Radeon R9-M295X

Shader counts, clocks and architecture are still up in the air.

 

Well thank goodness! It used to be a fatal disease.

 

Yeap, same as my 7970m. 1040 on 1.05v STABLE!!! not only for 5 min benchmarks. And 1090 on 1.1 also stable for gaming. Dear "AMD" I am very disappointed with this R9-M295x description ...
How we like to say in Poland "hope, the mother of fools".
I'm terribly sad, so terribly saaaaaaad...

 

AMD hasn't given any specifications of the R9 M295X for people to be disappointed or anything, what cloud posted wasn't real numbers it was just random guesswork.

 

R9 285 ITX Compact Edition

Also, OC Edition, 10 MHz

 

Good thing they made that. This means the form factor could handle the GPU as is on a notebook. If the 720mhz clocks are true, we could even see a full 285x downvolted as the M295X. At least we know 285 can fit easily now.

 

What gives you this idea?

 

back when 680m/780m were rumored to come out, things started to become real when the actual size of the card was released (Desktop GTX670). It gives hope and makes you realize that they can cram down the GPU to the notebook size easily, but cooling will be another matter.

Also, 7970m can run full clocks as a 7870 desktop, and that is supposed to be a 170W card. I am sure they can squeeze a 285 easily, or even a downvolted and downclocked 285x a la 780m.

For example, the GTX670 was also 17cm long, At least the PCB. http://www.anandtech.com/show/5818/nvidia-geforce-gtx-670-review-feat-evga/2

 

The physical size of the card means little in this context. They cut down the memory interface on the 285 plus there is a lot of wasted board space on the mid-range cards so it's not hard to shrink the PCB down.

But look at the power requirements. 270 is a 150W card and only needs a single 6-pin PCIe connector. 285 is 190W and needs two 6-pin PCIe connectors. In contrast, 750 Ti doesn't require a single power connector as it's so efficient that it draws all the power it needs from the PCIe slot itself.

Since there looks to be little if any efficiency gain with Tonga, if M295X is clocked low, it won't perform appreciably faster than a Pitcairn chip, especially an overclocked one. And if it is clocked high, it'll be another hot and power-hungry card like the 880M.

 

This seems to be the informed consensus, from cloudfires posts too, it seems like NVidia is likely going to be leading the charge with the next best mobile chip. As a real brief synopsis, I personally think that Kepler has been a real win of an architecture, the early adopters of the 680M have been laughing all the way to the bank for the past 2 years, especially if they were willing to flash svl7's modified vBIOS. 7970M has been a somewhat worthy contender, but overclocked not quite as well. I truly want AMD to bring some fire to the table, but it just sounds like false hopes if I'm gonna take the bleak approach! C'mon AMD, kick some a**, make those NVidia chips less pricey! (I feel AMD abuse winging my way potentially!)

EDIT: to create some balance: 880M sucks and you shouldn't buy one if you have a brain! (get a 780M if you can and overclock it, or better yet just wait till high performance Maxwell - like everybody who's waiting SO impatiently!)

 

*Puts on tinfoil hat*

 

Indeed, the 7970M is still looking pretty good next to the alleged M295X. A good analogy is how closely 7870 and 7950 perform stock-for-stock. Even though 7950 has 8 more CU's, a higher memory clock and 50% wider memory bus, it is often within 10% of 7870. This tells us one, that 7950 with its 1792 shaders has an overspecced back end, which is a good thing because it means that the 256-bit bus on M295X won't hurt it much. But two, it shows how important a high core clock is for M295X to pull away from 7970M. If M295X releases with <800 MHz and little OC headroom, it will barely be better than a 1 GHz OC'ed 7970M.

 

Yep, let's bring on the next true next gen GPUs!

 

"Next-gen" is such a trite and clichéd hence meaningless marketing buzzword at this point *glares at PS4 and Xbone*. Just give us something that's significantly faster and better than what we have right now.

 

Indeed, faster & better - bring it on! NVidia & AMD have been dragging on this recently, no real significant in-roads since 7970M/680M in April 2012!

 

It technically makes sense for consoles but it is very much overused and in the wrong way.

 

There was a reason why the original 7950 wasn't much faster than the 7870GHz, besides the low clocks it had the same number of ROPs which meant that it's ROP performance was very limited, and much worse than 7870 due to lower clocks, so in games that required high ROP performance the 7950 fell behind the 7870. Now the 285 only has 32 ROPs but considering its performance compared to even a 280X it is impressive that it can keep up with lower clocks. AMD must have improved the ROP performance of Tonga compared to Tahiti, maybe the doubleing of Geometry engines has something to do with it, but there is still a lack of data so we can't really tell much. Once we see other benchmarks it will become clear.
So if Tonga has an improved ROP performance it will be a sizeable improvement from R9 290X. It will also have a better clock boost system and even if those leaked clock rates are accurate they might just be base clocks. As for overclocking who knows how it will perform, we don't even know which foundry is making the GPU.
Personally though I can care less about Overclocking mobile GPUs, it has more negative consequences than benefits, especially when it comes to warranties.

 

Overclocking has no implications on warranty since nobody can prove you overclocked unless you flashed it into vBIOS. Maybe it doesn't make as much sense at the high-end since the cards are running at or close to their limits anyway and there is little OC headroom, but for the mid-range, it's a way to buy a cheaper card and increase performance 30-40% for free to match a more expensive higher-end card with little to no ill effects.

 

GF + FD-SOI, but that's just wishful thinking

 

Arghh, I'm sure if I was following AMD architecture & the like this would make sense! There's too many variables for me to follow currently in the AMD architecture for this to make sense to me - I may be an outlier!

