Radeon R9-M295X

The R9-M295X is not yet published. No one can test it. Maybe Alienware has already cards?!
The drivers for the R9-M295X are likely to be as the driver in front...



A more comprehensive overview of the recent data ... who finds errors may keep them

 

Are the 32 rops confirmed?

 

Okay ... for you the "all unconfirmed overview version" :thumbsup:

 

June 23 is when FirePro W8100 is unveiled.
AMD FirePro W8100 spotted in the wild | VideoCardz.com
Which is also part of "Tonga" which R9 M295X is based on.
So who knows, that M295X launch is probably not far away now.

 

End of the summer/autumn I guess..?

 

AMD FirePro W8100 has no Tonga GPU:



source

translate:

> AMD FirePro W8100 has no Tonga GPU:



source

translate:

/>> AMD FirePro W8100 has no Tonga GPU:



source

translate:

> AMD FirePro W8100 has no Tonga GPU:



source

translate:

/> />

 

That is what confuse me.

Either CodeXL is wrong or Tonga is based on Hawaii.
For W8100 they can have disabled 12CUs (44CU>32CU) and kept the 512bit bus.
For R9 M295X they have disabled 12CUs to 32CU and slized memory bus in half (512bit>256bit).
The Hawaii chip is a big one, 423mm2, but that havent stopped Nvidia from making GTX 480M that was 526mm2.

But FirePro W8100 can be a different chip than M295X, with W8100 being Hawaii while M295X being Tonga which is then a different chip.

I`m thinking that the two are most likely two entirely different chips, but I keep both options open.

EDIT: Oh I didnt read your post entirely.
Its confirmed that W8100 isnt Tonga but Hawaii. Thanks.
Speculation about Tonga continues :

 

Is this M295X the 20 mn Maxwell?

 

Not sure if serious...

 

R9 M295X is made by AMD
Maxwell is made by Nvidia
R9 M295X is most likely 28nm (not confirmed)

 

My bad. I'm such a noob.

Well, is there a future 20nm Maxwell this year? :^)

Sent from my SM-N900V using Tapatalk

 

The answer to that question will most likely be NO.

I saw a chart a few days ago where the W8100 had two variants, the 2560SP Hawaii Pro (clocked lower) and the 2048SP Tonga with unknown clocks (still higher). Not sure what happened that everyone started claiming that W8100 won't have Tonga, maybe that variant was delayed.

 

A cut down hawaii castrated that far and crammed into 100w would perform worse than the M290x and not be able to reach 800mhz also the memory controller would struggle to reach those vram clocks (they are designed to work next to each other easily but not for speed)

 

R9 M295X launching in September according to several sites.
Lets hope that is true.

 

yes 3 mo from now. this is gonna be a weird release, perhaps another rebrand. well tbh at this point none of these graphics card are practical if I'd need to have dual PSU mod to go with it, although I will probably still need it if im to push cpu close to 5.0.

 

It's not going to be a rebrand. In fact a lot of things are completely new in R9-M295X. First the Tonga GPU is new and unknown. All we know so far is that it will have 32CUs and for the mobile variant 800MHz core clock, 256bit memory bus, and 4GB memory clocked at 5500MHz.
The newest thing about the GPU is it uses GCN 2.0 architecture. We still know nothing about the changes in GCN 2.0, besides the fact that it will have great efficiency improvements. We don't even know if each CU will still have 64SPs or not. ROPs are a big mystery since they are not affected by CU count, there might be 48 and even 64, no one knows. TMU count will probably stay the same at 128, it's already a lot for a mobile GPU, no reason to change it. Lets not forget other performance improvements from optimizations inside the Compute Units that come with a new architecture.
It will probably also be based on a new manufacturing process if it is based on GloFo 28nm. Not only is it more efficient than TSMC 28nm, the GPU will most likely run at lower voltages, meaning less heat/mm2, so easier to cool.
Other things to expect are the usual: better mantle support and performance improvements (similar to GCN1.0 vs GCN 1.1), TruAudio, larger cache, crossfire improvements, UVD & VCE improvements, Enduro improvements, short idle power consumption reduction (long idle is already almost 0), better turbo and similar stuff.

 

thats a nice list of things you mentioned. however at the moment somehow only AMD gpu gives me the problem with hwinfo64 fan control in my laptop freezes whenever i try to change brightness so I wanted to try nvidia for once and see how that goes, BUT, if new GPU will run cooler, then its always a good thing.

 

Hello

I know that it is a little bit early but do you think that this M295X would fit in a MSI GT70? It assume it will be the same TDP of 100W...?

I hope that it will be a really different architecture otherwise this GPU would not even perform closely to a desktop 7970...

