcoming FreeSync models appear to be undercutting G-Sync in price - which also requires a monitor upgrade. I would argue that the same type of people willing to buy into G-Sync would also be willing to spend less to get the same experience.

Who recommends crossfiring midrange cards like the 270X? Nobody.

The same people that recommend 960 SLI, perhaps? I think we can all agree that CF/SLI for anything but top-end cards is generally a waste, but there are people that do it and don't regret it. So here's another option.

4k monitors and $150 video cards do not go hand in hand.

They most certainly can and do - maybe not for gaming at 4K, but with 4K video hardware decoding, many people with GoPro cameras and other 4K-capable devices would certainly benefit from the feature. Even if you had a 1080p monitor, the massive bit rate improvement of 4K YouTube videos makes hardware decoding even more appealing for low to mid-range GPUs.

Its a bit of a loss but the number of people looking for those features is pretty small.

CUDA development was non-existent before NVIDIA started it and look where it is now. Adaptive refresh rates started out as a power saving measure and look where it is now. Sometimes niche features stay niche, but the often end up greatly impacting the market over time. Good on AMD if they can get all those features across the whole 3xx series.

 

[shady28]

...
The mobile M295X is a 125W chip. Desktop version with higher voltage, who knows. But its over 190W obviously.

...

Look at it this way.

The mobile M295X at 125W has superior performance and much better power characteristics to the desktop version R9 280. It's actually closer to the 280X in performance.

That would be the perfect answer to the GTX 960. At 125-150W, you only need a single 6-pin.

AMD could ramp up / ramp down frequency and ramp up / ramp down SP count to fill out their entire R9 370 - R9 380X lineup with that one chip, and the entire line would be GCN 1.3 and Freesync capable.

 

[RussianSensation]

If they rebrand most of cards then AMD is dead and you have right no PR will HELP there.

Not happening, no matter how much some people on this forum dream of this. AMD could literally sell 0 mobile and desktop GPUs until 14nm and not die. f
http://www.trefis.com/stock/amd/mod...09c261e1bb9b92be289e7935994576189&from=search

As I said before, at the current time AMD's business model is such that standalone graphics cards are NOT the primary cash flow lifeline of the firm. However, on this forum, certain people cannot differentiate/dive deep into the finance vs. how they think the company is run from their outdated viewpoint of AMD 2, 5, or 10 years ago.

As of right now, AMD can have 0% desktop and 0% laptop market share in graphics until Q3 2016 and they would survive until Radeon R9 400/500 series. Don't pay attention to the usual FUD. I've been reading about the death of AMD for 10 years, and every single quarter of every single year since Phenom I launched. The one theme remains consistent - the same individuals who were predicting the demise of AMD years ago and still do today are never able to financially justify why it will happen.

Its already in the imac; 2048 core chip, 256 bit bus. And it overheats like mad (we are talking 105 degrees overheating)

I've seen you post this several times. Some users on Mac forums have reported that under full load, the chip operates at 103-108C. However, you are looking to squarely put the blame on Tonga XT while ignoring completely that the iMac Retina is to blame. 108C doesn't kill the Tonga chip and Apple has issued a statement before that the iMac's temperatures within operating limits. As far as I know, there have not been any unusual cases of failed Tonga XT chips in the iMac, unless you have data to the contrary?

There is absolutely nothing wrong with Tonga XT itself as you are trying to allude. The problem lies in Apple choosing to use the wrong TDP part in an insufficient chassis (125W Tonga instead of an 85-100W GPU) with a horribly insufficient heatsink. In fact, the entire iMac Retina throttles, both the CPU and GPU. This would have been no different had Apple used an 880M in there with its own 125W TDP.

"Just to chime in, my M295X reached 108C in Windows during gaming.

The fan should have been at 2700rpm since the CPU was throttling too. I didn't check that though as I'm not particularly Windows savvy. It could have been 2300rpm. It was damn loud, either way."

or

"The i7 CPU and M295X GPU are hamstrung by the 5K iMac cooling system. The cooling is not sufficient for the components, so when under sustained load they both throttle to avoid overheating. Tests online show that they throttle by up to 15%. I don't agree with this, so I returned my iMac."
http://forums.macrumors.com/showthread.php?t=1815601&page=36

This is Apple we are talking about. They have a track record of CPU/GPU throttling for MacBook Pros and iMacs for years and years.

