• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

GT 650M or 7970M

M00_cow

n00b
Joined
Oct 2, 2009
Messages
22
Hi guys,

Wanted to ask your opinion. I know most would suggest I grab the 7970M as its easily far more powerful, but what I really wanted to ask was, as I only really play Dota 2, will the 650M (GDDR5 model) do and last me for awhile or is it simply too underpowered?

Thanks for your opinions all!
 
Enduro is a hybrid integrated/dedicated gpu switcher. It causes a lot of problems with 7970s and maxing out gpu usage. It isn't a problem with alienwares though.
Is the 650m lappy cheaper?

Edit: Do you have particular laptops in mind?
 
Currently looking at these Metabox models (Australian buyer):
GT 650M model costing with customization = $1701

or

7970M model costing me with customization = $2153

Buying Metabox because I:
  1. Want a 15.6" model with a FHD screen
  2. Want to replace the optical drive with another harddrive
 
Last edited:
Australia Ftw! Yeah a mate of mine just bought a Clevo with a 7970m an he's having lots of very bad enduro driver problems.

Basically I see it like this with the clevos
- High end Amd = doesn't work as stated
- High end Nvidia = very expensive but works
- Low end Nvidia = works but a lot less performance headroom (1/3 as good as a 7970m incases where a 7970 works (alienwares))

Alienware's are really expensive if you want warranty in Aus though. Also they are chunky and not very understated however they have robust build quality. Couple of mates have had to rma their r2(3?)s.
 
So, what exactly are you saying?

From what I read, you are firstly suggesting the 7970M struggles with driver issues, and then you are suggesting that the GT 650M while working well, doesn't have alot of head room for overclocking?

In your opinion, which should I go with?
 
Tbh, I don't actually know which will be better. Sorry.
They both have issues.

By headroom I mean just general performance and longevity.
How about something like a gtx 675 or gtx 670? Problem with the 7970 is that with its gpu usage issues it will even have problems running games at lower settings.

I'd avoid mobile amd this gen.

Also, my mate reckons http://www.logicalblueone.com.au/ are better to buy from because they are the same price and have a quicker returns process if there is a case where something goes wrong.
Same clevos,
Same country,
Same price,
Better service....apparently.
 
Last edited:
LogicalBlue are an option, but I can't see any laptops they are stocking with the 650m :/

(Scrap that. I'll buy a GTX 660M)
 
Last edited:
Probably not a bad choice as the 660 is actually kepler and not rebadged fermi. Which means it will be cooler and just generally newer tech.
 
So for what it is worth I find that a Clevo with a 7970M works fine. It is a bit pesky, you have to run a program, then quit, after that it'll ask you what GPU to use and you assign it, then it works fine. I find performance to be quite good. I haven't sat around and sweated if GPU usage is up to whatever level as I don't care (and I've seen plenty of cases on my desktop where something else is a limiting factor) I've just played games and they work well.

I think most of the "problems" you hear fall in to one of a few categories:

1) People who are messing with things. AMD doesn't have a general driver for these cards on their site yet. There's a special driver that you get that yes, is an older Catalyst. However, it works fine for me. Use that until a new one is released, don't go messing with hacked betas.

2) People who are obsessing over GPU usage numbers. Those aren't really relevant since you don't play game based on usage, you play them based on how they feel. Also I can't even find any info for how things report usage numbers. Are they talking shaders? fillrate? memory bandwidth? There are all kinds of different ways a GPU can be used and it would be vanishingly rare to ever see all of them maxed out.

3) People who are bitching about temps because they don't understand them. The Clevo fan controller apparently thinks 90 degrees is about the temperature the GPU would be. If it is much less than that, it drops to a lower speed. However some people have decided that is too high and means there's a problem, even though the fan is running below full speed, which indicates the system itself thinks it is fine.

So while the 7970 isn't 100% free of problems or anything, it seems to work. Games run well, and I really don't care about much else.

In terms of DOTA 2 I can't help you, haven't played it. Notebookcheck doesn't benchmark it and they are the only site I know that does any real amount of laptop GPU benchmarks. Looking at the general list of games they test, you might have to turn rez down. Most of the games they had played fairly well (like 30fps or better) on their "high" settings which for NBC means 1366x768 rez and the high preset in the game.

However some games ran full bore. Diablo 3 runs at 1920x1080 with settings maxed at like 50fps, and ME3 runs at like 45fps. Even DXHR ran at like 29fps on DX1 ultra with options turned up.

So depends on how demanding DOTA 2 is graphically, and how optimized it is. It'll run for sure, however you might have to crank things down depending on. With a 7970 of course you have no problems. Only Metro slows down on Ultra settings and only because that game is optimized for shit in DX11 mode (it doesn't even really run that great on a GTX 680 desktop). Everything else, even BF3 can be cranked all the way up and still keep above 30fps.
 