 

Lets just wait and see they haven't even properly announced the m295x yet!

We may all be pleasantly surprised.... Or not.... Either way no point getting too down when we don't know what is coming

 

Its taking forever...
No leaks, no news, nothing Bah I`m getting sick of waiting for AMD and Nvidia :/

I guess we might have to wait til September 2nd until NDA ends

 

There is an interesting article about Tonga and some of its architectural improvements by Xbit Labs. Here is the article: AMD Quietly Reveals Third Iteration of GCN Architecture with Tonga GPU - X-bit labs
It seems like the lower memory bandwidth may not be an issue, and there are tessellation improvements as well.

 

Memory bandwidth is a HUGE issue bro... in all of the 15 games or so that make it an issue.

If you're doing stuff like rendering and you need to make use of that vRAM, it's gonna make a little chunk of a difference. If you're just gaming, you probably won't notice. However, AMD has had the slower memory bandwidth for their entire 7000 series & Rx 2xx series chips as far as I know. I think going BACK even further on memory bandwidth to push the core more is kind of stupid. Eventually games are going to come out which happily fill those frame buffers, and we've seen quite a few of them recently. Most of them are of course unoptimized and could have looked better with less frame buffer use, but it doesn't change the fact that more vRAM will become a benefit soon enough (2GB won't be the sweet spot anymore) and the ability to fill that vRAM quickly may definitely be an advantage.

 

That's not true. All of the GCN cards so far have had wider buses and more memory bandwidth than their respective Kepler counterparts. And like I said before, I doubt the reduced memory interface is gonna hurt Tonga Pro as Tahiti Pro's 384-bit bus was overspecced for its 1792 shaders. With Tonga XT (2048 shaders) it might be more of an issue though.

 

I stand corrected.
They have had wider buses but slower memory clocks only sometimes. The R9 290X with 512-bit mem bus and its 1300MHz mem clock only come out to 320GB/s bandwidth whereas the 780Ti and Titan Blacks cards have 384-bit, 1750MHz memory configs, which set their bandwidth at 336GB/s.

The R9 280X has 6000 effective on a 384-bit mem bus though, so that's a lot. I thought I always saw them with lower configs, but when I just looked for some on newegg they showed up as 1500MHz. Oh well, nice there AMD. Don't know why 290X was so low though. The 780 and Titan match this. Man they could have really ripped nVidia a new one if they set their R9 290X to so much as a 1500MHz mem clock.

And now I get to speak less stupid in the world =D

 

Except 290X's price/performance counterpart is 780 @ 288.4 GB/s, not 780 Ti.

 

Good read. Who knows, maybe we are in for a surprise. Still sceptical but we will see

 

Less than 3 days left and we still don't have any useful information, the fact that AMD has managed to keep things quiet at this late stage is just ridiculous.

 

Actually give credit to AMD in this day and age with so many leaks everywhere it's quite impressive.

 

Absolutely! The wait seems never ending! 2 days at this point = eternity!

 

any idea what m295x will cost alone? very excided to exchange my 7970m

 

Around 500 dollars probably

 

AMD Roy had a little discussion earlier this week and they did bring up the R9 285.
He said it draw around 85W less than the 280/280X. Which will put it around 120W if it was 280X he was rfferring to. You may also notice from the chart that the 7870 draw 120W for the same test, so I`m thinking there are chances that the R9 M295X will have the same TDP as 7970M and not like 880M which was previously discussed.

@14:34 here
https://www.youtube.com/watch?v=dTB2Uk43LKU

 

Okay, so i've been on this thread for idk how long. but what is happening the 2nd september?

 

Desktop Tonga is launching (same GPU in R9 M295X) and more importantly NDA will be lifted, so we will finally get an idea of how it performs in real world cases and more importantly how efficient it will be (hoping someone will try to do some undervolting but that is a bit too much to expect from a desktop GPU review).

 

Then i don´t understand a thing, why do they list it as 190W TDP if that is true?

 

LOL how ironic.

 

They state "board power" in the specifications. It should be the TDP but who knows if there is some other mumbo jambo baked in to it.

 

Tomorrow people! At least it is Monday morning where I live

 

Finally, 24 hours left.

 

I don't know why, but I have a feeling that so far there has been no leaks about new mobile card from AMD simply because they are going to announce nothing in the mobile sector tomorrow. Just my $0.02.

 

I have that bad feeling too. But soon we will know.

 

The GTX 980M coming a month after could have changed their plans...

 

No matter we hear about the 285X that is enough.

 

Same here.. There has been 0 info about this.. I'm expecting nothing great here...

I don't think so.. AMD has no excuses.. Its been like 2.5 years now.. They really need to stop rebranding and release something new to beat NVIDIA...

 

To do that, they have to know what nVIDIA have on their hands. That's after a month, so I guess we'll see the mobile then, while tomorrow we'll have the detailed specs of 285 and 285X to base our guesses on

 

now what? something on twitch? man im confused (as usuall) ^^

 

Sapphire R9 285 ITX is on sale already, but where are the reviews? AMD is probably keeping the NDA up until the end of the Launch event.

ShopBLT.com: Sapphire Technology Radeon R9 285 2gb Oc Itx Compact Hdmi / Dvi-I / Dual Mini Dp

 

10 incoming on 09-21-2014...

So no need to hurry and we can all watch the NVIDIA next-generation high-end GPU launch on 09-18-2014.

NVIDIA Announces Game24, Massive 24-hour Celebration of PC Gaming | techPowerUp