Radeon R9-M295X

Hi guys,

Read on a japanese thread that MSI
Launched a new lineup of notebooks .
The unannounced Radeon R9-M295X is in one of this machines.

未発表のAMD Radeon R9-M295Xの発売もラインナップに含まれており、徹底的にゲーミンググレードにこだわったブランドであることがわかる。

MSI


Do you guys know something about this GPU?

Thanks.

PS
Excuse my English. I'm japanese.
Tsubasa

 

Could it be that we're getting a new mobile GPU based on the Hawaii chip?

 

Bet it will be paired with an AMD processor!

 

Strange. Never heard about it. The site you link to plus 4gamer is the only sites that mention the name of that GPU.

AMD is announcing R9 295X2 this month, a dual GPU for desktops. Obviously we are not getting a dual GPU in mobile.

I have no idea to expect, but I know 4Gamer is pretty legit.

 

Looks like AMD is still in the game, but that doesn't mean they match up. I'm assuming it will be to compete with Maxwell (which is somewhat amusing to think about, ha ) which should be released during or after the summer.

 

Hawaii got absolutely nothing on Maxwell. Not even close in terms of efficiency.
I`m thinking this card is competing against GTX 780M/880M.

A little late but atleast its something. 4Gamer said the card is going inside GX series from MSI.

 

Yeah, that's what I was getting at when I said it was amusing to think about (funny).

 

Not sure if they'd be able to fit Hawaii in a mobile form considering the TDP, and they don't have anything in their roadmap that could compete with 780m/880m any time in the near future... I'm a bit stumped on what this card could be.

 

Perhaps it's the 3rd re-brand of the 7970M, but with 8GB's of VRAM, LOL.

 

That would be funny

 

I don't see why MSI won't put the AMD card into an intel based platform. Seeing as they just released a bios for support of the AMD cards. I am now able (at least as far as I understand it) able to rock an AMD R9 series card in my GT60. They've already released a laptop with an intel/amd combo, and for the budget gamer, this is one hell of a rocking combo. Much better option than the AMD/AMD combo.

 

That what it seems, it will be fitted in the "Destroyer Pro", maybe it is a Kaveri GX (there was a Dell inspiron Roadmap where seemed that laptop kaveri will be launched on april)?

 

You may be on to something here marcos.
R9 295x2 is a dual GPU card.
R9 M295x might be a Kaveri with Hybrid CrossfireX with R9 M290x. Giving it a small boost in performance.

The GPU is mentioned anywhere which means that MSI may be the only one using it, and MSI are the only OEM that use AMD GPUs with AMD APUs...

 

Heh, heh. The thought did actually occur to me although it would make AMD an utter laughing stock.

 

But the 880M is a 780M rebrand? If NVIDIA can rebrand, so can AMD

 

With higher stock clocks though. R9 M290X is exactly the same as 8970M.

 

I have been thinking and maybe it is posible that this M295X is based on the Tahiti LE chip found in the desktop 7870 xt?

 

Pffft M290X with Kaveri APU in Dual Graphics?!? I surely hope AMD isn't that stupid. The microstutter on that setup would be insane. Just imagine if the dGPU takes 16 ms to render a frame while the iGPU takes 50 ms to render the next one and so on and so forth in AFR. I would love to see the frame times chart on that one!

However, if the Kaveri onboard GPU is used as a separate co-processor utilizing HSA to handle non-rendering chores like physics calculations, e.g. OpenCL-accelerated Bullet, while the dGPU takes care of rendering? It becomes a discrete PPU setup, like having a dedicated PhysX card. Now that would be cool.

 

I think it'll be a r9 m290x with higher clocks. Hawaii is unlikely since AMD never put Tahiti into mobile, and hybrid crossfire is also unlikely because the driver overhead will be bigger than any performance gains to be had.