 

Next gen is ready and have been for a while.
Come forth R9 M295X





http://videocardz.com/50947/amds-radeon-r9-290x-hawaii-xt-gpu-might-full-chip

 

SinofLiberty on another forum? LOL!

 

Well that dude has at least 5K+ posts on that forum, Sin only made a couple hundred before run out of town here.

 

8 pack is a staff member at ocuk and their resident overclocker.

 

Overclockers UK - 8Pack Systems
The Man Behind The UK’s Number 1 Overclocker: Exclusive Interview With Ian ’8Pack’ Parry | KitGuru

Should be pretty reliable. He builds and test desktops which overclockersUK sell. He have probably already tested the chips out since he talks about NDA and that they have been finished for quite some time.
We already had a leak from ozone3d where a Clevo machine ran R9 M295X, so Im guessing that was an OEM testing the chip out for coming models.

Not long now until we see it all unveil.
Tonga as in R9 M295X is probably what he is referring to as "Next gen"

 

I know, I was jk. Sheesh.

 

I think the R9 M295X will have lower TDP than GTX 780M.

OEMs run the GPU Caps Viewer which according to the creator:

i7 4700MQ + GTX 780M: 82C with 99% GPU load. Fanspeed unknown.
i7 3930K + R9 M295X: 82C with 99% GPU load.

Clevo machine: 130W TDP CPU + xxxW TDP GPU
Unknown brand: 47W TDP CPU + 100W TDP GPU

I think the Clevo machine will run hotter than the other one, and since they reach same temp on the same benchmark, I`m thinking the R9 M295X have lower TDP than 780M.
Some unknowns here but if they tried to stress out the GPUs and the temp reported is max (since both are at 100% GPU load), that shouldnt be far from the truth.

 

I think the reason for this is that nVidia usually rates TDP at lower values than AMD does at the same amount of power consumed. Even with a 3X more heat created in the system with the Desktop CPU, it still runs at the same temps as the 780M. So TDP must be about 100W.

 

Those are for the desktop parts I believe, at least the Hawaii part, there are rumors that the next gen high end part might be an answer to the GTX 880 with new GCN 2.0 core and HBM memory.

 

I dont understand what you mean here.

Heat from CPU shared with heat from GPU > GPU Temp
47W shared with heat from GPU > 83C
130W shared with heat from GPU > 83C
If both cooling systems are equally good. How does this means same TDP for GPU?

R9 M295X is rumored GCN 2.0.
R9 M290X is GCN 1.0.
Hawaii is GCN 1.1

GCN 2.0 = Next Gen
Remember that Tonga is not just R9 M295X but also some desktop chips as well. If desktop = Next Gen, then so will the mobile chip based on it

 

For the first part what I meant is even if M295X is 100W it probably runs cooler than 780M because of the differences the way AMD rates TDP vs nVidia (similar to 8970M vs 780M).
The reason it probably runs cooler is because the 130W CPU dumps more heat into the system.

For the second part it's true that Tonga is GCN 2.0, but thats not really High end on the desktop. And besides Tonga will launch within 3 months anyways.

 

That's neither a Clevo nor even a Notebook...

 

Unless that GPU Cap Viewer stresses the CPU to the TDP limit as well, one simply cannot draw any conclusions about how much heat the respective CPUs are dumping. In any case, even if the CPUs are indeed consuming different amounts of power, as long as the cooling pathways for CPU and GPU are not thermally connected (as is the case with Alienware 17/18, real Clevo's etc..) the GPU temperature should not be affected appreciably. The effect of CPU temperature on GPU temperature is really just a second-order effect, if not more negligible.

 

What gives it away as not a p570wm?

 

A Clevo P570WM has subdevice_ID: 1558 - 0270

EDIT:

Though on a second thought they may encode the HWIDs.

If you want to know for sure submit a GPU CAPS Viewer result with your own P570WM and see if ID matches that result.

EDIT 2:

They indeed convert the hex to decimal for the online display.

BUT, a P570WM would show as:

subvendor_id 5464
subdevice_id 624 (or 625 for the 3D model)

That system on the other hand has:

subvendor_id 4098 (translates HWID of 1002)
subdevice_id 2816 (translates to a motherboard ID of B00)

The HWID of that card is: 6938

A Wimbledon XT M290X has a HWID of 6801
An Amethyst XT M295X has a HWID of 6921

EDIT 3: In English: That's not a Clevo, that's not a Notebook, that's not a M295X.

 

well R9 M295X with haswell-E since both are out in a few months, put them in clevo mmmmmm yeah anyone? i'd be down for that, or I can continue to chase after 5.0 with my ivy with HT disabled.

always welcome lower TDP on GPU, lower TDP and lower power usage for sure. 20nm is where its at, maybe.

