NVIDIA GT 650M Benchmarks

This is some kind of joke ... I would rather turn optimus completly off =/

 

I had a look through the INF's in the 304.48 and that driver only supports 2 GT650M's =s

 

Am I wrong in assuming that Optimus will still work even if we use these drivers with modified INFs? It seems to be the case when I installed the 301.42 driver with the LV2G INF, looking at GPU-Z and how it utilizes the GT 650M and the Intel 4k.

 

It seems they have acknowledged that some laptop models are still unsupported and so require the following information to add all our laptop models to the upcoming release. This is good news! Head over to official NVIDIA forums to request your laptop be added to their official supported list!

http://forums.nvidia.com/index.php?showtopic=232034&view=findpost&p=1422865

 

Good luck with that. They never support GPU's like that, that's for eye washing people. It's up to Samsung to let Nvidia release new drivers for your GPU.

 

Samsung are opted into the Verde driver program. So there's hope!

 

I will make a modded INF for you, it looks like Nvidia combined the Win 7 and Win 8 drivers with this version. So the INF has changed.

 

Thank you kindly. I really appreciate this!

 

Use the attached modded INF (supports Win 7 and Win 8). This modded INF supports only the Samsung 6xxM mobile GPU's.

 

Seems I was misinformed when I made that assumption. I was referring to: "Exceptions: Notebooks supporting Hybrid Power technology with Intel chipsets are not supported." in the supported product list on the NVIDIA driver download webpage. However, I have just seen this comment on the NVIDIA forums by an NVIDIA Technician: "The comment above is regarding 1st and 2nd generation hybrid graphics technology. Optimus is considered 3rd generation hybrid technology." So it seems we ARE officially supported

Works like a charm. Thank you! I had attempted to make a modded INF of my own earlier today but it failed completely. It installed, however, Optimus screwed up and the 650M ceased to function in any game correctly lol. Reinstalling this works fine! I think I'd better leave this modding to the experts!

It seems performance in some of the benchmarked games has improved yet again over 301.42 modded version. And so, I really need to get around with posting updated benchmarks! Batman: Arkham Asylum however, is bugged, crawling along at abysmal FPS :/

 

Anyone running BF3?

 

Updated with 304.48 modded driver results. (sorry mattstermh, I don't own BF3).

However, the 3DMarks (11, Vantage and 06) have ceased functioning. It's not a driver problem but something else with Windows... i'll try and get to the bottom of it.

Next up... it's deciphering what overclocking limits this card has! Along with installing FRAPS and showing you guys Skyrim playing smoothly on Ultra settings

 

For me the 304.48 drivers improved DX9 by the looks of it but made DX11 slightly worse.

i'm in the same boat as you cakefish, i've got the samsung 550p but the i7 version

 

Nice, you're the first person anywhere I've seen that owns one of these as well! In my country only the 17.3" version has the i7 and 900P screen. 17" was too big for me! Can you tell me what the temps are like with the i7?

Modded 304.48 decimates my 3DMark 11 score but surprisingly not in the GPU tests. It handicaps my processor in some way so it goes extremely slow in the CPU tests. So overall it falls from 2308 to 1456! What is going on there? :S

However, performance in actual games seems to be on par with or above modded 301.42. Except for Batman: Arkham City which takes a nose dive in benchmarks - seems Optimus is messing up somehow?

 

I'm also from the uk =p. Bought mine because of the vatback offer + 10% off cause it was a bankholiday. endded up being £675, so had to have it =p

Temps while running an WII emulater on quite high settings were. This was with the turbo switched on, 3.0ghz to 3.1ghz this core temp said anyway.
CPU's mid 80c's
GPU using just a desktop gadget was 75c.

 

2700 3dmark11 score busted. Guy is crazy

[M14x R2] - VBIOS mod for higher boost clocks - USERS FOR TESTING WANTED

 

Im planning to buy a laptop with the 650m with 512mb of vram. Do you think thats enough to play bf3?