 

4th times the charm lol. Taking rebranding to a whole new level. One can almost smell the desperation

No seriously, that wont happen. I refuse to believe it

 

As far as unfounded postulation and educated polemics goes, this topic is the most interesting of all the discussions on new-gen cards here, IMO. I'm not really that interested for my purchase decisions as I still believe AMD hasn't anything close to what NV will offer in 20nm Maxwell (or to compete with fully-enable Kepler, for that matter): I am more interested in the mystery of it. Is it a Kaveri-Pitcairn Frankenstein? Is it some "7800 XT" derivative (...doubtful, IMO)? Is it simply a Crossfire setup branded as a single card a la the w90vp's "4870 X2"?

Agreed, Cloudfire: AMD rebranding Pitcairn yet again for this card would be a massive face palm the likes of which we haven't seen since, well... ever. I'm a huge supporter of AMD for the mere fact that a GPU market without competition is horrible for the consumer, but if they do this, I'll be the first lining up on every forum from here to Engadget tearing them a new one to raise awareness of the rebrand.

 

I found some new info in and the R9 M295X won't be a rebrand. It is based on the AMD's upcoming Tonga GPU, which is expected to launch this summer. What is most impressive though is the fact that even though it is still 28nm, it will be based on the new GCN 2.0 architecture that will be in Pirate Island GPUs and even more important is the fact that Tonga is supposed to bring some crazy improvement in performance/Watt. Maybe this is a result of moving to one of the 28nm processes at Globalfoundries. We should expect R9-M295X to completely destroy anything nVidia currently has in mobile not only in performance but also efficiency.

Edit: Just to give some performance estimates, Tonga is supposed to be a replacement for Pitcairn because of 2GB memory and 256bit bus (rumors), with much higher performance. Some people even claim that the desktop variant might even be in the performance category of Tahiti. This claims are probably rumors of course but for the desktop cards they will be R9-2XX series and for workstations FirePro W8100. I just looked at the FirePro W series specs and what this does mean is that it is more powerful than the W8000 which is of course Tahiti Pro but with a 256bit bus and 900MHz core clock (faster than the original 7950). Now W9000 is Tahiti XT and W9100 is Hawaii XT, and judging by the Performance Gap between the W9000 and W9100, W8100 might have some sizeable performance increase even compared to Tahiti Pro.

 

That would be nice to bring back some good old competitioj.

Sent from my C1905 using Tapatalk

 

I do hope R9 M295X "destroy" GTX 880M, but I don`t think it will be so much better than Maxwell efficience wise since Maxwell is already leaps ahead of previous generations.

Hopefully its much better than 880M and 8970M
About time is all I can say. Hopefully its a start of a new generation cards.

 

The Bonaire based FirePro M6100 that darkydark has performs similar to the Maxwell based 860M. It is about 2% faster but power consumption numbers based on notebookcheck, although they are estimates, show 75W and 60W. So the difference in performance/Watt is less than 25% even when comparing a new Maxwell GPU to a one that came out last year based on a GPU that is over a year old. Considering the 16-17 month difference between the release of Bonaire and Tonga, the new architecture, a more refined manufacturing process or newer and better one (if GloFo makes it), and the fact that Maxwell probably won't be as power efficient at the high end (due to bigger die) on 28nm, it isn't unlikely for AMD to improve the efficiency by at least 25% and that's not even taking into account to possible inclusion of HBM (which is supposed to be 3X more efficient than GDDR5). I don't see nVidia getting ahead before their 20nm Maxwell comes out, but there is still a year before that happens and AMD will have Pirate Islands based on Globalfoundries 20nm process by then.

 

Maxwell 860M is 37W.
GTX 750 Ti (Maxwell) is leaps more efficient than Bonaire (7790)
http://tpucdn.com/reviews/NVIDIA/GeForce_GTX_750_Ti/images/perfwatt.gif

They got to bring out a completely new architecture to have a chance to catch up or beat Maxwell. Since the difference is so big between Maxwell and GCN is why I`m sceptical about AMD "destroying" Maxwell.

 

37W?
Where did you get that number, maybe you are confusing it with the 850M. As for Bonaire, the FirePro M6100 uses a cut down version of it with only 768 out of 896 cores, so it is much more efficient and cannot really be compared to the desktop.

