Incoming: AMD 9970M

Haha that picture There's a reason copper is what the manufacturers try to skimp on the most.

 

This is the original post. Sorry, it's not in English.

The machine is a Clevo W110ER. This one now has two fans instead of one, so there is a big increase of outlet size. The fan on the left side of the picture take the place of what was originally a 2.5" hard drive slot.

Even if the outlet is the same, increasing the efficiency of heat pipes still helps. Given the same temperature and heat flow rate at the outlet side, a more efficient pipe setup leads to the same heat flow rate at lower temperature difference between two ends, and therefore a cooler chip.



The temperature achieved with this mod is (i7 3632QM, GT 650M DDR3 UV):

Compared to stock W110ER this is pretty good.

 

too lazy to respond to the rest but this one is easy....who says you would keep the GPU fan stuck to GPU temp only -_- Go through all that design and be that dumb to just leave the fan to only shift up and down due to GPU temp -_-

 

Whee that does sound more impressive. I thought all they did was add some extra heat pipes between the existing heatsinks

Efficiency of heat pipes you say..? It is possible that the extra heat pipes reduce the heat pipe term in the total thermal resistance but I would wager that the die-heatsink interface term dominates the thermal resistance in the first place.. More pipes couldn't hurt I guess but gains are definitely asymptotic.

Sigh for temperature readings outside the contexts of ambient temps and power consumption!

So you need some logic gates to control GPU fan with both CPU and GPU temp sensors..? Should be easy in theory but how to implement in reality?

 

just an idea man ^-^

A program within OS could easily enough control it if designed that way. It monitors from OS instead of firmware

 

Ah that reminds me. HWinfo actually allows fan control using TWO temperature sensors. Fat hope it'll work correctly on a laptop though

 

ah no giggle from my first statement :/

 

I would love to give ambient temp and power but that post didn't mention anything.

Given the time of the post my best guess is the temp was somewhere between 15C and 20C.

 

This bit's not right I think. The rate of heat flow out of the laptop is what determines the temperature. If the amount of heat leaving the laptop is the same, then the temperatures will be the same. The amount of heat leaving the laptop in any given unit of time (aka the heat flow rate) is the important factor.

(But I do think the heat sink mod the guy did was a good job, I can see what he did definitely lowering the temperatures)

 

yes but the thing is most heatsinks does rest at the same temp as the CPU as least with all my experience with butchering notebooks and testing cooling that hasn't been the case...hence the extra heatpipe heatsinks for the m18x serious always helps. If you measure the CPU at 90C and the heatsink is at 60C more heat pipes will help. If CPU is 90C and heatsink is at 80C the gain is marginal. Laptop makers suck...well they suck at making balanced cooling systems...trying to keep the analogies clean ^^. They either have cheap heat sink on the die that can't even maintain even temps/pressure. I dabbled with my heat-spring and my m17x a few weeks ago with it running and trying to find a sweet spot. I couldn't make heads or tales of the readings on if i was getting an even pressure layout. (correction i got an even pressure layout (even temps) but i couldn't figure out how to get max pressure with even layout. I could go on in greater detail but it tires me with all the days i have spent) I got one core super cool and one super hot, all real hot, and all even but meh. The copper plates really need to be better made for contact along with over all more balanced and better designed cooling systems. They honestly dont even try. Also i have dabbled with at least 50 tests with pressure paper and it is 100% useless! Its pressure readings are meaningless for countless reasons. Partly how the paper works and the application presses will never give an accurate reading. I have at least 50 sheets and they all are different because of how it works. To be an accurate reading on the pressure layout is literally impossible. I even dabbled with using 1-8k grit paper and trying to smooth it but it is impossible due to the whole reading process is useless. if it was automated you might get some decent results but by hand it is pointless. Desktop side....i could see some possibilities with high grit paper.

 

You don't have to increase the heat flow rate to drop temperature. A more efficient heat conductor can achieve the same heat flow rate required for equilibrium with lower temperature difference. If the temperature on the sink side (outlet) of the temperature field is the same, which is something we don't care about, a lower temperature difference means lower temperature on the source side (chip) of the field, which is what matters.

This W230ST mod below uses the stock outlet and only adds heat pipes. CPU spec was not mentioned but the GPU under full stress is 12C lower than mine with stock cooling under similar ambient temperature.

