Radeon R9-M295X

Whatever you are just basing your assumptions on comparisons between older architectures and unreleased hardware, there is very little data yet you are even going against whatever little there is available. I still think my expectation are more realistic and since there is no evidence to go against it I will stick to it until more hard data is released. Until then there is no point to treat all this doom and gloom seriously.

And D2 as long as performance/price is better it will be fine. If it can be cooled properly there will be no issues, and also there shouldn't be any issues with crossfire either since unlike older cards there is no crossfire bridge anymore. As for overclocking the GPU will have a lot of headroom so it will all depend on the cooling solution. But since only the minority overclock their GPUs and even less for mobile GPU users due to thermal restrictions in notebooks and many people don't want their laptop to run hotter or louder than it is supposed to, I don't think overclocking will make much difference. Sure to the tiny minority it will but outside of that great stock performance benchmarks and performance/price is what will really matter.

 

The silence interests me too. The fact there are no leaks means they have nothing they want to reveal yet. Who knows perhaps they will push out 20nm tonga and surprise us all. I hope for something outrageous like that. 20nm tonga could perhaps take the fight to nvidia. I would love to see full 32cu clocked at 850mhz-900mhz with high speed VRAM and decent overclockability. Such a card would outstrip the 970m and efficiency wise be similar to maxwell.

By the time nvidia release 20nm amd might then have a new arch we can dream right.

As for 680m being crippled and a bad performer d2? Stock clock is 720mhz. Stock voltage nets average of 950-1000mhz. Pretty good. Also the card although bandwidth limited still far outperforms the non bandwidth limited 7970m with both clocked the same on the core.

 

I'm not saying the 680M was a bad performer... but it was artificially crippled, unlike say the 780M chips. Or the 870M chips, which can get a bit more mem bandwidth and should be OC-able to a higher degree, but are otherwise similar to the 680Ms. As for it being better than the opposition, I think it is where the core power outstripped the memory bandwidth limitation. Remember, core strength in 99% of games FAR surpasses memory bandwidth improvements.

I mean, the 680M was not a bad card by any stretch of the imagination. Not disputing this at all; but the 680M's low voltage memory and very low basic clock speed (which if people who bought it and didn't know about NBR and the modded vBIOS or anything would have kept it right at 720MHz core etc) meant that at STOCK it had way way way too much held-back potential.

In my eyes, for a mobile card, it should have been done better. There was no reason why it was not, and I'll stick by that to this day. It could have had 5000MHz memory, 850 core speed and still be well within the 100W drain and would have run quite cool (as the people who OC'd it within an inch of its life have noticed, it still stays very cool). I think the reason it was limited was because it'd have been too close to the desktop chips (namely the 670). I dislike that notion entirely, and I hope that kind of thing never happens again; though I think it's even happening now with the 980Ms not being a full GM204 GTX 980 chip with a downclock... because the 780M did it with the far comparably hotter kepler GK104, and the maxwell chip wouldn't get nearly as hot (and would drain less power too).

But oh well, we'll see how things turn out.

Sorry, you have to understand, I am not a fan of artificial limitations for the reason of "making one corner of the market feel better about themselves" and for no other reason. It's like how they dumbed down Watch Dogs' graphics to much more closely match the console games, so console users wouldn't feel bad about their "shiny new systems". And the same thing is apparently happening with The Division, and will probably happen with Far Cry 4. It just more rarely happens with mobile GPUs than it does with games, but when it DOES happen, I get very annoyed

 

Unless mobile Tonga is seriously delayed it won't be 20nm. There won't be any 20nm GPUs this year, maybe a paper launch of a flagship GPU at 20nm at the end of December but even that wouldn't be released until 2015. The best we can hope for right now is that mobile Tonga is made in GloFo instead of TSMC. GloFo 28nm is awesome for mobile, just look at how powerful the GPU is in Kaveri, 512SPs in the 35W FX 7600P clocked at 686MHz and that includes the CPU and Northbridge too. Also when they moved from Temash on TSMC 28nm to Mullins on GloFo 28nm TDP went form 8W to 4.5W and even still performance increased.

