Radeon R9-M295X

This is the original quote..

 

lol you guys.

When I said R9 M295X CF would be cooler than GTX 880M I was obviously talking about M295X CF vs 880M SLI.
And why does it really matter? As long as the heatsinks and cooling system works like they should in a notebook, a M295X CF could produce less heat than a system with single 880M. The GPU itself will be cooler, and any heat from the GPUs are mostly pushed out of the notebook anyway.
All in all, the M295X CF will probably expel more heat out in the room than a single 880M, but in the bigger picture, does anyone really care?

Extra wonderful heat in the winter and more heat produced in a world thats getting hotter and hotter each year anyway thanks to the global warming. I don`t think SLI and Crossfire configurations is the cause of that crisis. Unless the scientist have been wrong all along.

 

HAHA, information about R9 M295X was leaked today.

The information is a little bit hidden but what is certain is:

R9 M295X have 4GB GDDR5 running @ 1375MHz
R9 M290X have 2GB GDDR5 running @ 1250MHz

R9 M295X have a core base clock is 800MHz.
R9 M290X have a core base clock at 850MHz

The R9 M295X have been tested with OpenGL and according to the entries, it have many changes from R9 M290X which makes me believe this is a new architecture. But I`m not sure.
Does anyone know how to read these differences and translate them to what have changed?

R9 M295X @ ozone3d.net
R9 M290X @ ozone3d.net (8970M = R9 M290X)

Just a little bit of what was leaked:

 

32 compute units is the same number as the 7970.

Of course how many shaders per unit is the question.

 

Everything listed under " OpenGL capabilities" is the same across all AMD graphic cards from previous 2 years. R9 290X, 8970M, R7 260X etc. I have checked.
The stuff above have changed. "3.0 core" etc whatever that means. They gotta mean something but that is way over my head

 

Cloud, look at the bottom, I compared the 7870, 7970 and the one you have posted before posting.

CL_Device_max_compute_units

Or are you actually not asking for help from someone who has a better understanding of this information?

 

Thank you Meaker. I didnt notice that entry.

R9 M295X have 32 Compute Units while R9 M290X have 20 Compute Units
Holy crap!





R9 M295X will have the same shader count as 7970: 2048
A massive boost from R9 M290X with only 1280 shaders

!!!

 

Specifications of upcoming R9 M295X (Codename Tonga):

2048 cores @ 800MHz
4GB GDDR5 @1375MHz (256bit bus)

:thumbsup:

 

You mean a 7970, not a 7970m, that is amazing, 7970 performance in a laptop, not bad at all, quite a big jump

 

Those stuff that did change are OpenGL library extensions (probably programmed by AMD), there are 29 new ones and probably a few version updates. I didn't spend much time checking those but all I found was a lot of "String" and a bunch of code templates and not much code. Maybe those links are just descriptions of the functions of library extensions.
I am not familiar with OpenGL, or any other GPU languages. I have only programmed in Java, C++, and Python (nothing too advanced).

 

Oh this bad boy is gonna spank the GTX 880M for sure

We are talking 60% increase in cores vs R9 M290X.

Thanks. I almost came to the conclusion that 4.2, 4.3, 3.1, 4.4 are versions of OpenGL it supports on the various functions, but I`m not sure.

Yes correct. A typo there

 

If this card will be compatible with Clevo 170em, I'll buy it. F**K 780gtx MOAHAHA!!!

 

Have been doing some maths with those 800MHz should have the performance of a 7950 boost or r9 280

 

I think that memory interface will be 256-bit wide, not 384 as in desktop 7970 or 7950, so performance of Tonga will be a little lower, unless there will be some cosmic architecture improvements. I also hope for P170EM compatibility or for Prema to work his magic and make it compatible, for the sake of more donations and for us to be able to upgrade.

 

I'm sure if we paid him an hourly wage it could be arranged.

 

If MSI do a version with the new APU that came out I'd be interested. From what I have read, it gives Haswell a good run for its money, and of course is cheaper than Intel.

 

Nice I just saw the Compute Unit numbers, not sure if it's accurate or not for the M295X, but it sure is for the M290X, and if it is then it's a great news.
Honestly I was expecting Tonga to have 22-30CUs with very high clock speeds (for the desktop version, at least 1100MHz stock). For the mobile variant I was expecting at least 22CUs (Half a Hawaii core (new GPU though), with half the memory controller and everything else), 28CUs being a more probable number (same as a 7950, since that's its replacement, but with better performance at the same clock due to new architecture), with 950-1050MHz clocks speeds for a smaller core (like 22CU) or 900-950MHz for for bigger ones (28CUs). Anyway I will try to see if I can find any texture unit numbers, as that will give away the CU count.