 

[Cloudfire777]

Look at it this way.

The mobile M295X at 125W has superior performance and much better power characteristics to the desktop version R9 280. It's actually closer to the 280X in performance.

That would be the perfect answer to the GTX 960. At 125-150W, you only need a single 6-pin.

AMD could ramp up / ramp down frequency and ramp up / ramp down SP count to fill out their entire R9 370 - R9 380X lineup with that one chip, and the entire line would be GCN 1.3 and Freesync capable.

R9 280 have a GPU score of around 10 000. Thats 20% better than M295X.
R9 280X score around 12 000. 45% better than M295X.

Thats one thing. Another is that mobile GPUs run slower clocks than desktop GPUs plus they are binned, meaning they can take lower voltage to reach the same clocks as desktop GPUs. The result of this is higher price for the mobile GPUs.

I don`t think AMD will start binning chips for a desktop Tonga XT chip. Not only because they need to keep the price down, but it would also mean they produce far less chips than required because the quality of the mass produced chips will not be good enough for the product. Desktop GPUs are sold in much higher quantities than mobile GPUs. Especially the high ends. So requirements go down

 

[shady28]

R9 280 have a GPU score of around 10 000. Thats 20% better than M295X.
R9 280X score around 12 000. 45% better than M295X.

Thats one thing. Another is that mobile GPUs run slower clocks than desktop GPUs plus they are binned, meaning they can take lower voltage to reach the same clocks as desktop GPUs. The result of this is higher price for the mobile GPUs.

I don`t think AMD will start binning chips for a desktop Tonga XT chip. Not only because they need to keep the price down, but it would also mean they produce far less chips than required because the quality of the mass produced chips will not be good enough for the product. Desktop GPUs are sold in much higher quantities than mobile GPUs. Especially the high ends

You can't look at a single benchmark and come to any conclusion when comparing different arch. That might have worked when comparing Pitcairn with X SPs vs Pitcairn with Y SPs, or frequency differences only, but not different archs.

The same is true comparing a GTX 760 Kepler vs a 960 Maxwell. They're different in too many ways.

To illustrate :
http://www.notebookcheck.net/AMD-Radeon-R9-M295X.129043.0.html

Some comparisons :

M295X (mobile Tonga XT) @ 125W vs R9 280X (desktop Tahiti) @ 250W

Cloud Gate Standard :
21766 vs 22027 (1% difference)

3DMark11 ovearall performance :

9516 vs 10241 (7% difference)

Cinbench 11.5 OpenGL 64bit M295X vs R9 290X

118.3 vs 117 (yes, the M295X is faster than the R9 290X here)


To reiterate, the mobile 125W M295X actually beats an 300W R9 290X on OpenGL 64bit.

Again, you cannot just take one bench out of context on new arch cards and make a valid statement. That may have worked for the last 3 years where we just had more SPs and higher clocks on minor variations of the same old arch, but Tonga is not a minor variation. You have to look deeper.

 

[Enigmoid]

Allow me to play devil's advocate.

While more expensive than average non-dynamic refresh displays, the upcoming FreeSync models appear to be undercutting G-Sync in price - which also requires a monitor upgrade. I would argue that the same type of people willing to buy into G-Sync would also be willing to spend less to get the same experience.

You are still talking a $150 GPU and a $300+ monitor. Not your typical gamer though as I mentioned there may be a fringe few.

The same people that recommend 960 SLI, perhaps? I think we can all agree that CF/SLI for anything but top-end cards is generally a waste, but there are people that do it and don't regret it. So here's another option.

Who exactly is that? 960 already doesn't have enough vram is is overpriced.
[/QUOTE]

Look at it this way.

The mobile M295X at 125W has superior performance and much better power characteristics to the desktop version R9 280. It's actually closer to the 280X in performance.

That would be the perfect answer to the GTX 960. At 125-150W, you only need a single 6-pin.

AMD could ramp up / ramp down frequency and ramp up / ramp down SP count to fill out their entire R9 370 - R9 380X lineup with that one chip, and the entire line would be GCN 1.3 and Freesync capable.

The M295X is about equal to 285 on the desktop. M295X has more shaders but runs at a lower clock.

