New Nvidia Card - 7950 1GB GDDR3

Miscellaneous Forums/General Discussion/New Nvidia Card - 7950 1GB GDDR3

http://www.nvidia.com/page/geforce_7950.html

Check that out guys. Will ATI follow suit?

Don't know what to do now as I was going to build a new PC nextr week with an ATI X1900XTX but might instead go for this new card.

This card is a waste of money imho.

1 GB VRam? You'll only have 512 MB for games.
Also beware that these high end cards have a higher percentage of failrate and a shoorter lifetime. There have been articles on the net about this recently.

I have a x850xt-pe at and won't by a new card before dx10 cards are out.

Btw: What do you plan for CPU and Ram?

Ahem, VRAM is not the main issue here... read the "This dual-GPU board--" part... and imagine two of these in SLI. Quad GPUs, anyone?

It somehow doesn't make sense to me anymore. Most of the new cards consume too much power. I hope that nvidia and ati will follow what happens on the cpu side at the moment. Multiple cores, more power per watt and a lot less power consumption overall.

Same with the newest Futuremark. It was nice till the one before the last one but the range in what is sold in the pc sector is so large that the tool is rather useless for most of the graphic card owners...

If i remember the framerates i got on my 9800pro: well, slideshows are faster...


Same with the newest Futuremark. It was nice till the one before the last one but the range in what is sold in the pc sector is so large that the tool is rather useless for most of the graphic card owners...

If i remember the framerates i got on my 9800pro: well, slideshows are faster...



Each new 3DMark is meant for testing the top hw at the launch time and hw coming after that. Older hw like your 9800pro should be tested with earlier 3DMarks - it's always been like that. Every generation of hw has a suitable 3DMark... it would be useless to run latest Mark with GF4MX for example.

And the latest 3DMark goes like a wind when you run it with latest Nvidias in SLI, so...

And the latest 3DMark goes like a wind when you run it with latest Nvidias in SLI, so...
Having trouble keeping up with the times? Say it isn't so!

@Mustang

Yep, i knew this but:

a) How do i compare results between slow and fast cards then? How compareable are results achieved with different versions of futuremark?
b) It's kind of boring doing a years old test on a slower card isn't it?! Should be obvious for people who come from the demo scene...
c) Also how are features of newly released low range gpus tested where the featureset is richer than what could be tested in older versions of futuremark?
d) If i have a slow and a fast gpu doesn't this also mean that i'll have to buy more than one benchmark?

On power consumption, the 7950 draws less power than ATI's X1900. Bit of a bonus.

Do you really want to spend that much on a vid card though? I bought my 7800GTX when it was bleeding edge, less than 3 months later it has been superceded by about 7 faster cards. I could pay half of what I paid for my card and get something of equal or better speed today.

My advice, get a 7900GT and in six months get whatever is 2nd or third fastest again. You will still have spent less money than if you bought a 7950 today, plus you can eBay your old card and still get a reasonable return on it.

On power consumption, the 7950 draws less power than ATI's X1900. Bit of a bonus.

Yep, but it's still way too much.

My next card also won't be the fastest one. It will be the one which gives me the most power per watt in a perfomancerange above my old one and which lies in a price range up to €250-€300. I'm really not willing to pay any more for a graphics card.

I usually wait and then buy a generation behind. My current GeForce 6 is only a few months old, and plays everything I've thrown at it. Not always at high settings, admittedly, but I save a lot of money, and still get to play the latest games.


a) How do i compare results between slow and fast cards then? How compareable are results achieved with different versions of futuremark?



Result between different 3DMark versions are not comparable. And comparing GF4 and GF6 is difficult - usually you can compare just fine cards that are like generation 6/7 5/6 etc don't have too big tech gap.

3DMark usually focuses new (DX) tech like new shader versions and are meant for testing those card generations, like latest Mark was for DX9(SM3) cards. So it should be mainly used for comparing DX9(SM3) cards because SM2 cards can't run all the tests - thus results are "wrong".

You CAN however use whatnot cards and compare them using whatnot 3DMark if they both can run the same set of tests.

IMO 3DMark is especially good for deciding what new card should you get and comparing those. Older cards are obviously slower so it does not make sense to compare new cards to old ones?


b) It's kind of boring doing a years old test on a slower card isn't it?! Should be obvious for people who come from the demo scene...



? Didn't quite get what was the question?


c) Also how are features of newly released low range gpus tested where the featureset is richer than what could be tested in older versions of futuremark?



Well it's all about what SM level etc they have and use approriate 3DMark for testing. They will get bad scores but in their range the result are comparable.


d) If i have a slow and a fast gpu doesn't this also mean that i'll have to buy more than one benchmark?



No - 3DMark is free you know! :) Seriously, like I said it's all about the tech level of your GPUs. Different SMs (Shader Models) for example require different tools (3DMarks) to test them properly.


Having trouble keeping up with the times? Say it isn't so!



?

Result between different 3DMark versions are not comparable. And comparing GF4 and GF6 is difficult - usually you can compare just fine cards that are like generation 6/7 5/6 etc don't have too big tech gap.

It would be nice if i could compare the gfx-cards in a general and a feature specific way. First as a comparison thought for all cards and second for what you said a featuretest of a new/certain generation.

