Theoretical gt300 performance? vs 5870

A

Awesomedude

Weaksauce
Joined
Jan 9, 2005
Messages
93
Ok since we know a lot about the upcoming fermi gt300 w/e you wanna call it card from nvidia, I thought it would be fun to see if we can try to estimate the performance of the gt300, and see how it compares. We are going to have to make assumptions, and these are all purely guesses. I am going to try to look at game performance as well as extras, and am curious to see what the [H] crowd thinks of my numbers, and if the card performs like I guess what do you think of it?

Here is what we know and what we are going to assume for this post:

Fact: The gt300 will have 512 cores, this is 2.13 time as many cores as the GTX 285 which has 240 cores.

Fact: The gt300 has about 6.5% more cores then a gtx 285 sli setup

Fact: The gt300 is a wildly new architecture from the gt200

Fact: SLI doesn't = double the performance, there are losses due to sli

Fact: the gt300 has 2.14 times as many transistors.

Fact: The gt300 will feature gddr 5.0 vs gddr 3.0 for the gt200

Assume: Since according to a front page post nvidia has "eyefinity" technology that they will implement it

Assume: That the gt300 will run at the same clocks as the gt200

Assume: That the gt300 will cost between $500-$600

Assume: That with the gt300 supporting c++ that some apps will become accelerated by gpu

Assume: That there will be a linear increase in performance based on the core increase (this is kind of a big assumption I'm not sure that this will exactly be the case, but we are guessing here, in all actuality it will probably be less)

Assume: That there is a 10% decrease in performance due to sli. (I just picked a number, I imagine this would be lower end, but this is just to get an idea)

Assume: That since this is a new architecture that will be 10% faster per clock then previous architectures. (It is impossible to know if Nvidia's changes will have a positive impact on FPS, or a negative one, but I'd like to think there will be an improvement, but it makes more sense that there would be an improvement, even if most of the improvements are meant for CUDA. 10% is a stab in the dark the performance increase over gt200)

Ok now after a quick googling I could only find one site that compared the 5870 and gtx 285 in sli. I haven't looked into it much but it looks like these are canned benchmarks (not [H] quality) again I am just trying to get an idea on performance if someone has a better review we can compare numbers with then I'd love to see it, but this is what I found.

Now lets start with a gtx 285 in sli as our base performance, and modify that based on the numbers above. First we have a 6.5% improvement for the extra cores, then we multiply in 10% because this is a single card vs the sli configuration, next we multiply a 10% architecture boost. This gives us an overall 28.8% boost over a gt280 sli configuration. Now lets round that down to 25% and look at some benchmarks.

http://www.hexus.net/content/item.php?item=20354&page=4

Here is where I am using the bechmarks from, again these aren't the best benchmarks as far as I can tell but we are making a guess.

Farcry 2 - 2560 x 1600 4xAA 16xAF (average fps)

GTX 285 SLI - 75.5
GT300 - 94.4
5870 - 57.4
5870 XF - 99.1

Grid - 2560 x 1600 4xAA 0xAF

GTX 285 SLI - 106.8
GT300 - 133.5
5870 - 87.2
5870 XF - 151.9

H.A.W.X. 2560 x 1600 4xAA

GTX 285 SLI - 73.5
GT300 - 91.9
5870 - 56.0
5870 XF - 93.5

Call of Duty world at war 2560 x 1600

GTX 285 SLI - 121.4
GT300 - 151.75
5870 - 82.6
5870 XF - 123.5

Enemy Territory: quake wars 2560 x 1600 4xAA 16xAF

GTX 285 SLI - 113.2
GT300 - 141.5
5870 - 76.9
5870 XF - 135.0

Now I know a lot of these games are older games, but even in most of them the gtx285 sli beating the 5870 by a good margin, now if we assume the gt300 will best the sli setup, then we can see that the gt300 is comparing favorably with the 2 5870s in crossfire. All of the cards are providing more then enough performance for any of these games at these huge resolutions, even the GTX 275 in the original review seemed to be handling these well.

