Radeon R9-M295X

Lol, all I really want to see is if it will be powerful enough to MAX out everything at 1080P (the most suitable resolution for a 15.6'' laptop i believe, because anything more and text becomes too small to read) which still didn't happen right?

 

It's like with the CPU Turbo. They Turbo for a certain period of time and within a certain range before being throttled.

The throttle on NVIDIA cards is much more complex than just looking at a single value. That specific number you are pointing at is more like a 'fall back' target. It all depends on many factors, temperature, voltage, power draw...

 

So lets say the new MSI is announced at this conference, how long would it take for that to be available retail? I only ask because I am not very savvy to announcement and release's in this arena. I am literally about to order a new Clevo with 880m to replace my tired Aspire 5560G-7809.

 

I don't see how that changes what the TDP is. Yes, a CPU (eg. an i7-4900mq) can turbo past it's 47W TDP, let's say to 55W or something. But the TDP is still considered to be 47W.

Regardless, according to Nvidia's SMI documentation, the power limit is:
"The power management algorithm's power ceiling, in watts. Total board power draw is manipulated by the power management algorithm such that it stays under this value."

And that's the value in the VBIOS. KBT could be reading Maxwell vbios wrong, though.

On a different note, even if Tonga is 50% more efficient than Pitcairn, that would put the R9-M295x as barely faster than the 880m, since the 880m is like ~30-35% faster than the 8970m to begin with (assuming that the TDP stays the same).

 

TDP is not the actual power draw. It's a rating for OEMs. It's used for things like heatsinks. It stands for Thermal Design Power.

So when a 880M is a 100W TDP card, it is not referring to the cards actual power usage.

A rough comparison of TDP vs power consumption:

A Intel Core i7 3940XM processor has a 55W TDP.

For intel, 0.122 x TDP = continuous amps @ 12v. So 55w TDP x 0.122 = 6.71A. Now to get watts, 6.71A x 12v = 80.52Watts.

I got the 0.122 x TDP from here

Hopefully I did all of that correctly.

 

Great card. Much cooler than m8900 and same temps as quadro 4000m while having great performance. All games that i tried run fine. Most important thing i can use duplicate monitors option without issues - where on nvidia it was only working when card or drivers felt like it.

Funny thin is that in my experience nVidia cards had more quirks than amd but then again ive never owned amd high end cards which were the ones with issues - compared with certain nvidia mid series that ate my nerves.

Sent from my C1905 using Tapatalk

 

On paper, TDP is supposed to be for heat, not power draw. But both AMD(search for TDP) and Nvidia use (or have used) power draw as their definition. So it's a little fuzzy.

Anyway, this is off-topic, so let's keep the discussion about the R9-M295x.

 

The big questions for me are:
1. Is it going to be MXM3.0B?
2. Is it going to be offered as an option by Clevo?
3. Is it going to be possible to "plug&play" upgrade my P170EM?
4. If answer to #4 is "no" then is Prema going to make it possible?

 

I believe we understand that, the point is TDP is directly related to power draw, right? They are not the same thing yet they are strongly correlated: the more power it draws the more heat it'll generate.

How can a Kepler at TDP 75W and a Maxwell at TDP 50W draw basically the same amount of power? From my possibly incorrect understanding either these TDP numbers don't mean squat or Nvidia lied. TDP is directly related to power draw, if it's power draw is the same how can it's TDP be this much off?

EDIT:
@Marecki_clf
Regarding nº1, possibly MXM 3.1, apparently it's been around since 2012 already: http://www.mxm-sig.org/ I have no concrete idea though.

 

3.1 is the same as 3.0 it just adds support for PCI 3.0 speeds. Not yet necessary unless you run desktop cards in SLI. I doubt even dual 880Ms can bottleneck PCI 2.0. Maybe just if overclocked?

 

For a silicon chip:

Power in = heat out.

 

I found no conclusion about TDP and Kepler bios tweaker screenshots I posted.
No one is able to give a direct answer here. Not sure if people are unsure about what they know is correct or not.