Edit: More reasons why no 20nm GPUs this year. http://www.fudzilla.com/home/item/35743-apple-takes-20nm-capacity-from-nvidia-amd

 

Everyone is arguing different things, which is why there is so much.... fight? Let's just wait it out. There is actually a lot that can be tweaked into a gpu, and after seeing the leaks from nvidia, AMD will definitely keep silent until they come up with something. In the end, this is also how 290 came to be.

 

I really hope that the delay is due to slight redesign so they can take advantage of FD-SOI, which would mean power AND efficiency

 

If you wonder why I'm daydreaming about this technology: FD-SOI vs FinFET Oh, and GloFo has licensed the FD-SOI in 2012 and is working on it ever since. So if it's 28nm, better be FD-SOI

 

I`m basing my assumptions on a Tonga chip. The exact same chip which will be used on a M295X.

There are ton of reasons which why your assumptions is wrong, which I have written about, that you haven`t responded to yet.

#1: We know that Tonga is the exact same architecture as any other AMD card from the 7000 and R9 xxx series. The only difference is Audioengine, XDMA, and other minor tweaks like color compression and such. Which is why they call it GCN 1.x and not GCN 2.0
#2: We know that the R9 285 draw a ton more watt than the 7870 (185W vs 111W), which makes it very hard to make a mobile GPU out of it.
#3: There are rumors that a R9 285X with 32CUs will have 384bit, which is one of the many reasons why a mobile wont have this. Plus it will draw more than 200W. Yeah dream on.
#4: We know that R9 285 output more heat than the 7870 (190W vs 175W). Which will make a mobile GPU hotter than 7970M. Yet you magically expect even more cores than this.
#5: We know that Tonga are miles from Maxwell in efficiency yet you still think it can match a Maxwell chip in a restrained mobile enviroment where heat and power is everything. How does that even make any sense?

 

Love it if Tonga no longer featured considering Nvidia destroy it with Maxwell!

 

I personally think they completely ditched the plan of bringing a M295X out to the market when they heard about Nvidia`s plans with Maxwell.
They had it ready, we all saw the M295X from ozone3d.net tested. But it was only posted 2 times. One in a desktop, another one in a notebook. No leaks happened after that.

There are rumors about a R9 390X coming out soon since videocardz posted the chassis of it.
I think AMD is concentrating at bringing out the R9 300M cards with a brand new architecture instead of playing catch up with Nvidia`s GTX 880M.
Those cards should bring some serious competition to Maxwell.

That, or they have nothing new to compete with Maxwell until early next year.

 

The question is when....

 

The reason was that if nV would have released in late 2012 a fully-enabled card (780m) instead of the cut-down 680m they would not have had any "new" product to release for two years.

And as for this sort of thing happening again, the 980m will most certainly be a cut-down card and nV will follow up with a fully-enabled core (985m, 990m... take your pick) some time late 2015/early 2016 -- you can take that to the bank.

 

You misunderstand. The 680M could have been a 1344 core GTX 670 without low voltage RAM and a 800 or so core clock and been a proper "mini-670". The 780M could have then been a 900MHz core with 1536 cores; it'd have been a bit less of a jump between 680M and 780M, but it would have had a great thing going. The problem I have is that the artificial limiting via low-voltage vRAM and very low stock clock speed could have been fixed before the card was released. The 800M series didn't really need releasing either. The fixed 680M could have been rebranded as a 770M (maybe with a +50MHz core boost to 850/5000) and the 780M would have taken its kingship and then the only "800M" series cards would have been the maxwell 850M and 860M. I think the mobile lineup would have been FAR better off in that way.

I don't think we need a whole lot of rebrands in the mobile lineup unless it truly brings an actual benefit. I know the 500M chips had a bit of improvement over the 400M chips, but say... the 670M/675M chips? Those were pointless. Those rebrands don't help anyone, you feel me? I know nVidia likes its rebrands and not releasing the real bad boys at launch (they've done it since Kepler; though as far as I know Kepler did not have GK100/GK110 ready until when they made the Titan) but I wish they really wouldn't. It's just a stupid money-grab attempt. If you have improvements, then great! If you don't, then leave the cards up and make your proper release at once. I'm just old-fashioned though I guess.

