Zotac GeForce 9800 GX2

[intel igent]

Apparently Hipro5 used two cards in Crossfire X. source

my bad the 32xxx score was 2 x2's and on air!

i was reffering to his 3D05 single card x2 score of 37xxx on LN2

 

[mdm-adph]

Wow... I'm just stunned....

...I didn't think any card would ever offer less performance per dollar than the HD 2900 XT.

 

[newtekie1]

Wow... I'm just stunned....

...I didn't think any card would ever offer less performance per dollar than the HD 2900 XT.

W1z's relative performance, Performance per Watt, and Performance per $ are off because of the CPU limitting the lower resolution scores.

Because the lower resolution scores are CPU limitted, it makes it seem like the 9800GX2 performs the same at lower resolutions compared to the 3870x2 and other high end cards.

Though this is a prime example of what happens when you pair a $600 video card with a $170 processor. Perhaps an update to the charts is in order once W1z gets his E8400 as I think both the 3870x2 and 9800GX2 is severally limitted by the CPU at lower resolutions by any Core 2 Duo below 3.0GHz.

 

[Grings]

Im kinda shocked by the power usage and noise being so low, its not such a donkey after all.
And props to w1zz for running a 'single core' test, gives a good (rough) indication of how a 9800 gtx will perform

 

[Edito]

i loved the review... keep the things like this wizz ur reviews are sou clear i really liked it... i think this card rocks not for my wallet but rocks lol...

 

[newtekie1]

And props to w1zz for running a 'single core' test, gives a good (rough) indication of how a 9800 gtx will perform

Meh, overclocking a 8800GTS512 to 9800GTX levels would give a better indication of how a 9800GTX will perform. Since the 9800GX2 is clocked even lower than the 8800GTS, it isn't really a good indication of how a 9800GTX will perform.

 

[W1zzard]

because of the CPU limitting the lower resolution scores.

the sli architecture limits as well .. thats why you see lower scores with 2 gpus than with one gpu. but you are right, since we use all results for performance summary, 1024x768 scores are included there as well.

still performance/dollar wise the card sucks .. even if it was twice as fast as it is now on average

 

[phresh]

muchas gracias

Just signed up to say thanks for the review. After seeing that the GX2 is out on slashdot, I went looking for confirmation that it'll work on my p35 board. Found that information and more with this review. Much thanks.

Personally I bought an EVGA G92 GTS in early January with the hopes of 'stepping-up' to one of these beasts eventually and my plan seems to be coming to fruition. I'd never shell out $600 for a card in one motion but as it is I'm looking at about $250 for the 'step-up' so I'm not having much difficulty justifying the purchase. Now I've just got to decide if I want to push my luck waiting to see how the 9800GTX stacks up..

Cheers,
-phresh

 

[newtekie1]

the sli architecture limits as well .. thats why you see lower scores with 2 gpus than with one gpu. but you are right, since we use all results for performance summary, 1024x768 scores are included there as well.

still performance/dollar wise the card sucks .. even if it was twice as fast as it is now on average

What do the performance/dollar charts look like when all the CPU limitted results are thrown out? How does it look with just the 1600x1200 resolutions and higher(or 1280x1024 in Crysis)? Pretty damn good I bet.

Personally, I don't really care how a $600 performs at 1024x768 0xAA, it is CPU limitted with all but the highest CPUs, and I'm not hooking a $600 card up to a crappy monitor either.

I'd like to see the relative performance, performance/watt, and performance/price calculated just considering the higher resolutions, the resolutions that are more likely to be used by people that are buying the card, and the resolutions that aren't CPU limitted.

I'm not saying it is all CPU limitting causing the lower performance at lower resolutions. SLi and Crossfire both show much bigger gains at higher resolutions, in fact the common argument in all 2 card vs. 1 threads is usually that dual cards are only worth while if you use high resolutions. Multi-card solutions are simply more efficient at higher resolutions.

I think the Price/performance needs to be taken in context, at higher resolutions the price/performance is amazing. In fact, it seems this is the only single card configuration that is capable of playing Crysis@1600x1200 4xAA 16xAF.

This is essentially 2 8800GTS512 strapped together. A single 8800GTS is $290. So this is only $20 more than buying the two cards seperately.

The 3870x2 is two 3870s strapped together, with GDDR3 RAM in place of GDDR4, so the 3870x2 is actually a lower value than two 3870s. The 3870 goes for $210. So the 3870x2 is $20 more than two 3870s.

I think nVidia is suffering because they have made a card that is too powerful. It is expensive, but the price is justifable, IMO. The Price/Performance in the review seems low, but I think that is because it was tested in situations that no one will probably ever use it in. I highly doubt anyone is buying this card with the intention of playing any games at 1024x768, I don't even think there are people out there intending to play at 1280x1024 with this card, or at least not sane people.