So for what it is worth I find that a Clevo with a 7970M works fine. It is a bit pesky, you have to run a program, then quit, after that it'll ask you what GPU to use and you assign it, then it works fine. I find performance to be quite good. I haven't sat around and sweated if GPU usage is up to whatever level as I don't care (and I've seen plenty of cases on my desktop where something else is a limiting factor) I've just played games and they work well.

I think most of the "problems" you hear fall in to one of a few categories:

1) People who are messing with things. AMD doesn't have a general driver for these cards on their site yet. There's a special driver that you get that yes, is an older Catalyst. However, it works fine for me. Use that until a new one is released, don't go messing with hacked betas.

2) People who are obsessing over GPU usage numbers. Those aren't really relevant since you don't play game based on usage, you play them based on how they feel. Also I can't even find any info for how things report usage numbers. Are they talking shaders? fillrate? memory bandwidth? There are all kinds of different ways a GPU can be used and it would be vanishingly rare to ever see all of them maxed out.

3) People who are bitching about temps because they don't understand them. The Clevo fan controller apparently thinks 90 degrees is about the temperature the GPU would be. If it is much less than that, it drops to a lower speed. However some people have decided that is too high and means there's a problem, even though the fan is running below full speed, which indicates the system itself thinks it is fine.

So while the 7970 isn't 100% free of problems or anything, it seems to work. Games run well, and I really don't care about much else.

In terms of DOTA 2 I can't help you, haven't played it. Notebookcheck doesn't benchmark it and they are the only site I know that does any real amount of laptop GPU benchmarks. Looking at the general list of games they test, you might have to turn rez down. Most of the games they had played fairly well (like 30fps or better) on their "high" settings which for NBC means 1366x768 rez and the high preset in the game.

However some games ran full bore. Diablo 3 runs at 1920x1080 with settings maxed at like 50fps, and ME3 runs at like 45fps. Even DXHR ran at like 29fps on DX1 ultra with options turned up.

So depends on how demanding DOTA 2 is graphically, and how optimized it is. It'll run for sure, however you might have to crank things down depending on. With a 7970 of course you have no problems. Only Metro slows down on Ultra settings and only because that game is optimized for shit in DX11 mode (it doesn't even really run that great on a GTX 680 desktop). Everything else, even BF3 can be cranked all the way up and still keep above 30fps.

People have different opinions of "running well".

Those aren't really relevant since you don't play game based on usage, you play them based on how they feel.

Bad "feeling" games are a symptom of poor gpu usage.

Also I can't even find any info for how things report usage numbers. Are they talking shaders? fillrate? memory bandwidth? There are all kinds of different ways a GPU can be used and it would be vanishingly rare to ever see all of them maxed out.

Msi afterburner among others can measure the memory usage as well as the core usage. This should be maxed on all gpus (especially single gpus) not running on a battery.

So depends on how demanding DOTA 2 is graphically, and how optimized it is. It'll run for sure, however you might have to crank things down depending on. With a 7970 of course you have no problems.

For something like dota 2 it solely depends on how well the clevo utilizes the 7970. If it has good gpu usage it should run flawlessly at 60fps then worse case it could struggle to break 30fps (dirt 3 is a good example). Too much of a crapshoot to be a recommendable product.
 
Hello, I'm a clevo/sager owner myself. After looking at the recommended specs i would sat that the 660 would get you by just fine, the game looks more cpu intensive (which is common with strategy games). If you have the money to burn or want some "future Proofing" id say spring for a higher end card. I agree with craft the 7970 isnt a bad card just has some minor inconveniences that may or may not bother you
 
Msi afterburner among others can measure the memory usage as well as the core usage. This should be maxed on all gpus (especially single gpus) not running on a battery.

No not at all. First like I said, define "core" usage. Do you mean ROPs? TMUs? shaders? There are different parts to a GPU and it would be hard to make a game that will use all of them all the time. A simple example is older games will hit shaders very lightly if at all. So how do you measure core usage? Only if you know what you are measuring do you have a feeling for if it ought to be maxed by a given game.

Then there's the fact that you aren't going to max both core and memory all the time. Generally if one is maxed, the other won't be. One or the other will be the bottleneck. That's just how things go. Some types of games will be more memory bandwidth intensive, others more shader intensive and so on. Depends on the engine, the rez, FSAA, all kinds of things. To think they should both be maxed at all times is silly.