 

they did that, its called 680mx

 

(lower clocks, only very specific macs)

 

Doesn't matter

Its still the same gpu

 

No it doesnt count.
Its like saying GTX 780M is a rebrand of GTX 680

Have you seen the GTX 680MX?
It doesnt come on a MXM module (mobile). It got a 8-pin power connector, ie desktop connector.

 

lmao look at what AMD's response to game 24 is
Crashing The Party -- AMD's 'Red Team' To Infiltrate Nvidia's Game24 Event To Celebrate PC Gaming - Forbes

is this all you got AMD? first the 285 bs and now this.

sometimes I just wonder if AMD PRs used to be comedians

release the 390x you have a chance, otherwise just don't make fools of yourself even more

 

Why so serious? That's why I love them It doesn't hurt to have some fun from time to time.

 

the point is, you are going in there, knowing for 100% that you are gonna get owned. that's trying to laugh at someone when they got an 89% on a test while you got a 65% yourself

 

lol maybe there will be fist fights between AMD and Nvidia fans?

Did you see the live streaming from AMD where they show the R9 285? If you did, you know there are some serious dedication to AMD among some people.
The audience was ecstatic and was screaming and clapping on pretty much all AMD said and presented, even the most dullest thing.
It almost resembled the typical Apple shows where a religious gang gather to celebrate a new wonderful gift from the gods

 

I don't think the cult of apple is comparable to anything else. I've even heard apple fans complaining why the SSD isn't part of the mobo.

this is just completely ridiculous. at least last time when NV launched titan z, amd fanboys had the right to laugh at it with r9 295x2.

this time, they just got their a$$ kicked completely. you have literally nothing to show off. what are you even gonna do there? laugh at nv fanboys with the 980? get owned in everything?

 

There are quite a few people (and I personally know some of them) who will flat out argue "Why should I pay more for no reason?" and will go AMD forever regardless of what happens.

Then there's others who will be AMD forever simply because they hate nVidia or Intel for some reason or other. But will never admit to AMD doing anything bad. It's kind of sad to see such kinds of fanboys. But nVidia/Intel has some of them too... somewhere. I'm sure. When I see some of them who'll never touch AMD without any kind of good reason, I'll have found them.

 

see the problem is I wouldn't have minded if AMD has something better, but the reality is they don't, they got completely smoked.

back when TTZ was released, AMD actually had something better, what do they have this time? tonga with 32 CU and less TDP? my little pony that can talk? lol

 

I was actually hyped up and ready for the mobile Tonga, but since its the exact same architecture as all other GCN cards and offer nothing new, the interest suddenly crashed.

 

Exactly. I want to see a solid improvement and new architecture from them... but until I see such a thing, I'm just disappointed. Their current tech is not mobile-ready as far as I see. Not at all. They cannot be bringing power-hungry, hot cards and expecting it to compete with nVidia's more efficient, cooler cards in mobile market. Non-mobile market? Sure. It's cheaper in desktop land if it performs similarly. But in mobile? Power drain and heat are the limits, not the actual power of the chip. If a decidedly weaker chip can OC easily past the chip with a stronger base, then the "weaker" chip wins. Especially if it produces less heat/draws less power while OC'd.

 

It doesn't mean that they have to mass suicide or something, now do they I hope that things won't get ugly (like a football-fans ugly (soccer for the americans)), since there are fanatics in every camp, and everything would remain just a fun gathering.

