[REVIEW] MSI GT60 w/ GTX 780M 4GB GDDR5

GTX 780M 4GB UPGRADED in MSI GT60

The 670MX/675MX/680M heatsink fits the card nicely as it has exactly the same layout as the latters. There are some heat enhancements that can be done (replacing thermal pads and TIM) and this will be done at a later stage.
To install this bloody GPU, I had to install Windows 7 on my Windows 8 GT60 (turning off the Secure Boot and switching from UEFI to Legacy in the BIOS) in order to get the card fully operational - otherwise all the drivers I have tried in my Windows 8 partition made the card unstable or simply Windows 8 unbootable (black screen).

Once Windows 7 installed, I only got the 320.14 drivers to work and get the laptop to switch from iGPU to dGPU and vice versa - I have not tried earlier versions but I tried some leaked drivers and they failed to get Optimus to switch to the dGPU.

So here we go:


The current setup includes:
i7-3630QM - 16GB Vengeance PC12800 - 256GB M4 SSD - and originally fitted: GTX675MX 2GB GDDR5

Benchmarks and extra analysis will come very soon - also plan to make some gaming comparisons with my MSI 16F2 equipped with the powaful 7970M.

3dmark11 @ 771/2500
Generic VGA video card benchmark result - Intel Core i7-3630QM Processor,Micro-Star International Co., Ltd. MS-16F3 score: P7157 3DMarks

3dmarkVantage @ 771/2500
Generic VGA video card benchmark result - Intel Core i7-3630QM Processor,Micro-Star International Co., Ltd. MS-16F3 score: P25133 3DMarks

Unigine Valley is an AMD biased benchmark, however I used it as a comparison for the temps and power consumption vs. the 7970M installed in my 16F2 (cooling further optimized though with modded back panel and custom heatsink).
I am using a 240W Dell PSU and Gelid Extreme thermal paste on both GPUs.


The power consumption is very similar which is great - the temps are quite high though but as I said, the cooling is slightly improved on the 16F2.

 

which driver version are you using?

 

I had botched a 311.44 driver because one .dll was corrupted, this must be the culprit - will clean uninstall/reinstall HD4000/780M tonight with another driver!

 

311.48 is working fine.

 

Hmmm is it a generic driver I could download, or a MSI modded one ?

 

What!
How did you get a 780M? Amazing if it works.

 

Can`t wait for some benchmarks paul

Especially temperature measurements vs 680M

 

I am also wondering how you got the 780m already

 

I will carry on with the testing tonight and update this thread (yeah that's the point of creating one!) - hopefully everything should go fine, I have not spent enough time on it yet.
If there're no news within the next 3 hours, that means I hung myself - alternative option: I fitted the GPU in my 16F2.

 

Unfortunately, I still didn't find a stable driver now downloading my last resort, otherwise I may just go back to Windows 7 for a try.

 

Meh, just another reason to hate Win8

 

SUCCESSFUL ! Windows 8 was the culprit OFC.

OP will be updated soon, I need to re-install all the drivers etc as I just formatted with Windows 7.

 

OP updated with some useful information - testing will resume soon.

 

Finally we get to see what the MSI 780M can do

You should upload bigger pictures btw. I can`t see much from the GPU-Z on the first photo

 

Thanks for the piece of advice updated it.
Now waiting for the W7 updates to be done before performing some stock clocks benches.

 

Hello

GG paul

 

Very awesome! I am looking forward for more testing!

I will surely compare with my machine once it arrives mwuahaha.... whenever it gets released hahahah

 

3dmark11 P score added.

I hope there will be some better drivers in a near future! I think my 7970M stock scores about the same numbers haha!

 

Are these 780m available to buy? How would i go about getting one..?

Thanks

 

NICE! I always enjoy your mods. The card can do better, and I think surely the numbers will improve with better drivers. I have already seen preliminary 3DMark11 (WIN 8 GT70 2OD) scores in the 7300 to 8100 range.

Good Luck!