Also I read somewhere that flash might support acceleration through CUDA in the future, also since there already exist cpu optimized versions of firefox, one could assume there might be a gpu optimized version or perhaps gpu optimizations or extensions to other programs that will enable a speed boost using an Nvidia gpu, I imagine you could see a huge increase in converting music, video, picture editing and more. There probably won't be a lot of this at launch, but I'd guess within the first year you'd start to see everyday applications make use of CUDA now that you can write it in c++

Now assuming a $500-600 ish pricetag and the above stats how do you feel about the GT300? Worth the wait? Overpriced?

Also I'd love to hear others takes on what the GT300 will perform like, if you agree with my numbers or if you think they are way off.

Also I'd love to see a [H] review of the 5870 compared with a GTX 285 sli setup, maybe even their guess as to how a GT300 might compare based on what we know compared to the GTX 285 sli.

Again this is all just guesswork based on what we know and some benchmarks from some random site off of google, the fermi white paper has me curious on what the GT300 has in store, and this is a fun way to fill the "few more months"

NOTE: I am not in any way shape or form a GPU expert, and some of the stuff in white paper went over my head, I'd love someone with a better understanding to take a better stab at some numbers
 
on your "linear increase with more cores" and "SLI doesn't double performance", look up Amdahl's Law, and you'll understand whats wrong there...

basically to summarize:
double the compute power will never halve the run time, however it can approach double, depending on the level to which a given application, or portion thereof, can be parallelized (in other words, if crap can be sliced 9000 ways, it goes faster, if it has to run as a monolith, you need a bigger monolith (I mean Cray 1 style, not Earth Simulator style))

with SLI, the performance gain is variable, depending on the application, some games will see a 190% performance increase (theres your 10% loss), others will see maybe a 150% or 170% performance increase (I usually will estimate 170% because its roughly dead center, and you'll arrive at roughly +0.7 per GPU if you average benchmarks that compare single to SLI setups)

this also applies to adding processing units to an architecture, in other words, it won't neccisarily double performance by having double the processing units, because of how a GPU looks on the inside (once you get past the magic smoke that is), so it may be 2x the compute power on paper, but it may only be 180% faster in reality

now the fun part:
the GPU is going to have faster internal communication than relying on the SLI interconnect and software abstraction (in other words, its a bigger monolith (OH GOD THE CONTRADICTION)), which means that comparing it to a pair of chips is somewhat hard to do, mostly because its scaling with feature increase is going to be different than scaling an existing architecture in an SMP setup

with all of that said

honestly I think your numbers are probably fairly close to reality, close enough for an educated guess at least, based on a few things:
normally with a new GPU launch you can expect 30-40% more performance over the previous generation (so in the era of multi-GPU setups, basically the new gen is equivalent to whatever two of the last gen could do, give or take)

for example 5870 vs 4870/90, or GTX 285 vs 8800GTX/Ultra, it isn't double the speed, but its about half again, so based on that, your estimations of the GT300's performance are within reality (this doesn't mean that it couldn't be twice as fast, mind you, nVidia has released stuff thats close to 2x faster than their previous generation before (GeForce 6800 comes to mind), not to mention that nVidia is notorious for absolutely BEASTLY designs (its like Intel and nVidia want to race who can get more trannies on a block, Itanium vs GeForce))

so in short:
not a bad guess, and I won't be all that surprised if you aren't that far off, although I won't be all that surprised if the thing doesn't quite approach your estimations

if you wanted to refine the numbers, take GTX 285 #'s and compare them to your GT300 estimate #'s, just to establish better correlation

you could also take some past data, like GeForce 7800/7900 compared to GeForce 8800, to help establish a more accurate line of best fit
 
Awesomedude said:
Assume: That with the gt300 supporting c++ that some apps will become accelerated by gpu
Click to expand...

Not per se. C++ is a language that when compiled by software used on the CPU like gcc, borland, VC++ it gets converted to machine code. It's an intermediary to microcode and asm. I'm not familiar with what Fermi does but maybe it can help speed up compile times? I think I read it can compile C++ itself?