W1z, I think the review is great, don't get me wrong. I'm not trying to say the review is bad, it is fair. You can't just ignore the low resolution performance, and to keep the testing fair you need to test this card with the lower resolutions, just like you tested the rest.

However, it needs to be pointed out that the resolutions where the card performed poorly are resolutions that no person in their right mind would ever use with this card.

 

[MilkyWay]

how the hell do you get 123.2fps on unreal tournament on an 8800gt 512mb when i get 61fps on the latest leaked beta driver 174.50 windows vista ultimate 32bit granted a better cpu than mine was used but DOUBLE my average frames per second is ridiculous

i game at 1280x1024 because thats all my CRT monitor allows granted my crt monitor has a good clear picture

i had everything at max settings in unreal tournament 3 options

the cards overkill for most gamers okay for big 40plus inch hd tv users and high resolution gamers yeah its a must
really dear compared to the 3870 x2 isnt it

EDIT: how the hell did the 8800gs do so good at 1024x768

 

[newtekie1]

how the hell do you get 123.2fps on unreal tournament on an 8800gt 512mb when i get 61fps on the latest leaked beta driver 174.50 windows vista ultimate 32bit

He is using DX9(WinXP) and you are using DX10(Vista). That is why. DX10 is a lot harder on graphics cards. Plus the fact that you are playing it in Vista also causes a performance hit.

 

[ShinyG]

Good to finally see a direct comparison between the x2 and the gx2. Top review as always, W1zzard!
The card is 8% percent better that the X2 but comes at a 33% price premium! Damn, is it just me or do high-end buyers get ripped off like this all the time?
Just stick to magic quadrant: 3850, 9600GT, 3870, 8800GT I say

 

[W1zzard]

thats the relative performance for 1600x1200 and 2048x1536. feel free to guesstimate price/performance yourself

Relative Performance
perfrel;%;400;27;7;Higher
828282;HD 2400 Pro 256M|6
828282;HD 2400 XT 256M|8
828282;GeForce 8500 GT 256M|10
828282;GeForce 7600 GS 256M|15
828282;HD 2600 XT 512M|20
828282;X1800 GTO 256M|20
828282;HD 3650 512M|21
828282;GeForce 8600 GT 256M|22
828282;GeForce 8600 GTS 256M|27
828282;X1950 Pro 512M|29
828282;GeForce 7900 GTX 512M|40
828282;HD 3850 256M|44
828282;X1900 XTX 512M|46
828282;GeForce 8800 GS 384M|48
828282;HD 2900 XT 512M|56
828282;HD 3870 512M|56
828282;HD 2900 XT 1024M|57
828282;GeForce 9600 GT 512M|59
828282;GeForce 8800 GT 512M|66
828282;Zotac 9800 GX2 single|71
828282;GeForce 8800 GTS 512M|72
828282;GeForce 8800 GTX 768M|76
828282;HD 3870 X2 1024M|86
828282;Zotac 9800 GX2|100

 

[trt740]

for once in a loooong time a ATI card is the better deal

lol @ kthnxbye

now maybe prices will drop.

 

[yogurt_21]

thats the relative performance for 1600x1200 and 2048x1536. feel free to guesstimate price/performance yourself

Relative Performance
perfrel;%;400;27;7;Higher
828282;HD 2400 Pro 256M|6
828282;HD 2400 XT 256M|8
828282;GeForce 8500 GT 256M|10
828282;GeForce 7600 GS 256M|15
828282;HD 2600 XT 512M|20
828282;X1800 GTO 256M|20
828282;HD 3650 512M|21
828282;GeForce 8600 GT 256M|22
828282;GeForce 8600 GTS 256M|27
828282;X1950 Pro 512M|29
828282;GeForce 7900 GTX 512M|40
828282;HD 3850 256M|44
828282;X1900 XTX 512M|46
828282;GeForce 8800 GS 384M|48
828282;HD 2900 XT 512M|56
828282;HD 3870 512M|56
828282;HD 2900 XT 1024M|57
828282;GeForce 9600 GT 512M|59
828282;GeForce 8800 GT 512M|66
828282;Zotac 9800 GX2 single|71
828282;GeForce 8800 GTS 512M|72
828282;GeForce 8800 GTX 768M|76
828282;HD 3870 X2 1024M|86
828282;Zotac 9800 GX2|100

so a 14% Performance boost over the 3870x2 while costing 180-200$ more. not justifyable for most, but those with the money will pay for that 14%, especially considering that the 9800xt was only a few % faster than the 9800 pro 128mb yet cost 150$ more.

 

[newtekie1]

Very nice when you look at it in that light. The 9800GX2 is 14% faster than the nearest competition(X2) when considering only the higher resolutions. Thanks W1z for the graph. 14% is a pretty major jump, especially when comparing the highest of the high cards. Definitely still not a great price/performance, but when considering high end cards like this, every 1% increase in performance is going to cost you 2% in price.