I'm all of the above. I always had slight bias towards ATi/AMD, but recently the nVIDIA fanboys (and they are more than you are willing to admit as it seems) are the ones that swayed me to the ATi/AMD camp. I can go on and on, but would it matter in the end? Everyone is free to choose whatever (s)he wants. You choose to pick the latest and greatest, I choose to give my money for a good cause. Not only nVIDIA has higher markups for the same performance (most of the time), but the market-share is almost twice as big, so in the end, that's a lot of revenue that can be, and will be reinvested. What about the "catch-up"? Where they are supposed to get those money? How they are supposed to make a new architecture? Out of thin air?! OK it' not as black and white as I'm putting it, but it's not very pink either. You have to be thankful for the rest of your lives that there is someone to do at least some sort of push. Take a look at the preliminary info about Broadwell and you'll see what I mean. Also, what about this:

And you'll tell me that's not fanboysm? Buying something even though you realize it's intentionally crippled? "But it's because there's no competition." But why there is no competition?

I'm disappointed of R9-285 so far, but it means nothing. Until they actually release an R9-M295X, there's nothing to write home about. Sure there were some leaks, and then some nVIDIA leaks, and then nothing, but can you bet your home on it? Only AMD knows the truth and I hope we'll see it soon. I'm sure they do their best, like they always do.

 

The kind of fanboys I speak about are the ones who claim AMD can do no wrong. Not the ones who make a conscious decision that AMD is the better product for them, even if the largest factor is "cost". I have a friend who for years refused to so much as even acknowledge that Intel made anything good, and fought with hotfixes and core unparking and heavy overclocking to get his AMD CPU to work great. He won't buy a $300 intel CPU at all, but was all in for the FX-9570 CPUs on launch and the only reason he didn't "upgrade" from his FX-8350 to that was that he didn't have the $900 they charged for a basically factory OC'd 8350.

NO FANBOY OF THIS TYPE IS GOOD. FOR ANY REASON. Most of the people I know who use nVidia will call nVidia out on their crap, and that's what I've done multiple times on this forum. Will I buy AMD? No. Why? Because nVidia is better for the mobile market and has been for some time. I also prefer nVidia's features in NCP etc. Does that mean I think nVidia is the greatest thing since Morgan Webb? Not even close. I was pissed off it took so long for shadowplay to come to mobile, and I am still pissed off at the 680M because I know it could have been better, and I am pissed off at the 880M, and I am pissed off at the 980M because it isn't a fully cored GM204 even though we know it's possible just so they can appease to the desktop users. I still prefer nVidia... I just know they can do wrong. AMD's most vocal fanboys do not seem to say or admit such things (ever) and I have seen more AMD fanboys in that state than I have seen nVidia fanboys who won't admit that nVidia has done some utter horsemoo in their time.

Also, as cloud and I have been saying, if AMD has something under wraps, PLEASE let them bring it. I want them to. He wants them to. WE ALL WANT THEM TO. But if they are going to base it on ANY CURRENTLY EXISTING TECHNOLOGY THAT THEY HAVE SHOWN OFF AT ALL, then they have entirely lost the race. Yes, nothing's set in stone until the R9 M295X is released, but it is not looking promising for the red team unlike for the green team (even with their pointless artificial crippling and sale number tactics).

I know cloud and I are probably gonna wait for the GM210 or full GM204 chips to come out before we even consider trying to get new cards XD. What we got is good enough for now, right? =D

 

I agree, there are definitely more Intel and Nvidia fanboys than AMD ones, more so in a notebook forum and especially in this thread, where we just had one go off the deep end.

At the end of the day, I think it's a reflection of market share, an area where AMD has been the eternal underdog. But I like rooting for the underdog, and I hope they can pull off an upset.

OK enough about Cinderella stories, now I've gone too far with my sports analogies. :laugh:

 

To me, parts of what you say seem to be a load of b*LL**ks - people generally buy products that offer more performance or better value on balance, that's why NVidia have the biggest market share, it's just market economics right. There's nothing wrong per-say with being a fanboy in either camp, but if you're choosing your purchases on that basis, then you're just shooting yourself in the foot & missing out on products that would be better suited. Buying a 980M or 970M would in no way be down to fanboyism, at the moment it looks like it's gonna be the BEST choice if you want some good performance.

 

Buying a GTX 980M that is 50% faster than the GTX 880M with significant less heat. Or a 70% faster GPU than what AMD can offer.

Yup, thats fanboyism right there.