All I need is to go to my professor/program director and say "Yeah, these would be neat to have in the lab." Would turn into "We need those!" :D
 
The 5870 is literally two 4890s put together, yet it often performs worse than two 4870s in crossfire. This is due to things such as Amdahl's law (which obobski mention), limited bandwidth, not needing so many shaders, being cpu limited etc.

Thus you can not assume the nvidia 300 series will get double the performance compared to 200 series.

Second just because it is new architecture doesn't mean it will be more efficient compared to the old architecture for games. They can make the new architecture changes due to things such as using the gpu as a cpu (and the info released yesterday makes it kinda evident that this was the reason.)

Third you can't assume similar clocks, this is a new manufacturing process as well as a new architecture, furthermore since it is a new architecture similar clocks doesn't mean squat.

Fourth you aren't going to get double the bandwidth for even though they are moving to gddr5 they are moving to a smaller bus, we don't know the memory used (just like the clock speed) so you are just assuming.

Finally just because the gpu can do c++ doesn't mean it will do so in games. First you have limited marketshare for the old nvidia gpus don't support and the ati gpus don't support it thus developers won't waste time coding for it.

We learned very little on how the nvidia gpu won't peform in games, and any assuming today is almost useless.. That said I am not buying a 5800 series gpu today or later until the new nvidia cards come out to see how they compare price/peformance. My xfxs 4890s are doing fine (in fact I will probably sell one of them soon while the price is still good) and I don't really need more speed at 1080p gaming
 
there is no way in hell I expect gt300 to be that much faster than hd5870, according to your assumptions, Nvidia doesn't even have a workin part, the card that they showed was a dud tesla based card of fermi and it wasn't even working prototype, I am starting to believe that Kyle was right on when he said you won't see thse gpu's until feb 10, which makes perfect sense, and Anandtec asked Ujesh the gpu guy at Nvidia that why the hell they didn't have a working card yet, and Ujest said becuase designing a chip this big is fucking hard, literally, you can read anand's article on his website. so there you have specs on paper and they still don't have a working prototype, if they did why would nvidia ceo be holding a dud in his hand, and almost everyone is saying that the enhancements in the architecture do not seem like nvidia is too worried about being ahead in games or beating ati, I am sure this card will be fast, but the true potential is yet to be seen. what I wanna see is how fast nvidia would be able to get their shader clock to run, I mean with all those transistors it seems like we won't get shader clock as fast as gtx 285. who know I might be wrong, which I am fully prepared to deal with.
 
NKD said:
I wanna see is how fast nvidia would be able to get their shader clock to run, I mean with all those transistors it seems like we won't get shader clock as fast as gtx 285. who know I might be wrong, which I am fully prepared to deal with.
Click to expand...

AMD doubled transistor count and raised clocks. Do they have magic fairies working for them that Nvidia doesn't have access to?
 
Roland00 said:
The 5870 is literally two 4890s put together, yet it often performs worse than two 4870s in crossfire. This is due to things such as Amdahl's law (which obobski mention), limited bandwidth, not needing so many shaders, being cpu limited etc.


Thus you can not assume the nvidia 300 series will get double the performance compared to 200 series.
Click to expand...

The 5870 is two 4870s not two 4890s.

http://hardocp.com/image.html?image=MTI1NDIwODI1OUNsdmpXQmtnejVfMV8yX2wuZ2lm

http://www.hexus.net/content/item.php?item=20354&page=4

shows the 5870 and 4870x2 comparing about equally.

There are limitations I realize this, 2x was just a guess, its not listed as a fact, its an assumption I made to get some numbers, you could have went with 1.9 and still gotten close to my numbers, 2x was just easier.


Second just because it is new architecture doesn't mean it will be more efficient compared to the old architecture for games. They can make the new architecture changes due to things such as using the gpu as a cpu (and the info released yesterday makes it kinda evident that this was the reason.)
Click to expand...

I mentioned this, but I figured it will likely be more efficient in games, it could be worse, but I imagine it will be better, it could be the same, but because we only have a limited amount of information on the new card you have to make a guess, I made my guesses in the assumption part of the document, to state that they were my guesses. It makes more sense that this new architecture will offer some benefits over the gt200, but you can't know.