 

Dual Core Haswell yes, just don't expect it to be near the 47W+ Quad Core desktop chips in mobile socket packaging.

 

My quick calculation, probably inaccurate as hell, says R9 M295X will be 20-25% faster than GTX 880M.

It was dead on accurate on the 8970M from the same site and test.

Very true. Intel quads are much better. If kaveri can just operate without bottlenecking the GPU, it is good enough for most I think. If so then GX notebooks with R9 M295X will be good value

 

As I said they could change the balance of a compute unit, so we are not 100% sure of the shader count.

It will be 256bit at most due to the limitations of MXM-B.

 

Not very likely, as GCN is still a new architecture and 64SP per CU is the sweet-spot (at least for now).

 

Assuming that the SP design is still the same.

 

Congrats Cloudfire!

" AMD Radeon R9 M295X specifications
Some fresh leaks have been discovered by one of our most dedicated readers, Cloudfire. What he discovered are the specs of the upcoming Tonga GPU, which as you know is being slowly being prepared by AMD. Tonga is based on the same GCN architecture known from Hawaii or Bonaire. Tonga is actually a Tahiti replacement, but since it’s much more power efficient, AMD has decided to replace Pitcairn/Curacao/Neptune with something fresh."

You were mentioned in a Videocardz article. AMD Radeon R9 M295X with Tonga GPU has 32 Compute Units | VideoCardz.com

 

thats gonna be amazing.

please msi release a gx destroyer version of this for 1300$. hopefully they'll have a cpu better than the garbage 5750m by now, it's been over a year.

 

CPU would be certainly better, but still there would be a bottleneck, even more, the bottleneck would be probably be even bigger because that m295x is 50-60% more powerful than m290x whereas fx 7600p is not that powerful compared to a 5750m, at most it would be a 30% better than 5750m

 

That's lame, amd mobile cpus really are atrocious.

I wonder how much the m295x will cost, if it costs as much as the 870m then nvidia is finished. The 880m is stupidly overpriced anyway.

 

Price would likely be similar to the 7970M launch. It's pretty typical for amd to have the lead early in a generation.

 

Hasn't the m290x's price been stagnant for 3 years?

So another GX destroyer trainwreck would be 1300$ while an i7+m295x be in the 1800 dollar range? Was initially going to get a y50 next month for 900 euro, but if I can get an i7 plus that one at less than 1500 euro it might be cool, since it looks like it'll be twice as strong as the 860m.

 

I hope this comes out, I'd def get two. I know that in order to install after market nvida cards you need to modded the inf on the driver, would it be the same for amd after market cards?

Sent from my SM-N900T using Tapatalk

 

I am a big fan of cheap.

 

Urgh their naming schemes are terrible. So when it's a rebrand it deserves a new series generational name but when it's a completely different architecture it stays in the same series? Why AMD? Why NVIDIA? Whyyy?

 

I do not know whether that's for sure, because:

mobile GTX 880M = desktop GTX 680 ...1536 shader ... -52MHz base GPU clock ( -5,2%) ... -250MHz MEM clock (192GB to 160GB bandwith is -16,7%)

If we use the desktop HD 7970 to gross comparison

mobile R9-M295X ≅ desktop HD 7970 ... 2048 shader ... -125MHz base? GPU clock ( -13,5%) ... -128Bit MEM interface (264GB to 176GB bandwith is -33,3% )








image source

 

Finally, a new core. Time to upgrade.

 

@SFVogt:
Here is my rough calculation:
R9 M290X: 1280 shaders @ 900MHz = 1152000
R9 M295X: 2048 shaders @ 800MHz = 1638400
1638400/1152000 = +42%

According to games tested at Notebookcheck, GTX 880M is about 20% faster than R9 M290X:

R9 M290X: 100FPS in a game
GTX 880M: 120FPS in the same game
R9 M295X: 142FPS

142/120 = 1.18 = 18% better.

This is if AMD doesnt nerf the core by using too little memory bandwidth and if M295X is the same architecture as M290X.