1) Does there exist any data sheet or official information about mobile graphic cards and TDP? Yes, where? No, how can some say that GTX 860M is 50W then? Again, I think Nvidia only give out TDP about desktop cards to the public, while letting OEMs decide TDP based on clocks and type of VRAM on the mobile, and neither them or Nvidia have said anything to the public about current Maxwell cards. Kepler bios tweaker can find the particular power limit through looking at voltage tables and such and is pretty accurate regarding TDP from that particular chip.
2) The power target on Kepler bios tweaker, how can TDP be higher than 37W when the vbios throttles the GPU beyond that power draw? 37W is listed as "100%" meaning maximum, TDP is worst case scenario for a chip = It can never be higher than peak power consumption (=Power target 100%)

 

1. I gave you the correct numbers...for obvious reasons you will simply have to trust me on this.

2. As said before that number is nothing more than a fall back value or a base line if you want. The card can go beyond that value, which again is set by many more values.

 

Yes, we are on the third implementation of turbo now on Nvidia chips I think, each iteration has added more complex monitoring of current stats to define the envelope it is operating in.

Also be aware that 37W may be the TDP of just the GPU while 50W could be for the whole "card" if you will. That's another important definition.

 

See, this is the problem and why I say I have got no conclusion on this all. What obvious reasons is it that TDP (look below what I wrote to Meaker, you say it is TGP) is 50W? I want to know where you have seen that the TDP is 50W. Or what makes it 50W.

The vbios say 37W is 100% but have another max limit called "233%" which is 100W. Are you saying that the TDP is usually higher than 100% because its worse case scenario and the vbios allow it to go all higher than 37W because there is another entry there?

Exactly what I said earlier but according to Prema 50W is TGP and not TDP. http://forum.notebookreview.com/gaming-software-graphics-cards/747646-radeon-r9-m295x-5.html#post9662204

Extremely confusing

 

Your are mixing TGP & TDP

Read it again:

http://forum.notebookreview.com/gam...rds/747646-radeon-r9-m295x-5.html#post9662156

NVIDIA rates MXM in TGP and on-board in TDP, not the other way round.

I only wanted to share the correct numbers...I won't post a document that has "confidential - Internal use only" written all over it...I enjoy what I am doing and want to keep doing it a while longer...

 

lol isnt that what I`m saying?
MXM cards have VRAM + power supply + cores so that will be higher than soldered which only looks at the cores?

MXM = TGP
Soldered = TDP

860M:
MXM (if it existed) = 50W
Soldered = 37W

?

Its not that I don`t believe you because I know you are waaay over my head about these things, but I always wanted to know why about things instead of just taking a number and settling with it.

 

The GTX860M Maxwell onboard has a TDP of 50W, there is no MXM only the Kepler...

@ALL sorry for hijacking this thread...

 

Ah, now I get it lol.

TDP is TDP always. When TDP is given it only looks at the cores but disregard everything else like vram.
I thought they gave different TDP if it is MXM model or soldered model from the same card.

That explains why Kepler bios tweaker shows 80W for GTX 765M MXM from Clevo and 60W for GTX 765M soldered from Alienware. Its not TDP but power and its higher on the Clevo because vram is calculated along with it? Or is it still only for the cores and because Clevo allows higher power supply for scenarios where the cores might throttle while Alienware don`t want anything to do with it?

 

This thread is about AMD Radeon M295X, not about TDP of Maxwell/Kepler. Could you stay on topic, gentlemen?

 

ouh ouh ouh let me understand this well,
my older radeon 7970m have TDP like 84 with Power limit between 56-84 and if i change it for example like TDW for 90 with limit between 80-90 it will bring benefits in performance?
If yes then i'll know how VSS reached the magic level of 1150 on core haa!

 

Changing power table is something that helps in some scenarios.
Miners did exactly that with GTX 750 Ti to increase performance.

I know overclockers do that too on some cards. I don`t know anything more than that, but it helps to feed the GPU enough power if you go over the limit set by the graphic card manufacturer by for example overclocking. The card will become unstable and start throttling if you increase clocks over a certain limit and it can`t get the juice it want because the power table don`t allow it more power. Vbios is ultimately the boss over programs like RivaTuner or Afterburner so thats why these power limits are changed sometimes
Probably not without risks though.

But on topic again since Marecki asked so politely.

 

atiwinflash -p -f 0 new1.rom is juuuust w8ing for me to push "enter"
But i'll wait for more reviews about this. And i see even prema became interested in the topic and for him I have a question.
My favorite favorite favorite stranger from the internet, when you gonna stop working with p370sm bios and come back to p170em with gtx 780m?
I'm not sure if you know but some of users further waiting with teeth impinged on the table and clenched hands in pockets.
Regards!