Third you can't assume similar clocks, this is a new manufacturing process as well as a new architecture, furthermore since it is a new architecture similar clocks doesn't mean squat.
Click to expand...

Look in order to make any kind of comparison you need to assume similar clocks, it could be clocked slower faster, no one knows. This is just a guess an idea of how it will run. A stab in the dark. Replace the word assume with my random guess, or my random to make the math not ridiculous, or my random because we can't know either way I will pick what seems like a good idea.

Fourth you aren't going to get double the bandwidth for even though they are moving to gddr5 they are moving to a smaller bus, we don't know the memory used (just like the clock speed) so you are just assuming.
Click to expand...

http://hardocp.com/image.html?image=MTI1NDM1MDgzM3RHa3EwSHBHQWhfMV82X2wuanBn

I never said anything about doubling the bandwidth, merely mentioned that we were moving to gddr5 from gddr3, and thats what the white paper says, it supports 6GB of gddr5 dram, take from taht what you will.

Finally just because the gpu can do c++ doesn't mean it will do so in games. First you have limited marketshare for the old nvidia gpus don't support and the ati gpus don't support it thus developers won't waste time coding for it.
Click to expand...

WHAT?!?!?!?! I was talking about it accelerating non-games every day apps (as an additional feature to the gfx card, this may or may not happen, but I've already read somewhere about flash being accelerated with fermi.), and again with this and the memory bandwidth, I didn't include any percentage increase or decrease in performance based on these, just pointed them out.

We learned very little on how the nvidia gpu won't peform in games, and any assuming today is almost useless.. That said I am not buying a 5800 series gpu today or later until the new nvidia cards come out to see how they compare price/peformance. My xfxs 4890s are doing fine (in fact I will probably sell one of them soon while the price is still good) and I don't really need more speed at 1080p gaming
Click to expand...

We won't know how it performs until it comes out, but I felt with the white paper basically explaining on the gpu works, we could take a stab on how it will run in game. My numbers used my assumptions, I'm sure someone could refine my numbers/assumptions and come up with a better answer. This is a stab in the dark, I'm not an expert not even a little bit, I just thought it would be fun/interesting to try to see how the gt300 might run in games. You have to make assumptions to be able to make any kind of comparison, and thats what I did, and outlayed what those assumptions were so everyone could see what I did. If you think you can make a better guess please go ahead I would love to see what other think. But just bashing my assumptions because we don't know doesn't help. Thats why I made an assumption because we don't know/can't know I made what I thought was the best guess at this.
 
trinibwoy said:
AMD doubled transistor count and raised clocks. Do they have magic fairies working for them that Nvidia doesn't have access to?
Click to expand...

well may be who knows? haven't they always been better than nvidia for the last two gens when it comes to power consumption and performance range, it funny how you pick on one sentence out of the whole thing and quote it, and nitpick on one sentence when I was saying that I could be wrong, may be you couldn't read that, it is a new architecture and the clocks always don't work out the same, I say it again may be I am wrong and I don't care if I am wrong, because we are all making assumptions here, never said I hate nvidia, if that ticked you off. shit I could care less if they run faster or slower, and I hope it is fast enough to keep the competiton going, it is good for all of us.
 
Speculation threads can be fun. It's especially fun for posterity, when everyone has "forgotten" what their early impressions were, and have overnight become deadly accurate prognosticators. Everybody's having fun, right? :p

Personally, I hope the GT300 is as beastly as all this. Hopefully prices will drop a bit for everyone's second 5800 in CF. :cool:
 
20%-30% faster tops.

IMHO unless Nvidia pulls some crazy magic with MIMD shaders best case would be near equivalent of GTX 285 tri sli?
 
Awesomedude said:
The 5870 is two 4870s not two 4890s.

http://hardocp.com/image.html?image=MTI1NDIwODI1OUNsdmpXQmtnejVfMV8yX2wuZ2lm

http://www.hexus.net/content/item.php?item=20354&page=4

shows the 5870 and 4870x2 comparing about equally.