Finally there's the fact that the GPU is only one component, one area you can bottleneck. The CPU is another potential issue. Some games need a lot of CPU power. Other games just fail at CPU optimization. I remember I hit that with the second expansion for NWN2. It ran quite poorly, yet nothing seemed pegged. The answer was that it wasn't using the CPU well, it was basically single threaded and having wait state problems there as well. A similar issue is present with Rift. Though it is multi-threaded, it relies heavily on one thread over the others. So despite having a quad that was not maxed, it was under-using my 5870. Getting a Sandy Bridge fixed that, since it was so much faster per core.

My metric for it if works well is "Do my games play well?" So far the answer has been "yes". I haven't bothered to do much in the way of synthetic testing, other than 3Dmark which performed as expected.
 
No not at all. First like I said, define "core" usage. Do you mean ROPs? TMUs? shaders? There are different parts to a GPU and it would be hard to make a game that will use all of them all the time. A simple example is older games will hit shaders very lightly if at all. So how do you measure core usage? Only if you know what you are measuring do you have a feeling for if it ought to be maxed by a given game.

Gpu usage is measured (as far as I am aware) by the percentage of the total capacity of processes able to be executed by the core. Think like cpu usage on cores. Simple stuff you see in task manager.

Then there's the fact that you aren't going to max both core and memory all the time. Generally if one is maxed, the other won't be. One or the other will be the bottleneck. That's just how things go. Some types of games will be more memory bandwidth intensive, others more shader intensive and so on. Depends on the engine, the rez, FSAA, all kinds of things. To think they should both be maxed at all times is silly.

Memory is utilised by a combination of active use as well as reserve use. ie: bf3 can use 2gb of video memory maxed at 1080p yet that doesnt mean it can't be maxed on 1.5gb of vram. Yet at most bandwidths 1gb won't cut it at ultra. Memory will only cause significant drops in gpu usage if it has a massive simultaneous transfer of data associated with a lack of video ram to run a program. 7970m's have 2gb and should not have any issues here with current titles.

Finally there's the fact that the GPU is only one component, one area you can bottleneck. The CPU is another potential issue. Some games need a lot of CPU power. Other games just fail at CPU optimization. I remember I hit that with the second expansion for NWN2. It ran quite poorly, yet nothing seemed pegged. The answer was that it wasn't using the CPU well, it was basically single threaded and having wait state problems there as well. A similar issue is present with Rift. Though it is multi-threaded, it relies heavily on one thread over the others. So despite having a quad that was not maxed, it was under-using my 5870. Getting a Sandy Bridge fixed that, since it was so much faster per core.

Cpu power can be an issue in causing gpu to be underutilised. This is not the case here as Alienware laptops with equivalent cpus and the ability to turn enduro off at a hardware level do not have this issue (ie: aren't gimped).

My metric for it if works well is "Do my games play well?" So far the answer has been "yes". I haven't bothered to do much in the way of synthetic testing, other than 3Dmark which performed as expected.

I don't follow synthetic benches as I don't fins them applicable to real world applications.

It's like the old days of multi gpu scaling all over again. You won't know what game is going to run terribly. Why play Crysis 2 at sub 30fps when someone with the same gpu is getting 50+?

My mate who has the laptop sums it up well:
"i shouldnt be getting worse performance on some games on a 7970m than i was a 5870m,but i am and for 2400 dollars, thats something to complain about"
 
The 7970m is SIGNIFIGANTLY faster than the 650m, this is not even a contest. The 680m and 7970m are about the same speed, however the 650m? Is way worse than the 7970m.

Go to notebookreview for benchmarks, the 7970m destroys the 650m, period.
 
The 7970m is SIGNIFIGANTLY faster than the 650m, this is not even a contest. The 680m and 7970m are about the same speed, however the 650m? Is way worse than the 7970m.

Go to notebookreview for benchmarks, the 7970m destroys the 650m, period.

I know. The question wasn't, 'Which card is more powerful than the other'.
 
I picked up the GTX675m for my Clevo instead of the HD7970m primarily for better driver support, including Optimus. The GTX675m has been flawless, capable of running BF3 at high settings at 1920x1080. Switching between it and HD4000 is instantaneous and transparent.

The HD7970m is faster of course, but I'll take the stability and usability, especially in a mobile workstation.
 
I have a alienware m18x r2 with dual 7970m.
I can tell you that dota will run with at the highest setting with one card but cannot do any Super Sampling AA while doing it (the fps will dip in big fight below 30 fps), with 2 cards I can run it at 4x SSAA just fine (apart from the chat box text flickering issue with crossfire/sli).
With that said I highly doubt a single 650/660m will be able to keep the game maxed at 60fps, but surely if you drop some of the setting down it could do fine.

There is still hope that they could improve the gfx performance since its still in beta but I would not count too much on that.
 
Back
Top