@Link4:
Thanks. I shouldnt get credit for it though. The dude who tested the graphic card and published it intentionally or by mistake should.
I noticed that he tested the R9 M295X with a i7-3930K so I think that this was benched with one of those Clevo notebooks with desktop CPU.

 

Sweclockers just shared some news today regarding Tonga from AMD:

They say the desktop chip (which mobile will be made of) will be R9 280X performance but the power consumption will be similar to R9 270X.
R9 280X = 185W
R9 270X = 119W
Source: http://tpucdn.com/reviews/MSI/R9_280X_Gaming_6_GB/images/power_average.gif

7870 which R9 M290X was based on had a power consumption of 111W so this is why they can build a new mobile GPU which have many more cores than R9 M290X

Launch of Tonga will be in Q3, August or September according to Sweclockers.

 

That means is around 50% more energy efficient, not that far from Maxwell

 

would be impressive with a more hungry memory bus.

 

36% more efficient if the numbers are correct.

Not exactly as much as Maxwell (50%) but very impressive anyway. What Im curious about is this is thanks to a new architecture or production advancements in the 28nm process.

 

A very optimistic estimate. 42% theoretically more GPU power (linear increase) but only 10% more memory bandwidth. A visible imbalance. There are probably to be expected about ~30% in real terms. Notebookcheck.net writes " Nvidia's premium model defeats the "more reasonably" priced Radeon R9 M290X by approximately 25%."

my rough calculation: 130/125 = 1.04 = 4% better

I think it will be a close thing. But AMD should win clear the situation actually. The GK104 (GTX 880M) is now over 2 years old. The GTX 880M performance is known to all since march 2014. :thumbsup:

 

Depends on the situation but the 7970M will scale almost linearly with memory and core clocks so it should be almost a direct increase as much like maxwell they are likely squeezing more efficiency out the mem bus with larger caches too.

 

I have a question.

If R9 M295X comes out, would it work on Alienware M18x R2?

Also, don't AMD have inconsistent drivers? That would bottleneck, no?

Sent from my SM-N900V using Tapatalk

 

Fortunately the shader count isn't the only thing to effect the performance. If I remember correctly, 7950 while having having more GFLOP processing power, lost in some benchmarks to 7870, because it had less processing power from the ROPs side.

There was a rumor a couple of weeks ago, that a high-end card coming this year would be using High-Bandwidth-Memory, but because we got a rumor about XDR2 before the launch of 7000-series, I don't think we will see HBM soon.

 

Nope, 7950 smashes 7870. It has a stronger core config and stronger memory back end.

 

We will see. I don`t think AMD would put out a GPU that are memory bottlenecked. It happens on some Nvidia GPUs but not on the flagship, and I think AMD feels the same about M295X.
Like Meaker says, could be more than just memory clocks in the picture. L2 cache, efficiency (Maxwell 750 Ti with 86GB/s matched a 144GB/s Kepler), ROPs etc

Interesting. It still did fine in gaming despite lacking ROPs?
EDIT: Noticed that both had 32ROPs.
Thats one way to elevate the need for high bandwidth maybe?

 

Yes may be, see:

HD 7870 = 1.000 MHz @ 32 ROPs = 32.000 MPix/s
HD 7950 = _.800 MHz @ 32 ROPS = 25.600 MPix/s




image source

 

Fillrate = ROPs*MHz. More fillrate doesn't automatically mean it's faster. You're not accounting for all the other factors in the card specs. 7870 has the highest fillrate in that chart, but is it the fastest card? No, of course not. As a matter of fact, it's the slowest one.

 

The 7870GHz did beat the original 7950 in a few cases, and even though they had the same number of ROPs, the ones on the 7870 were much higher clocked. The case was even worse for the FirePro cards. The Pitcairn powered FirePro W7000 destroyed everything on the Professional GPU market with the exception of the W9000(Tahiti XT). That includes the Tahiti Pro W8000, the 7870GHz (oddly the FirePro drivers did much better for the W7000) and everything Quadro nVidia had at the time.

 

But has the 7870 a higher ROP power as the 7950? Yes nothing else i have written

 

My tech bros. Can you guys answer my question? (Sorry, I'm a bit excited about this new card)

Sent from my SM-N900V using Tapatalk

 

You would have expected them to work hard on any major bottlenecks, if the ROPs were holding it back then they will have been worked on.