 

6800 in GPU score. That can`t be right.
Clevo P177SM score 8000. MSI GT70 scored 7700 from the 780M thread in gaming subforum,

 

It's consistent with Reborn's scores considering Optimus always slows down the card a lilttle bit, I suspect the 3dmark11 P scores from the P177SM/GT70 have been done with a clocked gpu as you cannot expect such a bump over the 680M with only 14% extra cuda cores on 3dmark11, especially because the extra bandwitdh does not have any impact on the P score.
I hope I am wrong though!

 

Reborn doesn`t have GPU Boost in his 780M. Looks like you could miss that as well? What vbios are you running?

GTX 780M not only have 15% more cores, it is also clocked 80MHz higher than 680M cores. Then we have the GPU Boost that clocks the GPU up additionally. Plus 1250MHz VRAM vs 900MHz VRAM.

12% higher GPU score than GTX 680M doesn`t make any sense. Surely you must see that it doesn`t add up? It should score atleast 25% better (without GPU Boost active)

 

New cooling systems for msi notebooks should be soon .

 

Do you mean the bridge between CPU and GPU?

 

Maybe boost is not working which is why performance is more inline with a very small OC?

 

GPU boost isn't working at the moment.

 

Yes boost not working would impact the scores, though I am not allowed to say what they should/should not be

 

Rofl.

This card does not work in the MSI 16F2 so expect it not to work in Sandy Bridge based system.
For now, only stock clock and no boost are allowed on my GT60 ! Better than nothing!

 

OP updated softly!

 

Currently testing a MSI vbios where the GPU BOOST and OC works!

Generic VGA video card benchmark result - Intel Core i7-3630QM Processor,Micro-Star International Co., Ltd. MS-16F3 score: P7157 3DMarks stock clocks

Generic VGA video card benchmark result - Intel Core i7-3630QM Processor,Micro-Star International Co., Ltd. MS-16F3 score: P7547 3DMarks 850/2985

Generic VGA video card benchmark result - Intel Core i7-3630QM Processor,Micro-Star International Co., Ltd. MS-16F3 score: P8063 3DMarks 967/2985 with some heavy throttle in the first test mainly.

 

I think it. is not worth to pay so much for gtx 780m. It will be better to buy r8970m.

 

Yep, exactly.

 

Yup, we need to wait for the duo Prema/SVL7 to unlock the vbioses - there are some heavy throttles due to the TURBO BOOST when I clock too high (between 950-1000mhz on the core) - at the moment I'm running @ 910/2950 without a glitch, I'd need to get some games benchmark to compare to the 7970M at full HD rez - will probably do this tonight
I'd never replace my 7970M but I love tweaking those NVIDIA cards.

 

OP updated for good.

 

Is it an MSI card? What memory chips are fitted?

 

pau1ow did you try bumping the core clock + testing and the memory clock + testing to see what increase of Mhz causes the throttle to occur? It might be the memory that don't like overclock past 1250Mhz (like it was with 680M) or the overall voltage on the chip

 

I'm wondering as to from where did u get the 780m gpu card?

 

@Meaker
Clevo card flashed with a MSI vbios - regarding the RAM -> Samsung k4g20325fd-fc04

@Red Line
Yeah unfortunately I have tested a lot of different configuration, and the throttle starts really at 950mhz...depending on how heavy is the app.

@pingz7
Sorry, I cannot reveal this.

 

Hi mate! it would be very interesting, i'm planning a new barebone, as you know, but i want to know if the big price gap between high end vga is someway justified in real games (synthetic benchmarks not always explain the sistuation of course)

 

it is hard to justify on "which one is worth going for" since it is never an idea world

