Incoming: AMD 9970M

isn't the 880 just going to be a rebrand of the 780? i doubt it will drop that much if it is a rebrand. (780 is a 680mx so it isn't a rebrand of the 680m)

 

I`d take that GTX 880M anyday of the week though

I`m 99.9% certain that GTX 880M will not be a rebrand.
Source: GTX 6 75M was a rebrand. GTX 6 80M was not.

 

No, not exactly. Actually, the 780M is basically the closest thing to a 're-brand' we will see on the "high-end" mobile side from NVIDIA. It has the same architecture as the 680M, but features more CUDA cores and a slightly higher clock rate. The 880M is going to support a new architecture [22nm] and probably have more CUDA cores and an even higher clock rate, which should provide a substantial increase in performance over the 780M. Rumors are that it will be at least 30% - 40% better than the 780M, if not more.

Of course it's too early to tell, and I can't back up anything I've said with credible sources, but rumors tend to be somewhat accurate...though, usually on the 'high side' of things.

EDIT: We should start hearing more about it in the next couple months. :thumbsup:

I would, too.

 

from my understanding 22nm was not going to be ready summer of 2014 hence why i said rebrand. unless there is new news....

 

Nvidia could still make a brand new architecture and put it on 28nm. Then do 22nm when that time comes.
Intel does that with their CPUs. Ivy Bridge: 28nm. Haswell: 28nm, new architecture.

Of course 22nm will bring a lot of improvements on the efficiency scale, meaning they can build faster GPUs (especially in the mobile department due to thermal/power restrictions), but a new architecture in itself will also boost performance

 

Well, let's hope they surprise us. I'm not one to live by rumors, but I consider them.

 

I'm not surprised that there is little news at the moment. NVIDIA usually launch the new high end mobile GPUs in summer. It was summer 2012 for the 680M and summer 2013 for the 780M. Logical for 880M to be summer 2014. Seeing as AMD haven't provided any real competition yet, there's no need for NVIDIA to rush another GPU to market. Anyways, I think I'd rather a fully-fledged Maxwell 20nm in summer than a 28nm or Kepler rebrand just so we get it a few months earlier.

Sorry I know this thread is meant to be about AMD but I was just responding to the posts about 880M here.

 

This is the same nvidia we are talking about right? Nvidia is THE king of rebrands.

 

i was surprised 8970m was a complete rebrand and didn't offer any extra cores or processors or pixel pipes or stream something....or whatever they call them today.

 

If NV continue on the "122W is acceptable" path, will AMD follow?

 

i hope so a decent laptop with a decent cooling system can easily handle it

 

It will be interesting to see what happens from both camps.
TSMC is starting mass production for partners February 2014. Wonder how long it takes for them to reach the market? But we will undoubtly have 20nm by summer 2014 for sure. What I wonder, if AMD and Nvidia do the next architectures on 28nm first. If they do, we could have new GPUs just around new year.

Read the recent reports about 20nm here:
http://www.xbitlabs.com/news/other/display/20131022230815_TSMC_Shares_More_Details_Regarding_16nm_FinFET_and_20nm_Progress.html

LOL. Oh there will be rebrands for sure.

 

Nvidia is seeing greatly diminishing returns on 28nm. I don't know that they'd waste the R&D for a new architecture, with this process.

 

Pretty sure the next big release will be around Spring next year? Probably includes any offering from AMD. That was when the 7970M was released anyways...

 

there will be leak and a few ES cards going around but the mass wont be able to get it until june or july

 

The 780M is a 110W card by the way, not 122W.

 

i only see 100w where do you have 110?

 

780m is ~ 20% increased power consumption over a 680m. So 680m is either an ~ 85W TDP card or the 780m is ~ 120W.

Source: Me.

http://forum.notebookreview.com/sag...s-sager-np8250-clevo-p157sm-review.html#power

 

if i recall the 6990 was more power hungry than the 680 and 7970 so whats all i am saying

 

R290Mx LETS GO TEAM

 

dude you cant just left a bomb like that and leave??? give us more details >.>

 

That's not a bomb, it's a wish.