 

Nope, GTX 860M = 37W
It is that effective

http://forum.notebookreview.com/gaming-software-graphics-cards/745074-new-details-about-nvidia-s-maxwell-185.html#post9629219

 

About rebranding does anybody remember Nvidia with the 8600M GT, 9600M GT, 9650M GT, GT130M.

4 times has happened before is all I am saying although even here the clocks at least rose.

 

Not sure if that is correct though, Notebookcheck did compare Maxwell and Kepler variants of the 860M to each other and the difference of power consumption with systems that had the same CPU was only a few watts (3-5). Review Nvidia GeForce GTX 860M Maxwell vs. Kepler - NotebookCheck.net Reviews

Wikipedia seems even less accurate about these numbers. They list the the Kepler 860M at 75W and the Maxwell variant at 45W. The difference under load is 10W at best, maybe nVidia started lying about their TDP numbers (heat generation instead of power consumption?). 37W just seams impossible to me considering every source lists 40W for the 850M. 860M should be 45-50W at best considering that 750Ti which is clocked lower requires 60W.
Personally I rather believe load power consumption than fake TDP numbers.

 

Yes, i do remember to the point that i still have a laptop with a 9650m gt (ddr2), XD.

 

There were node shrinks between 8600M GT, 9600M GT and 9650M GT (from 80nm to 65nm to 55nm), so while architecturally the same, these three were not entirely pure rebrands.

 

860M is 37W. End of story.
Notebookcheck and your other "sources" dont now squat, but just guessing. I tried that earlier too with desktop cards and estimating the mobile variant, but its just impossible. Too many factors like voltage reduction, binning and what stepping they are using and so forth.

The program that calculated the 860M was using voltage tables and such to calculate it. The same program also list 750 Ti as 60W so its pretty accurate.

So AMD have a lot of work to do to match Maxwell

 

I hope they do their homework. I hate monopolies.

 

Well according to many sites they do say the Tonga series, which R9 M295X is part of, will rival Maxwell. But unlike GTX 860M, it should be a high end card. So fingers crossed it beats 880M severely

A pretty good sign is that the FirePro W8000 which are already out on the market with 1792 shaders, is replaced by another FirePro card based on Tonga, W8100. So it should be better because of the name.

Which means maybe 2000 shaders++ I think. Remains to be seen what R9 M295X have though, but I think we should atleast expect 1792 shaders, which is a lot more than current R9 M290X with 1280

 

Ahem current 7970M

 

After a bit of digging I think I know what is going to happen in June:

1) The new mobile Kaveri is announced at Computex
2) AMD will announce the newest GPU architecture, "Tonga", and showcasing both desktop GPUs and of course the R9 M295X.
3) MSI will show their new GX series called "Destroyer Pro" with both the new Kaveri mobile APUs and with the R9 M295X which is listed here

 

Well damn Cloudfire, my interest in the GX series has peaked again now that the Kaveri Mobile chip is a hybrid of steamroller and excavator...

 

Sorry but I just can't understand that. How can their power consumption be the same yet the TDP changes that much? It just doesn't make sense to me, TDP is directly related to power draw, isn't it? It's pretty clear at least thinking this way that Nvidia lied about 860M Maxwell TDP. Where did you get that 37 W from? Apparently from some vbios but which one? How can you be sure it's just not guesstimation as well?

Notebookcheck does not guess it's power consumption, they tested it. The difference in power consumption between Maxwell and Kepler 860m does not exist:
Review Nvidia GeForce GTX 860M Maxwell vs. Kepler - NotebookCheck.net Reviews

EDIT:
Also this source states that the peak power for a microprocessor is usually 1.5 times the TDP rating:
Computer Architecture: A Quantitative Approach

 

I think i remember reading somewhere that Amd lists total power consumption of a card as tdp while nvidia only lists tdp of gpu chip itself.

Dont get me wrong new maxwell 860m is great stuff but 37 for entire gpu seems too good to be true.