In an ideal situation, we have to make assumptions to on the growth in GPU performance per year, performance scaling in GPU performance per 5 unit increase and performance requirement scaling in games per year

by the following assumptions, where game requirement scales perfectly with GPU performance in a per year basis, and GPU performance scaling per 5 unit where every 5 unit version increase yields you an extra half year meeting gaming performance need, where we assume that the GPU can last for n years for without turning down graphics settings

here we can assume that, IF gtx670mx lasts for n-1 years, gtx 675mx lasts for n-1/2 year and gtx 680m last for n years

given that in an ideal world every 365 days/ 1 year nvidia will release a new GPU for which the top of the line model last n years we get that if by some years where our original GPU goes out of date and needs upgrade, it will take us n-1 years before 670mx needs to be upgraded, 675mx n-1/2 year before upgrade and so on

now we split years into 4 quarters of purchasing and assume that a new release will be made in the 1st quarler, the TOP model lasting n years will go out of date in sync with a new release, the 3rd model last n-1 years in sync with a new release as well whereas the 2nd model which last n-1/2 year is out of sync, in this case we choose only the top or 3rd model and now we weigh it based on the price difference IF it is worth the price difference for an extra year, IF purchase is considered on 2nd quarter, then all 3 models are out of sync with new releases, where both top and 3rd model will last an extra 1/4 year meaning that you wasted 1/4 years worth of money assuming that the upgrade price do not change within that year making it no difference upgrading by the start of the release or 1/4 year after release

here the 2nd model kicks in where you will only suffer 1/4 year after its out dated before newer release, and if that 1/4 year is justified for you it will be of a good buy

by quarter 3 obviously it is best to buy 2nd model in sync with new release and by quarler 4 the buying choice will be between: not buying and wait for new release or buy top or 3rd model and suffer 1/4 year dated

if you consider buying at quarter 1 in an ideal world with the above assumption, check on the price difference and weigh them yourselves if the extra bucks are worth the 1 extra year.

assumption made have flaws where n doesn't sit exactly on an integer value, the 5 unit - 1/2 year long run is not accurate and so on.

 

thanks for your detailed answer, indeed very interesting to evaluate longevity aspects about the choice to do, planning a new gaming laptop is not so easy if you need to balance price and performance...
I've found this comparison review:
NVIDIA GeForce GTX 780M - NotebookCheck.net Tech

780m is due for 20% more than 7970m but almost the same of new 8970m...

 

The 780M will be far more powerful than the 8970M, disregarde Notebookcheck scores, they are wrong (Tesslation was OFF).

Regarding gaming performances, I'd need more time (and games) to test because I'm not really a gamer - only playing Battelfield 3 and CS:GO lol

 

The 8970M is just a rebranded 7970M with higher clocks, don't get me wrong it's a very,very good card but the 780M should be better, either way you can't go wrong but the only advantage AMD has is pricing which is a pretty big one if you ask me.

 

Ah that explains the low clocks

 

As i figured, indeed...elsewhere i found some comparisons between "old" 680m and 780m and the architecture step is sensible, so i hope better gaming compatibility and boost performance compared to the "rebranded" 7970m..

 

Well I was just curious how my 680m compares to the 780m with the SAME clockspeeds
I have an MSI 1762
i7-3630QM @2.4ghz
16gb ram corsair vengeance @1600mhz
680m (stock core is 771mhz for this model)

Running Unigine Valley 1.0, on the Extreme HD preset
his 780m stock: score 1373

my 680m stock (core@771 and mem@900mhz):

FPS:
26.0
Score:
1089
Min FPS:
14.1
Max FPS:
49.5

680m(core @771 mem@1250), same clockspeeds as the 780m

FPS:
30.5
Score:
1276
Min FPS:
15.5
Max FPS:
55.7

so at the same clockspeeds 780m did 1373, 680m did 1276.

which means at the same speeds atleast in that benchmark I came within 7%.... which does not sound right because the 780m has 15%more shaders.
So the 780m must be bandwidth limited in this benchmark also?

It kinda made me feel good about my 680m though
But I also see games showing a different story in FPS. I wish I had the same games to compare against.

 

Found some more tests with 780M:

 

It's hard to compare the results since the two benchmarks suite differ, but most of all i'm supposing a boost with new cpu model, even at the same frequency of 2,4ghz...am i wrong?

 

We won't really know until they are released and members start testing, it could be the CPU boost's the overall score but how benchmarks translate into gaming will tell the real story. If I had a 680M i would be happy too, it's still an amazing GPU so I wouldn't worry