 

Spill the beans if you know anything dude '

Its said to come out early 2014 according to wccftech, so leaks are bound to happen anytime soon I suppose

 

Its now December, care to venture why there has been nothing from AMD about this?

 

well they did announce R9 series on desktop side.

 

I know, kinda irrelevant though.

Desktop discussion isnt really for these forums, as we are only really interested in Notebook GPUs.

 

Like long explains, AMDs Hawaii cards was announced in October like I posted in my first post.
I just thought that new Hawaii cards + the Clevo reseller listing 9970M was a pretty good sign. Sadly it didn`t turn out that way.

Recent news now say early 2014 is when the mobile cards are coming out. Hopefully presented at CES

AMD Kaveri APU With SteamrollerB Core Features 20% CPU and 30% GPU Performance Uplift Over Richland - Platform Details Unveiled

Pretty relevant because AMD didn`t say which cards they were gonna put out on the market. Wether it was desktop or mobile cards.

 

i only got my speculation from the desktop side lol. but i do believe crystal series will be the same or similar to desktop volcanic series

 

Someone who has a Watt Meter could extrapolate/estimate what their 780M or 680M consumes from:

(Total Watts used by the system while gaming) - (CPU overhead/other component overhead) = (GPU Watt Consumption).

That's what I did for my system anyway, because I was curious how many Watts my 670MX was using when fully overvolted & overclocked, I worked it out as about 90W for my fully overclocked GPU. Working out the same thing for stock clocks it only comes to about 45-50Watts, and the 670MX is supposed to be a 75Watt card. So, I think that the TDP listing of these cards by NVidia is not really indicative of how much power these things actually suck! I reckon the 680M consumed less than 100W, and I reckon the 780M consumes 100W, but that's mostly a guess on my part, but I know that Fermi used a load more power than Kepler 600 series for their mobile chips - my previous card in this machine, a 560M, used a full 75W and that was listed a 75W TDP. All this makes me think 680M uses less than 100W, and 780M very close to 100W.

(I used Throttlestop to work out my CPU Watt consumption, and I used Heaven Benchmark for load testing).

 

780m is 110w.

 

That sounds about right to me, based on my thoughts I would think it might be 5 or 10 Watts lower, but I think 110W is plausible. How did you find out it was using 110W?

 

so it is higher than the 6990?

 

I never owned a 6990 so could not say, the 780M power control is specifically set to 110W though.

 

Ah right, is that figure of 110W listed in NV Inspector or something on the Power Slider? Where do you see the power control set?

EDIT: Actually according to screenshots NV Inspector lists the Power Slider in percentages, not on Watt values, so I'm still confused how you know power control for the 780M is set at 110W?

 

It's probably somewhere in the vbios.

 

Yes, I had thought that might be an option, but I'm waiting for Meaker to say where it's listed or how he knows - it adds a bit more credence to his statement if he comes back with an answer.

EDIT: @Meaker: where did you see this 110W figure you keep mentioning? How do you know the power control is set to 110W, where is that listed? You've been asked quite a few times now, by more than one person, I think we just want a simple answer from you. If you just heard it from someone then just say so. Anyway, I'll leave it at that - I find worming out of questions annoying, but I will leave it there even if I don't hear back from you.

 

I`d like to know the power consumption of the card as well.

Isn`t anything listed as power in the vbios a limit anyway? Like you have stock clocks, and then you can increase clocks as long as its within that limit (TDP).

 

That's not very accurate. You have CPU, RAM, LCD, HDD/SSD, mobo, etc to take into account. There's too much margin for error to make it useful. I think the best way is to compare two cards in the same machine and report overall % differences, but that still won't give you an exact amount.