 

Yep, so good you had to like it twice.

 

Haha, must be a glitch, I thought it's impossible to like something twice - normally just likes or unlikes right if you click on it twice!

 

70% faster than AMD in gaming (and probably more thermally efficient). In terms of OpenCL however, Nvidia is hopelessly outmatched.

I think it is a bit unfair to judge a gpu solely on its gaming capabilities whilst ignoring another segment. Even gamers don't spend all their time gaming... a lot of them use OpenCL programs.

Shouldn't a comprehensive analysis of the gpu capabilities encompass everything?
Even official reviews that test the cards in question seem to put forth gaming capabilities ahead of everything else in judging how good a card is. Nvidia might also make their gpus more thermally efficient due to not worrying about OpenCL.

Seems a bit unbalanced... though to Nvidias credit, they did apparently move forward in terms of thermal efficiency whilst AMD has not released a new product in a while.

 

Well we have OpenCL benchmarks of upcoming Maxwell GPUs as well

Here is GTX 970 which GTX 980M is identical to. Against the 8970M/R9 M290X, AMDs mobile top performer at the moment.
CompuBench - OpenCL Device Comparison

Here is GTX 970 against R9 285 which an eventual R9 M295X is based on
CompuBench - OpenCL Device Comparison

So GTX 980M isnt just good in gaming but also computation (OpenCL)

 

I would hope so. 7970M and it's derivatives are getting old!

AMD had better bring it as Nvidia are really destroying them! I like competition......! I can't afford a brand new 970 or 980M lol

 

Hm...
That is a different leaf for Nvidia.
Did they reverse their policy of gimping their gpu's in GPGPU/compute?

EDIT:
This article seems to indicate otherwise:
http://www.tomshardware.com/reviews/geforce-gtx-750-ti-review,3750-16.html

 

What more did you expect out of Nvidia?

 

I see you your BS and raise you with more BS Seriously, it's my word against yours. Why bother?

They screw you over and you bend over. If that's fine by you, go ahead.

A question to all of you, if you are after the superior (note, without quotes) product, why bother dealing with us mere mortals? Let us be happy (or unhappy, time will tell) with our choice.

 

Lol I was right with most of the Maxwell specs probably a month before any real specs came out, I would like to find it and post it here but considering the size of the Maxwell thread and the huge amount of pages with off topic conversations and other stuff like that after it I won't bother looking for it right now. Anyways I said that GTX 980 and the full GM204 would have 16SMM and very high clocks while most people were claiming 15SMM and I said that 980M if based on GM204 would be cut down quite a bit and end up slower than desktop 970. I even said that 970M should most likely be based on GM206 with up to 12SMM and that it will have 192bit bus before the 192but bus was leaked (of course the SMM count hasn't been leaked yet, though considering the lower performance it might also be cut down). Then on this thread I posted that if AMD releases R9 M295X (that is if they didn't change their plans and delay it because of either optimizations or switching from TSMC to GloFo, or canceling it in favor of something else in the works) there was a good chance that it may be clocked at 800MHz and have the full 32CUs (like the earlier leaks pointed out) because of it being more efficient at lower clocks and having great ROP performance with no lower clock penalty unlike Tahiti, estimated (based on calculations of course) the Firestrike GPU score to be about 8000 at those clocks outperforming 970M, and that using unoptimized cheap desktop GPUs to calculate the efficiency and therefore possible performance of Mobile GPUs (which we don't even know is made in TSMC or GloFo) is unrealistic. Then I was called a "Fanboy" for unknown reasons.

The fact that I was right in so many cases above means that I did a decent job of sieving through leaks and separating the more realistic from the BS and also considering different factors that would hint at performance levels as well as calculations to make very close estimates. If the M295X is released anytime soon (the 285X launch should be the earliest) I expect to be right this time also. As for the lack of leaks so far, how is it any different from the desktop 285 which should have had even more leaks? All we know about that thing is it may have 32CUs and thats it.

 

my question would be, what makes you think tonga at lower clocks would outperform a 970m, given that r9 285 can't even match a 770 at even higher clocks?