 

The Macbook Pro? Yes it will but if you can afford the 1GB version it'll be even better

Also, the result of overclocking on 304.48: NVIDIA GeForce GT 650M video card benchmark result - Intel(R) Core(TM) i5-3210M CPU @ 2.50GHz,SAMSUNG ELECTRONICS CO., LTD. SAMSUNG_NP1234567890 score: P2625 3DMarks

 

Sadly I cannot adford the 1gb model.

 

but isnt it important? im sure 512 is good for a lot of stuff, but i think for decent resolutions you really need 1 gb of vram?

people will tell you that macs are not build for games. I don't know.

 

Im only going to be on the native no buff resolution. Its either the 512mb lf vram or the 6770m with 1 gb.

 

Tricky one. The 650M is itself much more powerful than 6770M. It's just hard to say whether the 512mb will be enough to cope with 1440x900. Hmmm...

 

Would anyone care to use this Call of Pripyat benchmark? download | Joystiq

Cakefish, the setup you have in your sig is very similar to the setup I want to buy. Do you have the GDDR5 version or the DDR3?

Samsung 550P series:
Intel Core i5-3210m (@2.5 - 3.1 GHz), 6GB DDR3 RAM (@1600MHz), NVIDIA GT 650M 2GB, 1TB HDD 5400rpm, 1366x768 LED Matte display

 

Already done (among many others): http://forum.notebookreview.com/sag...10-clevo-w110er-first-look-review.html#gt650m

 

Ah okay, thanks! 23 FPS min with sun shafts is pretty great!

Cakefish, if can still do the benchmark, that would be great. Your laptop has the i5-3210, and HTWingNut's laptop has the i7-3610QM. I was wondering if that would make a difference.

 

I'll get on it

 

DirectX 10, 720P, like HTWingNut's benchmarks

 

That is a very interesting topic you got there! Have you tried it?

 

2700? 2931 with my trusty Clevo W110ER

NVIDIA GeForce GT 650M video card benchmark result - Intel Core i7-3610QM Processor,CLEVO CO. W110ER score: P2931 3DMarks

http://forum.notebookreview.com/sag...-w110er-first-look-review-82.html#post8636234

Basically the new beta drivers helped. It somehow allows for clocking GPU to 1084MHz instead of only 950MHz previously, and DDR3 vRAM to 1150MHz vs stock 900MHz.

 

Wow, that is a pretty big difference. These are the average FPS results of HTWingNut.

SunShafts
720p: 23
1080p: 28

Rain
720p: 36
1080p: 46

Night
720p: 33
1080p: 44

Day
720p: 35
1080p: 50

It doesn't make sense to me that the average FPS increases with a higher resolution, but what do I know...

 

Im going to write futuremark and report the 3dmark11 benchmark tool. The difference between the gddr5 and the dd3 should not be large but the GDDR5 in theory should be faster. The metro2033 benches between the two show a clear performance advantage of the GDDR5 by almost 5 fps and it goes up from there depending on GPU clocks. So why is 3dmark11 scoring the ddr3 higher? Because i think its only looking at the clock speeds and not actual rendering.

And yes, I am now using the updated vbios for Alienware m14x. Max clocks up to 1070 for GPU (up 100 from bios limited max of 969).

 

Yeah, that's why I don't put a whole lot of faith in Futuremark products. Actual in game FPS or game benchmarks are more realistic. The M14x R2 with GDDR5 had better score higher, otherwise something is seriously wrong.

 

Uploaded with ImageShack.ushttp ://img534.imageshack.us/img534/7556/copbenchmark.jpg

650m 2.0 GDDR5 (m14x)
CPU 3610qm at stock speed. 128SSD

I still believe one of the best current benchmarks with true rendering is the metro2033 test. It just works...

 

:-O That is impressive! Are you using the 304.48 beta drivers? I got a MSI GE60 with 650m gddr5, updating to 296.xx allowed me to overclock both core and memory but when I modified the .inf of the new 304.48 beta drivers and installed them I still still can't overclock more than core +135 (to 969 MHz). Do you think that the extra overclock is only for the ddr3 version? That would be sad

May I ask if you could elaborate a little on how it works to update the GPU to a custom vbios. It sounds really interesting

 

Will there ever be a custom gt650m vbios released for laptops other than the m14x? Specifically, the dv6t-7000?