I've seen you post this several times. Some users on Mac forums have reported that under full load, the chip can operate at 103-108C. However, it doesn't kill the chip and Apple has issued a statement before that the iMac's temperatures within operating limits. As far as I know, there have not been any unusual cases of failed Tonga XT chips in the iMac, unless you have data to the contrary?

Takes a while for heat to start killing stuff. Either way its unhealthy and definitely not good for the solder.

Would you honestly consider a GPU that ran at 105C under gaming load?

 

[Cloudfire777]

You can't look at a single benchmark and come to any conclusion when comparing different arch. That might have worked when comparing Pitcairn with X SPs vs Pitcairn with Y SPs, or frequency differences only, but not different archs.

The same is true comparing a GTX 760 Kepler vs a 960 Maxwell. They're different in too many ways.

To illustrate :
http://www.notebookcheck.net/AMD-Radeon-R9-M295X.129043.0.html

Some comparisons :

M295X (mobile Tonga XT) @ 125W vs R9 280X (desktop Tahiti) @ 250W

Cloud Gate Standard :
21766 vs 22027 (1% difference)

3DMark11 ovearall performance :

9516 vs 10241 (7% difference)

Cinbench 11.5 OpenGL 64bit M295X vs R9 290X

118.3 vs 117 (yes, the M295X is faster than the R9 290X here)


To reiterate, the mobile 125W M295X actually beats an 300W R9 290X on OpenGL 64bit.

Again, you cannot just take one bench out of context on new arch cards and make a valid statement. That may have worked for the last 3 years where we just had more SPs and higher clocks on minor variations of the same old arch, but Tonga is not a minor variation. You have to look deeper.

lol I give up. Believe what you want.
"Cant compare different arch"
Tonga and 280X ARE the same architecture.

GPU score (not total score) doesnt discriminate. You can compare a GTX 260 vs a GTX 980 just fine. Will you get the accurate gaming performance between them? No, but roughly the idea atleast

OpenCL means squat for gaming.

You are grasping at straws here mate. 125W desktop Tonga will never ever happen. The TDP of that card will go well over 200W.

Takes a while for heat to start killing stuff. Either way its unhealthy and definitely not good for the solder.

Would you honestly consider a GPU that ran at 105C under gaming load?

Whats the meting point of plastic? I would not buy that iMac

 

[Gloomy]

lol I give up. Believe what you want.
"Cant compare different arch"
Tonga and 280X ARE the same architecture.

GPU score (not total score) doesnt discriminate. You can compare a GTX 260 vs a GTX 980 just fine. Will you get the accurate gaming performance between them? No, but roughly the idea atleast

OpenCL means squat for gaming.

You are grasping at straws here mate. 125W desktop Tonga will never ever happen. The TDP of that card will go well over 200W.


Whats the meting point of plastic? I would not buy that iMac

It can happen if AMD limits it to 125W. Power consumption isn't set in stone, frequency and voltage can be lowered or raised as the manufacturer deems fit.

You're right though, the R9 285 is already drawing 176w gaming, I'm not expecting them to add more SPs and limit it to 125w and suddenly have a good GPU. But stranger things have happened, it could be okay (I have doubts though)

 

[geoxile]

Perhaps these rebrands are respins at GloFo. Didn't AMD imply they would be moving all their products to GF?

 

[Cloudfire777]

Perhaps these rebrands are respins at GloFo. Didn't AMD imply they would be moving all their products to GF?

Yup that could def happen. The upcoming chips could have better thermals and maybe performance too with GF chips. Identical specs as the 200 series (rebrands) but better characteristics due to better silicon.

AMD did say they were gonna manufacture at GlobalFoundries. They mention GPUs two times in this slide. But then again, it could be related to the GPU part of their APUs. I dont know

 

[RussianSensation]

Takes a while for heat to start killing stuff. Either way its unhealthy and definitely not good for the solder.

Would you honestly consider a GPU that ran at 105C under gaming load?

I added more info to my original post. Please read it again. The iMac Retina itself throttles, both the CPU and the GPU. The cooling is insufficient and it's Apple's fault for choosing a 125W mobile part to integrate into a chassis built for an 85-100W one. Tonga XT is not to blame here. It's not like AMD advertised it as an 85-100W part but it uses 125W. AMD was transparent about what the chip was from day 1. It's 100% the manufacturer's responsibility to make sure they choose the appropriate product for a specific chassis/cooling setup. Apple is 100% to blame for this.

lol I give up. Believe what you want.
"Cant compare different arch"
Tonga and 280X ARE the same architecture.