 

Probably based on AMD Tonga to showcase the next gen power efficient design. There is a new R9-275X coming supposedly so that same GPU would most likely be the R9-295X for laptops just as 7870 was the exact same 7970m/8970/R9 290m etc.

Hawaii is not meant for power efficiency. It surely competes in performance but hoping a Hawaii design on a laptop won't have much of a future.

 

A screenshot I've seen at computerbase.de seem to confirm that indeed R9-M295X will be based on Tonga core:

AMD-Tool deutet an: Tonga mit neuer Architektur - ComputerBase

 

Pretty nice. I really look forward to it. A Tonga upgrade would allow them to compete with higher end performance and even open up other product offerings. At the very least it should offer 880m performance for less power than current offerings. With proper driver support we can have a new nice contender for high performance laptops.

 

Cheaper, cooler, faster. What's not to like!

 

The model number R9-275X for Tonga is not confirmed, its just speculation. It might be called R9-285 or something similar, because it is supposed to be faster than 280.

 

i doubt its going to be cheaper than currently available mobile gpu.

 

It would likely be cheaper than the 880m.

 

Yea, most likely it will be cheaper because it's going into the GX series.

 

Well, with a Kaveri build yes, but chances are Clevo and Alienware will have an Intel build available.

EDIT: Maybe just Clevo, looked at Alienware page and they aren't listing models with the R9 M290X anymore... either they stopped using AMD or no more inventory.

 

Yeah other brands will for sure feature the M295X but MSI will be cheapest among them with the AMD APUs.

Interesting that they dont sell M290X anymore Any_key. Wonder if Dell are empty or just replacing it with M295X soon

 

And magically the R9 290X is back as a selection on some units (as well as 770M :/)... think Dell's ordering pages are having problems right now.

 

Mobile Kaveri APU lineup is on AMD's pages.
AMD APUs for Laptops

Was hoping that their R9 M200 page would be updated, but it's pretty devoid of information.
AMD Radeon™ R9 M200 Series Graphics

Hopefully this is in prep for launch at/around Computex.

 

Nice Any_key. Here is hoping those FX/A10 APUs are good enough to drive the GPUs without bottlenecking them this time.
Also hoping AMD come forward with R9 M295X at Computex in a week and that it spanks the 880M

 

If you check the alienware site on a daily basis, this happens pretty frequently. Sometimes it can change multiple times within 1 day.

 

Too optimistic, the bottleneck would be reduced (by 20-30%), that is not bad, but it would still be there

 

Yeah from the sounds of it the APU's for 2014 and 2015 aren't gon'na be anything to write home about. But the rumored one in 2016 that is being built from the ground up on a new architecture that will support multi-threading sounds very high performance focused... only problem is what will Intel be into at the same time?

 

Mantle finally for notebooks with Enduro: AMD Catalyst 14.6 Introduces Brand New Eyefinity and Mantle for Notebooks with Enduro

Still no more news about r9-m295x though.

 

I want o see how much the MSI GX series improve in BF4 and Thief

 

Has there been any rumors whether the card would be GCN 2.0 or 1.1? Should I be waiting something with beefier ROPs or something like 3/4ths of Hawaii? I really would like to have a mobile card that could push 2560×1600 resolution with comfortable performance.

 

Exactly my thoughts.

 

Rumors are that is based on Tonga, which would be GCN 2.0

 

This is interesting. If Dell release Alienware M17 with R9-295X, I'll definitely buy it.

 

I seriously doubt this M295X will be doing that.

 

Never say never... With the way technology is progressing, it might be possible.. Look at the 7970M... It gave such great performance 2 years ago and the 780M barely exceeded it,...

 

The 780M leads the 7970M by 20-25% in a number of games. Not quite "barely".

On the topic of resolution, 2560x1600 is around 2x the pixels of 1080p. It would take a GPU massively more powerful than the 780M, to run games at 1600p, using the settings the 780M does at 1080p.

I do agree with the mantra "never say never", but I just don't see a new architecture on 28nm giving us a mobile GPU which is near 2x the power of the GTX 780M.

 

78% more pixels at 1440p to be more precise (no 1600p 'cause of 16:10). If 28nm Maxwell's 2x power efficiency holds true at the high-end we could certainly be looking at mobile GPU's capable of driving 1440p at the same TDP envelope.

 

I just wonder whether that same efficiency can be maintained, when the die size and transistor count balloon to what's needed for a high end 28nm GPU.

Please hurry up, 20nm.