　 　

 

as what i stated above it is possible that a more efficient system is irrelevant (it wouldn't actually be more efficient at that point) assuming the heatsink is same temp as chip. The only increase....or decrease i should say would be by increasing how much heat is dispersed by the heat fins/fan

I was also daballing (sp? messing around) with trying to see if there was several patterns with cooling systems. Particularly how much ambient temps affected cooling in low and high temps along with low and high airflow. I never got around to finsihing that part. The testing and data logging to just get enough data to make an intellectual conclusion took 4-5 days :/ I gave up after that with all of my frustration with getting the damn HS to work.

not sure what he was attempting to do with those little copper pieces on the base of the fan...if anything those hurt the process due to solder doesn't conduct well plus added almost no surface area...a sticky copper heatsink would have worked better or a soldered copper heatsink :/

I remmember looking into trying to solder a copper slab mount onto my heatsink to attach a water cooling heat sink but i stopped at that point because soldering heat pipes is supposedly a total pain and i never wanted to bother.

 

I think you might have missed my point, when I was talking about 'Heat Flow Rate' out of the laptop I wasn't referring to speed of air flow out of laptop, I was talking about the amount of joules of heat leaving the laptop (the 'Amount of Heat'). More heat pipes or more efficient heat pipes have the effect of transferring more heat more quickly to the cooling fins - this allows more heat to leave the laptop as the air flows past the fins. I think we probably agree, but may be getting bogged down in terminology. The other effect of huge amounts of copper heat pipes is the sheer heat capacity of the weight of copper, which will allow for lower CPU/GPU temperatures before a steady state equilibrium has been achieved, which is useful when CPU or GPU will be stressed to the maximum, but for relatively short periods of time.

Looks to me like he put those little heat strips there in order to try and use some of the inflow air to the fan to cool a very small portion of the heat pipe - negligible effect probably.

Yes, the whole mounting pressure & evenness of pressure of the heatsink on the CPU or GPU is really important, and it does seem like a bit of an unknown quantity and a bit hit & miss with laptops!

 

Don't know if it was mentioned before. but check this out - MSI Showcases Complete High Performance Computing Lineup during CES 2014 | techPowerUp

 

What I did with my Asus G73 is I removed all the screws to the bottom access panel and then cut velcro tape into long thin strips. Then I put them around the edges where possible and now the bottom panel is easily removable. When I''m ready to play, I tear off the panel and let my laptop cooler blow air directly into the chassis. When I'm done the panel goes right back just as easily. It is good for 3-5c.

 

Hmm....most likely a "typo" to generate page hits. Considering there is know way a R9 M290x could be implemented in a AG2712A.

 

? I wasn't talking about the bottom panel, I was talking about the thin copper strips soldered to the heatsink near the fan.

 

Right, but when I saw those pics the first thing I thought about was that it would work much better with some direct cooling on that copper ...so I dropped the suggestion.

 

Which is exactly what I was talking about (when divided by unit time, or as dQ/dt).


At equili, given the same CPU_power/fan/outlet/ambient_temp, Ts is stable. Then we have:

P=k₁(Tc-Tb)=k₂(Tb-Ts)
Tc=Ts+P/k₁+P/k₂=Ts+P(1/k₁+1/k₂ ​

To achieve minimal CPU temp Tc, k₁ and k₂ should be as big as possible.


More heat pipes helps for exactly the same reason better thermal compound helps. The heat flow rate through the system at equili doesn't increase but the temp slope is less steep.

As HopelesslyFaithful has pointed out, when your heat conductor is good enough, it's no longer the bottleneck and improving it further will bring little benefit. (When k is big enough, 1/k is always close to zero.) But given our experience with different compounds and the successful W230ST mod, I believe it's safe say that what we get now are far from "good enough".

 

Nice! Count on the engineers to keep the equations in a discussion

It hadn't exactly occurred to me that Ts would be fixed but now I see the logic in that. It's easy to estimate k1 given the die area, typical thermal conductivity of the paste and a guestimate for the die/heatsink gap but estimating k2 requires some tech specs. of those heat pipes..

 

Yep, I see what you mean I think - spent about 10 mins looking at that and interpreting it. (I've got a background in Food Technology, and Food Processing was a big part, so involvement with theories of heat transfer in things like heat exchangers I have experience with - but a long time ago now!)