 

Subvendor ID: 5456 brings in these results

GTX 765M
GTX 680M
GT 650M

All notebooks.


The reason why I thought it would be a Clevo is because its the only brand that use desktop CPUs. I`m not sure what the "Subvendor ID" means and what "Subdevice ID", but Subvendor ID looks to be notebook and Subdevice ID looks to be desktop since I can find many desktop cards tested with Subdevice ID 2816.

So perhaps its a middle ground here. A mobile graphic card (R9 M295X) tested in a desktop, ie for testing purposes to validate that it works like it should?

 

Hello

According to its spec, the M295X would not even perform closely to a desktop 7970.. which is almost 3 years old. Do you think that the fact that it is a newer architecture can make it more powerful, for the same amount of shader and frequency?

 

7970; 2048 shaders @ 925MHz
R9 M295X: 2048 shaders @ 800MHz (Still unconfirmed if it have a turbo boost over 800MHz).

If its GCN 2.0, there could be some tweaks involved that makes the M295X more powerful than it seems yes. But I think they will mostly be power and thermal gains (= less power consumption and heat).
But we will see

 

Yes but the memory bandwith will surely leads to a significant drop in power, doesn't it? As well as the lower memory speed..

 

Not if they cut away the bandwidth slurping ROPs and/or neutured its GPGPU performance (Less FP64 cores). Maybe they added more L2 cache (more than 7970s 768KB) to help with bandwidth with the GCN 2.0. Nvidia already started the trend with 2MB L2 cache on a low end GPU with Maxwell so AMD might follow.

 

That's what I said. It would have to have that ID to be 1558 = Clevo, but it doesn't have 54 64.

Since the Clevo is the only DTR and it is using a desktop CPU, it is desktop.
Since the GPU also has the wrong HWID it is not a M295x.
The HWID it shows for the card is a yet unbranded R200 Series desktop card.

A 'proper bluff' with Clevo P570WM and M295x would have to look like this:

vendor_id 4098
device_id 26913
subvendor_id 5464
subdevice_id 624

 

It doesn't have to be Clevo, maybe it's a Eurocom, or do they still use clevo designed chassis? Even if they do not sure the GPU ID would be the same.

 

That unbranded card is the upcoming Tonga. And if its not R9 M295X, its the desktop card its based on but with R9 M295X clocks. Pretty much a R9 M295X.
I don`t know how OEMs and AMD work with cards, but previous official slides from AMD was results presented with desktop CPU and 7970M running together. So nothing new here Prema. The ozone3d.net data was perhaps AMD just testing the chip out maybe.

AnandTech Portal | AMD Launches Radeon 7700M, 7800M, and 7900M Mobile GPUs

 

I really hope this card totally blows the 880M out of the water. Maxwell too.

GPU prices been getting super high lately. This might just push them down if it can achieve superiority..

 

In a perfect world each side would be in competitive balance

 

Oh boy this is interesting. Lets just say that a well known leaker have spoken about R9 M295X:

R9 M295X 3DMark11 score (total) 10300
GTX 880M 3DMark11 score is 8300.
Which means that M295X will be undoubtly the fastest out there, beating GTX 880M by almost 25%.

Coming in notebooks in September

The leaker also said that the price will be cheaper than GTX 880M

 

Well that's good news... After the crap NVIDIA has pulled with the 880M, this would probably be a better upgrade then the 880M unless the 880MX is much much better then this..

 

Wow thats insane... I hope we can pair it with an Intel CPU...

Sent from my Nexus 7 using Tapatalk

 

Why would you not be able to? I have yet to see one you can't.

 

Way better upgrade than 880M.
60% more shaders than R9 M290X (if 64 shaders per CU is still being used with Tonga). Take 680M and the 880M (2 years apart). 14% more cores, that was all Nvidia could offer. Its pathetic. 680M was the only GPU worth investing in with Kepler. Rest are just mediocre upgrades. I will hardly call them upgrades because its the same chip (GK104) being reused and reused by a ton of different mobile chips. Pathetic!

I`m sure if the P10300 in 3DMark11 is legit (which I believe it is since the leaker is well know), AMD have again put pressure on Nvidia to release the GM204 Maxwell. But R9 M295X looking to be a beast anyway, beating 880M by 20-25%. So it will sell like crazy I`m sure

 

Not from a technicality standpoint, just from brands that offer it. For the high-end AMD mobile GPU most folks associate it with the MSI GX series. Alienware over priced, Clevo/Sager not as well known/branded outside notebook enthusiast community.