 

[zOaib]

Very nice when you look at it in that light. The 9800GX2 is 14% faster than the nearest competition(X2) when considering only the higher resolutions. Thanks W1z for the graph. 14% is a pretty major jump, especially when comparing the highest of the high cards. Definitely still not a great price/performance, but when considering high end cards like this, every 1% increase in performance is going to cost you 2% in price.

u wud make a great financial advisor for nvidia on - performance on par to how the pricing shud be ............. j/k .............. i am sorry cudnt help saying that , its just the way you phrased the above post

 

[newtekie1]

u wud make a great financial advisor for nvidia on - performance on par to how the pricing shud be ............. j/k .............. i am sorry cudnt help saying that , its just the way you phrased the above post

Well when you look at it, it is true.

Look at the HD3870 and 8800GTS512.

The HD3870 MSRP is $210, the 8800GTS512 MSRP is $290. That is a 38% price increase. The 8800GTS512 is 16% faster than the HD3870. So for a 16% increase in performance you are paying a 38% price premium.

Look at the HD3870x2 and the 9800GX2.

The HD3870x2 is $450, the 9800GX2 is $600. That is a 33.3% price increase. The 9800GX2 is 14% faster than the HD3870x2. So for a 14% increase in performance you are paying a 33.3% price premium.

The 9800GX2's price seems on par to the performance to me. Yes, the card is expensive, but the performance is there too.

 

[candle_86]

Yeah, I would definitely guess the CPU is hold the GX2 and X2 back at the lower resolutions, which is causing the results to be a little unclear. These cards should be scoring much higher scores at lower resolutions.

The card isn't perfect, but it is good to see the cooling doing a much better job than the X2's cooling. NVidia has really put some thought into it, and it shows. The design definitely is complicated, and after market cooling will be a pain, but with temps as good as they are, who needs aftermarket cooling? It is nice to see it overclock so well also.

I definitely thing the card is a very good card, but not worth $600. Buying two 8800GTS 512MB cards is cheaper than that. Then again, buying two 3870s is cheaper than buying a 3870x2. I'd expect the price to drop on this once the initial hype of the card is gone. I expect to see a the price drop to the $500 range in a few weeks.

I'm interested to see quad-SLI up and running though, I hope it is better supported than now than it was before.

Great review W1z, I love your reviews. They are so detailed.

yea look at some of the scores not changing alot the higher the res goes, and finally we have a card that plays crysis

 

[DaJMasta]

A good review, as always, quite through and a very powerful card it seems... of course at the obligatory insane price.


One thing that has bothered me about video card reviews from this site..... though they include a batch of other cards and info like performance/dollar and such, the choice of resolutions seems very random.
It seems to me that there's a huge prevalence of 20-24 inch widescreens..... and that their class of resolutions aren't represented at all by the chosen resolutions. 1024x768 and 1280x1024 i think are really necessary on the low end of cards, but 1600x1200 I rarely see in actual use. I think it would be more useful to the readers to see 1680x1050 or 1920x1200 or even 1900x1080 resolutions with moderate levels of AA and AF, which I think is more representative of midrange and high end computer systems bought and owned today.

2048x1536 seems like a less common resolution to use... though since it's bigger than 1920x1200and with all the eye candy it probably gives a better picture of the overall huge monitor performance (multiple large monitors, 1080p tvs, other huge options) but again, I can't imagine that it's used by all that many people now.


Not that the chosen resolutions don't paint a good general picture and offer performance comparison information, but if they're supposed to be representative of the reader base, I think the higher end resolutions could use an update.

 

[W1zzard]

i dont have a widescreen monitor for testing .. 1600x1200 = 1.92Mpixel, 1680x1050=1.76Mpixel so even less.

 

[ArkanHell]

dont like it, it could be better.

 

[vega22]

top review as always, i do wonder why it seems to be outperformed by the almost identicle 88gts 512 in single gpu mode but hey

...i also ordered an e8400 for our future gpu reviews rig...

any chance you can re-run a few cards with it when you get it setup? if you only did maybe 5 of the top sellers at the time to show how they all compare with the 45nm tech powering them to try and gauge if there is an increase over the current 65nm in games as well as synthetic benchmarks.

edit

@darren
your amd is chocking your gpu m8 check the 3dmarkscores.

 

[kaosII]

Apparently Hipro5 used two cards in Crossfire X. source

You are right George used them crossfired.

 

[W1zzard]

any chance you can re-run a few cards with it when you get it setup? if you only did maybe 5 of the top sellers at the time to show how they all compare with the 45nm tech powering them to try and gauge if there is an increase over the current 65nm in games as well as synthetic benchmarks.

i will rerun all cards i have and also use newer drivers, numbers will be included in the next review i publish using this setup