 

Both of those are faster than 970M.

 

Here's a good read "on" the matter - When TSMC advocates FD-SOI… (well sort of ).

 

Oh look, who would have thought. Even desktop 32CU Tonga have been cancelled. I wonder why...

Spoiler

Maxwell

(Not that mobile Tonga with 32CU would have been remotely possible anyway)
Did AMD just cancel Radeon R9 285X? | VideoCardz.com

@Link4: I hate to rain on your ego boost Link4, but GTX 970M is GM204 but with 64bit cut away.
I said it so many times now: GM206 doesnt exist yet


----------------------------------
Now continue with your wonderful AMD worshipping. The rest of us have already settled with the fact that AMD is long behind Nvidia since Tonga wasnt a new architecture like everyone hoped.

 

Very interesting find! And it is for the best too. Now AMD will push for a good revision and bring the performance in.

 

They for sure are feeling the pressure of Maxwell now.
Like said earlier: Pretty sure GTX 970M and 980M crashed their plans about a R9 M295X. It would barely match a GTX 880M and since 980M is 50% faster, well there goes there plans for mobile. Here is hoping they got a new architecture ready as a backup plan.

On the desktop side Nvidia is releasing GTX 970 for a price of $350 but with enough performance to match a $500 R9 290X but with a lot less power. A R9 285X Tonga for $350 but with a lot less performance and a ton more heat output since its a 3 year old architecture, would be laughable.

 

Whatever isn't confirmed on this thread is noting more than rumors. I said that probably over a month ago and we still don't have complete specs. So lets say the GM206 doesn't exist yet, then what is GTX 960 supposed to use when it launches in October, an even more cut down version of GM204? Considering the aggressive pricing of the 970 it doesn't really make much sense to launch a 400mm^2 GPU at < $250 unless TSMC is giving them away. And don't forget that mobile cards are launching in October too, which is Q4, the rumored release time of GM206. Now I hope they release the mobile GPU specs as well in their upcoming event especially after the Tonga livestream which with the exception of the Never Settle Space bundle announcement was as boring as it could possibly be.

As for worshiping WTH are you talking about, you are just being rude at this point. Competition is a good thing and that's all I care about, at this point I can care less about spending any money on 28nm GPUs, be it mobile or desktop.

 

Just noticed on 3Dcenter's GPU roadmap published on September 8th that 285X is missing.



That's long before the 970's aggressive pricing leak so I don't think Videocardz reasoning that it had to do with 970 pricing had anything to do with it. Makes me wonder what AMD is thinking, what's the point of releasing 285 and not 285X, yield issues could be the reason but everything they do seems too random at this point.

Also how do you guys make these images bigger, every time I post one it's always tiny and the same exact size.

 

GTX 980, GTX980M, GTX 970, GTX 970M and GTX 960 are all GM204

GTX 980: 2048 cores, 256bit - Confirmed
GTX 970: 1664 cores, 256bit - Confirmed
GTX 980M: 1664 cores, 256bit
GTX 960: 1280 cores, 192bit
GTX 970M: 1280 cores, 192bit - Confirmed

GTX 960M should ideally be GM206 though because I think it will be 1024cores or half GM204 so it will be wasted silicon like you say, but if they don`t have it ready, they could use GM204. Haven`t seen GM206 mentioned anywhere from reliable sources so who knows

Specs are all wrong for Tonga. They say it got 40CUs.
The Tonga die undoubtly contain more than 28CUs. AMD denying it is just part of their attempt to not admit that Maxwell crashed their plans. Thats why that roadmap isnt listing R9 285X because it was never announced by AMD. R9 285X would undoubtly be much better than 285 if priced correctly
Also R9 380 and 380X is not Tonga.

 

Lol I didn't say anything about the specs, they completely changed from a month ago, also they changed GM206 specs from 12 SMM to 10 SMM and 1280 cores. They changed Fiji and GM200/210 also, with both being close to 4000 cores down to 3500 and 3000 respectively. Makes me wonder where they get all those numbers.