There are limitations I realize this, 2x was just a guess, its not listed as a fact, its an assumption I made to get some numbers, you could have went with 1.9 and still gotten close to my numbers, 2x was just easier.
Click to expand...

Actually wrong, the best comparison to the 5870 spec wise is the 4890s (which is a refinement of the 4870s with an increase of the clock speed as well as the pipeline which made it about 6% more efficient for the same clock speed.) The 4890s have the same clock speed while the 4870s have 750 mhz clock speed (vs the the 5870/4890s 850 mhz speed and the 5850 725 mhz clock speed).

http://anandtech.com/video/showdoc.aspx?i=3643

Everything is literally double except memory bandwidth, and the only practical way to double the memory bandiwidth was a bigger bus which AMD didn't want to do for it was one of the many lessons learn from the 2900
 
Is your assumption based on a GT300x2 chip? If the GT300, single core, can get 50% better performance than a 285 then it would be doing decently in terms of progression. However it's going to need to raise the stakes if it wants to beat the 5870. How many of those 3 billion transistors are actually being used for faster gaming, and how many of them are dedicated to the fermi announced goal of better cpu processing/off-loading/folding? The GT300 will be handicapped vs the 5870 from the start if it is dedicating a signifcant percentage of its transistors for non-gpu purposes.
 
GT300, all else equal, would probably need to devote 2.8 billion transistors in order to be equal to 285 SLI, and consequently, ~20-30% faster than 5870. Although throw in 8XAA and the 5870 suddenly looks REAL GOOD.
 
astrallite said:
GT300, all else equal, would probably need to devote 2.8 billion transistors in order to be equal to 285 SLI, and consequently, ~20-30% faster than 5870. Although throw in 8XAA and the 5870 suddenly looks REAL GOOD.
Click to expand...

Yes, because NVIDIA can't match or improve 8xAA performance :rolleyes:
 
vjcsmoke said:
Is your assumption based on a GT300x2 chip? If the GT300, single core, can get 50% better performance than a 285 then it would be doing decently in terms of progression. However it's going to need to raise the stakes if it wants to beat the 5870. How many of those 3 billion transistors are actually being used for faster gaming, and how many of them are dedicated to the fermi announced goal of better cpu processing/off-loading/folding? The GT300 will be handicapped vs the 5870 from the start if it is dedicating a signifcant percentage of its transistors for non-gpu purposes.
Click to expand...

And who says that the consumer cards will have the same features as Fermi dedicated for super computing ?

NVIDIA hosted a "GPU TECHNOLOGY CONFERENCE" and not a "GPU GAMING CONFERENCE", so don't mix gaming and super computing together.

GF100 seems to be very scalable architecture, just like G80 was and many of its features may not see the light of day on consumer cards. NVIDIA has businesses outside of the gaming market and Fermi covers all of them. They have their hard work done now. All they need to do in the future is expand on this architecture.
 
Regardless I perdict that the GT300 will out perform the HD5870 - it has to. Nvidia can't afford to be 3 months late to the ball and then NOT have a faster product.

Other than that I hope that Nvidia makes a competitive product, if not then I'll end up with a HD 5850 this time around :)
 
Darth Bobo said:
Regardless I perdict that the GT300 will out perform the HD5870 - it has to. Nvidia can't afford to be 3 months late to the ball and then NOT have a faster product.
Click to expand...
I think they can.

Anyways, I would be seriously impressed if the GeForce "Fermi" card can run Crysis at 1920x1200 w all settings to "Enthusiast" without any AA at 60fps ave. But I think thats maybe for a refreshed and overclocked "Fermi" Ultra.
 
I think pretty much everyone expects the GT300 to beat the 5870 by a good margin. However, since the GT300 isn't due for release any time soon, AMD does have some time to come up with a card to counter it.
 
cost really needs to factor in as well. You are going to end up paying for all of this stuff this card can do that may or may not affect the performance of the card gaming.

If the card is 30% faster but costs 150% more than the 5870 well that is not something I would go for myself.