So, I think what you're saying is that more heat pipes doesn't change the amount of heat leaving the laptop at the heat sink fins, but instead more influences the temperature at which Tc will stabalise at in order to flow the same amount of heat out of the laptop. Yeah, that makes sense, the bigger the temperature difference between Tc and Ts, then the greater the rate of heat transfer would be through any given amount of material, but given that you've got more heat pipes in that mod we were talking about, then it can transfer the same amount of heat at a lower temperature difference between Tc and Ts (because it's more efficient, k2) - which is how the low CPU temperature can come about. OK, cool, makes sense. Thanks for that info.

 

The name will be AMD Radeon R9 M290X.

Apparantly its already been tested in Furmark. A notebook with 4800MQ and M290X have benched it.
It scored 4276 points (71FPS).



Can anyone download Furmark and run the 720p (0xAA) mode to see what they score? I don`t have time right now. Off to a party.

 

Great find) Submitted a month ago...

 

So which brands can we expect to hold the next AMD cards?

 

Where did you find that picture?

 

Well, ASUS and MSi will most surely have it.


Sent from my baked potato

 

I'm not too sure about ASUS since they didn't have AMD cards in their gaming notebooks for the last 2 gens. I'd expect MSI, Clevo and Alienware at the very least though. I just hope MSI puts AMD cards in their GT models this time around instead of sticking it into the budget GX model..

 

of all the benchmark software and they picked Furmark to test? seriously?

 

R9 M290x is being showcased @CES by atleast MSI.
Not sure if they mean the same card as 8970M or if it is two options.

According to the Furmark information I posted earlier, 290X looks to be clocked @ 950MHz. If that is 1536 shaders running at that clock, it will surpass GTX 780M.
But then again it could be 8970M clocked slightly higher. Which would be weird because that would mean we have 3 identical cards: 7970M/8970M/R9M290x

MSI to Showcase AMD "Kabini" Tablet at CES 2014

 

my thoughts as well.

 

It may have been notebook OEMs testing its stability and to see the temperatures it get up to in their notebook to ensure the cooling system is proper.

 

that or it's a scheme to get us burn out our old cards and upgrade =))

 

"Look how much score I got in Furmark. I dare you to beat it"

 

thats exactly my thought lol, think of how many 580M they can get rid of lmao

 

I just cancelled my Clevo P150SM with 4800MQ/8970M VGA order. Going to wait till end of the month.

 

That's their all in one series and it looks from that statement like a re brand. Hopefully it at least has faster vram.

 

Personally a poor choice IMO! You should have just gone for it and upgraded at a later stage if necessary.

I bought a Clevo P370SM with that same spec and couldn't be happier. It runs everything!!

 

he just want to have the last of least radeons

 

Of course the P370SM works better because of no Enduro issues. I want a 17" laptop but have been waiting for 20nm Maxwell. May need to wait another year it seems. Unless Mantle proves to gain some significant edge, I may opt for the next gen Radeon.

 

These guys visit the forum I guess: AMD R9 M290X Mobility Flagship Spotted - Appears to be a Rebranded 8970M with Increased Clocks

Sadly they can't see that the maximum temperature was 78º and not 57º.

 

Yeah he got it from this thread. Not the first thread he have been snooking around and posted news about. And he never bothers to post link to this forum either.

Much better to say "our sources say...." to make it more mysterious and trustable than linking to a forum I guess

 

I can imagine you like this

 

Ah the wait is painful if you dont know what you want.

 

just run FurMark v 1.12 with my 7970m at 950/1250 like m290x on the pic and got only 2650 points and avg fps 44, so if this pic isn´t fake and my scores are correct it means m290x isn´t a rebrand :thumbsup:

 

720p tested ? i doubt they can do a 200% increase in perf just with a new architecture + @ the same 28nm process..

 

Just FYI: 0x6801 is the device ID of the 8970m. So that is either a fake or your system has an issue, or you used the wrong preset.

 

Indeed you are correct. 1002-6801 like listed in the R9 M290X Furmark result is the ID for 8970M :/

Thanks. I discovered earlier while snooking around for Furmark results that its very unreliable performance benchmarking tool. Here is a 7970M scoring lousy 1066 points. I found the same weird gap with 680M too.

 

Surely not a rebrand of a rebrand!? Can't be! I run my 7970M at 950/1450 on stock voltage. AMD must be able to do better than that 2 years on...

 

AMD didnt care that much about the mobile cards, just look at the HD8670M P1250 3dm11, come on my hd5650 OC from 2010 does that score
the HD8850M cant even beat the GTX 660M.. if they oc that 660 then they can even arrive to P3.5k 3dm11
I'm not even talking about the 760 765 770M..

 

The 7700M and 7800M were very good, but were still not used very much.