 

Yeah, there's a lot to take account of, and it's the best I could do. I'll explain how I worked it out:

1) System idle, 0% CPU usage = 30W, but CPU at 0% usage using 5W according to Throttlestop, therefore 'system overhead not including CPU' is 25W
2) Overclocked & Overvolted Heaven Benchmark = 140W
3) CPU usage during Heaven benchmark = 20% => shown as 25W being used by CPU in Throttlestop.
4) Using previously acertained figures in steps 1 through 3: 140W - 25W - 25W = 90W

So, that's how I calculated 90W. It's the best I could do. I think it's reasonably accurate. It does take into account the things you mentioned though.

 

6990 was a true 100w from what i have been told on this forum so if it is more or less power hungry than that, it would give a better idea of its true power usage. Also what do yyou mean set to 110 watt? set to throttle at that point?

 

isnt 110w as in TDP? or is actual power usage

 

Pretty sure its TDP, actual usage at that level would be ridiculous lol

 

so now we know 880m will likely be a rebrand of 780m with 8gb vram instead of 4. harder to clock it higher cause of more ram (memory clock) onto waiting for 9970m or Rx 290m.

desktop Rx 290 doesnt just replace 8970 but completely out class it theres a good chance we'll see it catch up to 780m/880m, or perhaps a better performer of this coming entire year. of course if nvidia doesnt go and do something like 880mx?

 

No.

High-end GPUs on laptops are thermal bottlenecked. The new Rx 290X is simply bigger, not more efficient, so it won't make much different on a laptop platform unless AMD decides to follow NV and increase TDP limit.

 

well my 7970m runs cool as can be...it needs a much bigger TDP. with it overvolted on F@H i can't even get the fan to stay on. It hits 65 fan turns on and drops to 60 and turns off -_- I wish my CPU had that fan :/ I would get an XM chip but with meh temps at 3.4GHz i don't even care. I doubt my CPU could touch 3.8 GHz, which is max if i unlock TDP -_- So pathetic. Laptop GPUs had a lot of headroom for TDP in my eyes. I bet the upper limit would be 150 TDP. If we have CPUs pulling 100+ TDP with decent temps on smaller heatsinks and fans i am sure we can squeeze better GPUs.

I hope my next upgrade is GPUs 125-150W.

This isn't counting if Dell decides to just merge the CPU/GPU heatsinks in the m17x/m18x systems. I bet that would give both a good bit of room.

debatable. If you mean 290x it is faster but only because it is a larger die with more crap crammed in it. Exact same die but more ROP and blah blah. I forget. Lost interest in it a few months ago. Waiting for die shrink as usual :/ I doubt i'll upgrade laptop. Desktop will become my gamer. get ride of that crap 650 lol.

 

umm im talking about mobile, Rx 290m.

@hopelesslyfaithful guess so, need another die shrink lol maxwell won't come until 2015 I guess

 

Yep. But if even the top of the line desktop card is made like this, we can't expect anything magically better on laptop platforms.

I do agree with HopelesslyFaithful that mobile GPU TDP can be increased further.


Gaming notebook and mobile workstation designers, how about something like this next time? We might squeeze one XM CPU and one 100W dGPU into the next ultra portable if you guys do manage to beat the efficiency of some hobbyist's hand-made weekend project.

 

More info.please

 

Holy crap thats a lot of copper. Almost like its too much and blocking the air circulation.

What temps improvements did he get?

 

Why would you expect improvements necessarily? There's no enhancement to heat outlet as far as I can tell. The most one could expect is a higher heat capacity (longer time to reach max. temps).

The possibility of balanced cooling load is not realistic:
1. If both GPU and CPU are under load, sharing the heat doesn't help either party.
2. If, without loss of generality, the GPU is idle while the CPU is under load, the interconnected heat pipes will transfer some of the heat from the CPU to the GPU heatsink initially. However, due to the higher thermal resistance between the heatsink and the GPU as compared to between the heatsink and the fins, little of that heat will go into increasing the GPU temperature. As a result the GPU fan will not be activated and the GPU heatsink quickly heats up to choke out any heat transfer from the CPU heatsink.

If anything, I would like to see someone come up with a viable idea of extending the exhaust fins. The issue would be to minimize the frictional pressure loss lest the exhaust air flow velocity (which is critical in forced convection) drops drastically..