 

Well I did mine at 'Ultra' preset so it's not a fair comparison. I'll do it again on 'high'

 

Hmm I wonder if NVIDIA is planing to allow for higher clocks for kepler with the next drivers and all hardware ids (kepler) in the drivers can overclock further but because I modified the .inf to install the drivers I missed the other part of the drivers where the extra overclocking is unlocked :confused2:

If that is the case I should hurry to nvidias forum and request the hardware id of the 650m in my notebook to be added to the list of officially supported hardware ids in the new drivers...

 

Yeah let's hope so, but my GT650m in my dv6t-7000 happened to work without using a modified inf, although it was the beta, but the overclocking still wasn't unlocked.

 

I can clock mine to 1085MHz and memory unlimited?

 

It is pretty ridiculous.

 

I've since found out i was wrong. I assumed 'hybrid graphics' and Optimus were the same thing on their website but 'hybrid graphics' actually refers to the manual switching that came before Optimus (which they consider 3rd gen switching tech appaerently).

Also, more benchmarks coming soon, promise!

 

Thanks for the hard work Cakefish! I STILL haven't bought the samsung yet although I most likely will (the fact that the processor is a dual core niggles me somewhat - have you had any luck overclocking it beyond turbo boost parameters by the way?)

It would be useful if this thread remained as a gt 650m ddr3 or GDDR5 discussion thread as it's a relatively new card and the info from recent benchmarks has been very informative.

The capabilities of the card at 1366x768 seems quite promising and I have no need for the eye candy that you get at higher resolutions (opinions vary massively on this point however).

 

Crysis 2 (ultra DX11, high res textures on, 1366x768):

Central Park: 25.8
Time Square: 28.2
Downtown: 26.1

I'll add these & STALKER: Call of Pripyat benches to main graph soon as well as Arkham City and Asylum.

 

Lets see now. You got the ddr3 version, right? Stock overclock to 950 MHz and 1085 MHz is only +135 MHz which compared to the gddr5 version looks like a lot (+135 MHz gives about 970 MHz) but actually is no more than the +135 MHz extra overclock. Seems like I confused my self at first

But if that means that the ddr3 version can be made superior to the gddr5 version that does feel rather wrong

That should mean that the memory never was a bottleneck with the ddr3 version and the gddr5 version only got gddr5 because it is better on papper. Or did I just confuse myself again

 

But
clevo core 1084 MHz => P2931
m14x core 1020 MHz => P2724
can 64 MHz really do all that difference?
Just asking because it seam as the core speed alone cannot cause the entire difference, maybe different GPU drivers?

 

Yes, slower DDR3 is faster than the Fast GDDR5. Because 3dmark11 can't be wrong.

 

That's not what I meant

I'm just curios to why there is such a large difference in favor of the ddr3 version. I mean, there has to be some reason why 3dmark scores one higher than the other. Isn't it supposed to have some connection to reality?

 

3dmark11 is only 1 bench out of many. This kepler tech is not bleeding edge new but its pretty new and its possible that futuremark needs to change the way it generates performance figures during testing. Some tests like metro2033 actually utilize active processing during the bench. Its not an estimate based on the GPU clock speed. Just Cause 2 is another one thats a good bench (older but not bad).

 

Hi guys, I have the np6165/w150erq with 1gb 650m gddr5. I have oc'ed gpu to 970 mhz core and 3000 mhz memory(1500x2) with the 304.48 beta drivers. I also used throttlestop to unlock the turbo for this but the temps reach 93 C and it throttles. This is all with stock thermal paste. My IC diamond paste is coming next week with my notebook cooler. Hope to see better temps then:

http://3dmark.com/3dm11/3726076

EDIT: P2819 (im not so pro with urls). Some other benches here as well:
http://forum.notebookreview.com/sag...ager-np6165-owners-lounge-24.html#post8642963

 

Example LH6KEAN - Cakefish did me a favor and benched Metro2033 with the exact same settings this guy did. Cake said he got 37.67 on the same test when the m14xr2 got 42.67 on the same test. Its not a huge difference and I wouldn't expect it to be. 5 fps seems pretty fair and plausible.

M14x R2 Day One Review