Not even close. 285 has higher pixel-fill rate than an R9 290X despite the latter having double the ROPs (!!!) and Tonga has DOUBLE the geometry performance of a 280X + 40% more efficient color fill-rate/memory bandwidth efficiency compared to 280X/290X. Tonga can process 4 rasterized triangles per second vs. only 2 for Tahiti.
http://techreport.com/review/26997/amd-radeon-r9-285-graphics-card-reviewed/2

Seriously, after countless threads that discussed/proved Tonga's GCN 1.2 architecture's improvements over GCN1.0/1.1, people still think Tonga is the exact same architecture as a 280X? Tonga's problem is its both TMU and SP limited, which is why in 99% of games its architecture's advantages don't even show up. Take Tonga's advancements and put them inside a 4000+ SP and 256 TMU chip, and it's going to show up, guaranteed.

While we aren't going to see a 2048SP 125W Tonga on 28nm, it's possible AMD can still refine Tonga and release a faster Tonga part than a 280X with lower power usage than a 280X.

 

[shady28]

lol I give up. Believe what you want.
"Cant compare different arch"
Tonga and 280X ARE the same architecture.

....

I'm no AMD fan but what you've said in the last couple of posts is garbage.

Here's a place to start educating yourself. Do these chips sound like they are the same to you :

http://www.extremetech.com/gaming/1...cn-1-2-torpedo-that-takes-out-nvidias-gtx-760


"The R9 285 has a 359mm2 die and 5 billion transistors, compared to 352mm2 and 4.3 billion transistors for the old Radeon HD 7970/R9 280X. AMD has confirmed to us that the new chip has 32 CUs total (with 28 functional), implying that AMD has the option to bring a new version of the core out if it chooses to do so — but even the additional CUs don’t explain why a GPU with a smaller memory bus is both larger and more dense than its predecessor."

 

[tviceman]

*chuckles* actually it lives......not that we'd ever see it as all of them are going to Apple......

Wait for it.....

http://wccftech.com/amd-tonga-xt-r9-m295x-benchmark-surfaces-impressive-results/

Its already in the imac; 2048 core chip, 256 bit bus. And it overheats like mad (we are talking 105 degrees overheating)

It faster than the 780m but thats already an old chip, the 880m had been out for quite some time.

Is the M295X faster than the 780m? Yes.

Is it faster than the 880m? Yes.

Is it faster than the 970m? No.

It uses a lot more power than the 970m though.

All of the above. It's not 384-bit as many think it is, it doesn't come close to Hawaii performance (we're talking 280x at best), and six months has elapsed since Tonga was debuted. Keep in mind that at 359 mm2 it can't possibly be suffering from bad yields due to die size. Yet no desktop release.

Current performance, on the whole, matches an R9 280 and probably only matches an R9 280X (at best) fully functional. If Fiji is 290x + 60%, we're looking at a 100% performance difference between Tonga and Fiji. So, as I said, the gap between Tonga and Fiji will be massive and unless Fiji has terrible yields and has 4 total desktop SKUs, then Hawaii will be needed to fill the gap.

 

[shady28]

...

Current performance, on the whole, matches an R9 280 and probably only matches an R9 280X (at best) fully functional. ...

R9 285 1080p and 2160p index beats the R9 280X :

http://www.tomshardware.com/charts/...tml?prod[7297]=on&prod[7279]=on&prod[7276]=on


2160p R9 285 vs R9 280X vs R9 280X :68.81 65.71 64.60

1080p : 78.41 72.70 71.85


This is with 1792 SPs. With 2048 like the M295X it should be another 10% or so faster.

And no it doesn't win on every game. It depends on the game / benchmark. But there are enough where it wins that the aggregate (overall) is better.


Biggest gripe people had with the R9 285 was the 2GB. That negatively impacts some newer games and games at higher resolutions. That is an artificial limit though, its probably a marketing decision.

That a 256bit 2GB card can walk over their older 384bit 3GB card on anything should be saying something (including things like - this isn't the same architecture).


i.e. :