Also i could be wrong and mixed facts i read up long time ago.

Regardless of what team you preffer we shoul all hope that new amd mobile gpu lineup ends up awesome as it will stirr up and bring best from both nvidia and amd.

Sent from my C1905 using Tapatalk

 

The GTX860M Maxwell has a TDP of 50W.

The GTX860M Kepler has a TGP of 75W.

 

Why is it 50W Prema?

Kepler bios Tweaker says it is 37W.
Me and R3d had a discussion about this earlier and I think I found out that soldered GPUs have lower TDP than MXM versions of the same GPU despite clocks the same. I don`t know if they lower the voltage further on soldered GPUs to fit in to some specifications or VRAM is calculated with MXM but without on the soldered, but that is what I saw on the GTX 765M.
Clevo had 80W for the MXM GTX 765M while Dell (AW14) had 60W for the soldered GTX 765M.

Kepler bios tweaker reports power table the same as TDP with the Nvidia desktop cards. I checked several of them and they all match TDP. Even the higher clocked desktop versions (like EVGA OC) have slightly higher power table, so they seem to be related.

 

With the addition of some early Beema/Mullins devices thats about it, and maybe some info about their upcoming GPUs and the Globalfoundries process they are built on.

 

Kepler BIOS Tweaker is a user made software...NVIDIA's own numbers differ.

GTX765M MXM T GP is 70W
GTX765M onboard T DP is 55W

 

But the software takes voltage and such in to consideration before giving a limit right?

Whats the difference between TGP and TDP, and where did you find Nvidia saying that the GTX 860M is 50W?
I`m just curious since I was in the understanding that only desktop TDP are given.

Enough for me

If R9 M295X is much better than 880M and Nvidia is taking forever to release GTX 880MX or whatever it is called, I might switch camp. I`m tired of waiting.
I like the fact that GloFo is getting back in the game with GPU makers. Hopefully that will boost TSMC production on 20nm and down as well. Things are going painfully slow there it seems

 

TGP = Total Graphics Power-draw (GPU + RAM!?)
TDP = Thermal Design Power

(The software just shows a few vBIOS power-throttle targets, that has nothing to do with the hardware design itself.)

I can sell you a car with 6 gears but then lock 2 of them... aka GTX880M...LOL

 

I think you are right, I have also seen something that Indicated GPU only TDP on nVidia mobile cards, which is a sad thing really as it creates confusion. When Notebookcheck lists power consumption numbers though I think they use the estimated total power consumption of the GPU by just measuring total system power consumption and then subtracting the power usage of the other components or just subtracting load system power consumption with only the iGPU active.
Anyway the FirePro M6100 has a really high clocked GDDR5 memory and that probably uses a sizable margin of its 75W TDP.
I think this is also the case with Maxwell 860M, where the memory uses so much power that nVidia decided not to include it in the TDP. This is why I hope AMD uses HBM in R9-M295X. Even with 2+1 stack instead of the high end 4+1 stack, HBM will perform as fast as 256bit bus at 8000MHz (double that for 4+1) yet it will be much more power efficient.

Also how is your experience with your M6100?

 

Aha, so thats why the Clevo MXM had higher power target than Dell soldered 765M.
Since its MXM the software calculates power target based on everything, while on soldered it just looks at the core but disregard VRAM and power supply and such?

But to conclude on the 860M:
TGP = 50W (doesnt matter though, since its only soldered and not MXM)
TDP = 37W
?

Or is the TDP higher than power target (which is weird to me)?

http://forum.notebookreview.com/gaming-software-graphics-cards/745074-new-details-about-nvidia-s-maxwell-185.html#post9629219

 

Nope, that software just reads values written into the vBIOS.

In other words it shows you where the gear was locked not what the car was designed for.

NVIDIA designs the car and then they are locked as to per customer wishes.

 

But if OEMs put a limit there, it cant go over that limit now can it.
So if say Asus put a limit on 40W while MSI put it at 50W, one can`t say that the correct limit is 50W?

The 37W limit was from a Schenker notebook