3DMark usually focuses new (DX) tech like new shader versions and are meant for testing those card generations, like latest Mark was for DX9(SM3) cards. So it should be mainly used for comparing DX9(SM3) cards because SM2 cards can't run all the tests - thus results are "wrong".

Even then there will be such a wide gap between a Intel G965 and a Geforce 7950 GX2 SLI. And the gap will become wider and wider...

? Didn't quite get what was the question?

Futuremark from my point of view isn't just about getting a number out. It's also about the eye candy. Otherwise you could also spare the art (and your job?) and just throw some random stuff at the users eyes. Just rolling out a old futuremark for beeing able to test a slower card isn't so much fun.

Well it's all about what SM level etc they have and use approriate 3DMark for testing. They will get bad scores but in their range the result are comparable

Yep, but not in a whole context.

No - 3DMark is free you know! :) Seriously, like I said it's all about the tech level of your GPUs. Different SMs (Shader Models) for example require different tools (3DMarks) to test them properly.

Oh has this changed? The last one i downloaded, there i needed to buy a licence for if i wanted to do specific tests beside of the defaults or if i wanted to do all tests. Wasn't even switching the resolution disabled?

you know what, I love the fact that the graphics card industry is still competing. Compared to the CPU, the evolution has been amazing.

Gee, 24 billion texels/second is nice :-)


Oh has this changed? The last one i downloaded, there i needed to buy a licence for if i wanted to do specific tests beside of the defaults or if i wanted to do all tests. Wasn't even switching the resolution disabled?



Hey, we (I) have to eat too! :) But you are right about being able to run only the default benchmark with free version. First 3DMarks gave away too much for free -> no money to us...

So yes, you need to buy it if you want to use all the features like selecting what tests to use. But it's a measly $19.95 (upgrade is $14.95) so I don't think that (even) anyone here can call it expensive.

Are you kidding?! Intel are having to slash prices to compete against AMD. There is a very healthy rivalry between Intel and AMD. Can't you remember those few years ago when Intel (and Intel fans) classed AMD chips as second rate, not built for gaming? Boy, that sure got turned around!

Not long ago, my Athlon 3700 was pretty darn fast. Now you can pick up ten CPU's with either a better architecture or faster clock speed, or support DDR2. I certainly haven't seen the pace of progress slowing.


Hey, we (I) have to eat too! :) But you are right about being able to run only the default benchmark with free version. First 3DMarks gave away too much for free -> no money to us...


Oh don't get me wrong. I've no problem with that you earn money with it. So eat as much as you enjoy. :O)

Now offline up to football..... Jipieee! :O)

On the energy point: ATI and Nvidia already sent letters / fax to taiwan so they can prepare PSU for DX10 cards that themself will consume 150-300W ... it is quite sure that first 1000W and 1200 PSU will make their appearance (especially as PPU / GPU for Physic is another mather) as well as secondary PSU just for graphics in 5.25" slots in towers and the like ...

I don't see the point in that brainburnt energy usage anymore ... gosh I've a notebook with a 130W external PSU which is quite fast and powerfull compared to the 300-400W P4 - Intel trashcombos sold by Dell and the like ... somehow I fail to see the relation there ...


And Intel will turn aroud the game by the end of this year. First tests of testing Conroes against FX-60 showed that intel is even then faster if the FX-60 is that far overclocked that it has the same clock as the conroe on base ... and that was a far preproduction Conroe not overclocked to its max ... So it is quite sure that AMDs step to AM2 socket itself won't be enough ... they need new chips ... and I don't think that their decision to move "smaller die sizes" to next year was their most intelligent idea they had the last years ... If they weren't working together with IBM, this could end quite bad for them (remember, there have been competitors to intel once long ago that thought they can hold up as well).

Intels decision to move on with the P-M core was the best for years and if they get a decent FPU unit into conroe / merom, then AMD will get some serious problems. (they already got a pushback due to D805 as it is extremely cheap and can be overclocked with nearly no end. With a usefull fan colling, you can get it up to 3,66Ghz dual where it is nearly as fast as FX-60 for 10% of its price ...)

I bought the geforce cards in the past (1-4). But I wouldn't buy another graphics card from Nvidia, while they concentrate on hardware and neglect their drivers - ATI are very solid.

What's wrong with Nvidia drivers? I can't remember being forced to update my drivers once since I've had this 6600GT. Not because anything was crashing, not because games wouldn't run, nothing.

ATI were pretty decent too, but I did get thoroughly sick and tired of every game that came out working only on one specific driver version. One time, when I bought 3 games together ( 1 was KOTOR, can't remember the other 2 offhand ) I had to change the driver version back and forth every time I wanted to play a different game. Not fun.

That's the impression I get from Nvidia drivers - not ATI, judging by the problems mentioned here for various Nvidia cards with various driver versions - glitches rather than not completely working.

A few Quad SLI setups have been tested and they perform poorly, in a lot of cases WORSE than Dual Card SLI.

A lot of money for not much IMHO, purely for the "more money than sense" crowd.

Darkheart

I agree with Steve. Beside of some openGL quirks i had zero problems with ATI. With nVIDIA i had bluescreens...