<-- hasn't bought an ATI yet but I might go this round.
 
/sigh as much as I'd love to chime in here it seems this thread has been invaded by idiots. Another time another thread I suppose.
 
well I found that the 300 tended to burn my toast and its lasers were not that good at heating coffee.
It can purchase groceries but takes it forever to get them put away.


LONG STORY SHORT

they are blowing smoke trying to kill the 5800's

DONT buy their junk wait for something we might bring out....we are the greatest.

IF they have something bring it out, if its better and cheaper than the 5800's I will order one today.
 
Let's just wait and see aye? We all know they'll get one up on ATi, they always do.

FACT: nVidia can do drivers, ATi can't.
 
Rossi~ said:
Let's just wait and see aye? We all know they'll get one up on ATi, they always do.

FACT: nVidia can do drivers, ATi can't.
Click to expand...

huh? I can throw useless shit out too..

FACT: (and this is true) Nvidia sucks at drivers (remember vista at release? BSOD Paradise)

FACT: They disabled the tray icon making changing the smallest settings take 5 years

see?
 
Shouldnt there be some kind of ratio of # of transistors, die size, clock speed and power consumption in order to figure out gpu1 performance relative to gpu2?
 
DeadSkull said:
Shouldnt there be some kind of ratio of # of transistors, die size, clock speed and power consumption in order to figure out gpu1 performance relative to gpu2?
Click to expand...

Only if the architecture is the same. It is kind of like saying, shouldn't there be some kind of ratio of volume, cylinders, boost, compression ratio, max rpm, valve size, etc used to compare to engines. I mean yes, there is some horrifically complicated math you could do, practically no. Especially considering Nvidia's new architecture is like going from a V6 to a rotary in terms of being similar to the old architecture. You need a new base line for the architecture you don't have.
 
Rossi~ said:
Let's just wait and see aye? We all know they'll get one up on ATi, they always do.

FACT: nVidia can do drivers, ATi can't.
Click to expand...

Both driver sets have their ups and downs. I like the nvidia control panel a lot more than CCC but I also had A LOT of BSODs with my 8800GT until the drivers matured. It was pathetic to say the least.
 
No matter if i got flamed by this,or anything idc ill just point the truth in the past i had many cards.Ati & Nvidia (I"M NOT AN FANBOY) when i bought my 3870, i had problems with drivers game crashing etc.sold it,bought 8800GT everything work at once instal,& play no issues at all! My next card Ati 4870 = epic fail awful drivers,rma it bought GTX260 instal and play no issues at all! Next card,GTX295 instal and play the only issue slow control panel,and minor problems.And now if you think i hate Ati or something just look at my rig and my future cards.The point is with Nvidia i had 0 problems and this doesent mean i won't buy Ati card,but generally the drivers problems with Nvidia was minor from my personal experience,with Ati it was impossible to play like 1/3 of the games i played.

And yes i waited 3 months for 4870 drivers to fix my problems guess what that never happend.The problem is i'm an addict and ill get any next top end card no matter what kind of problem i had and keep telling my self hey! "Ati improve the drivers,that was long time ago! Guess ill see in 1 week or so when i get my 5870s.Sorry for offtopic.

BTW GT300 vs 5870..

IMHO Nvidia will be faster, the question is how much and for what price?

500$ vs lets say reduce price on Ati 300$? for +100 you could get +1 5870 and rape the GT300 prooly,but can in the end any of this cards play Crysis ? :D

P.S How long till will play Crysis @ 1920 x 1200 4XAA at constant 60 fps ? :D:D
 
Last edited:
Nvidia's drivers blow. I still get constant "display driver has recovered" messages after my games crash. I did not have that issue with my 3870x2. Not once.

Nvidia *can* afford to be 3 months late and not live up to the massive hype. Why? Fanboys, and ATI did it with HD 2000. It'll take recovery time, but thats really a non-issue.

While the specs are impressive, I'll wager it at roughly 39% faster than GTX 285 with equivalent clock speeds. Since clocks are REALLY unknown ATM I'd rather not base a prediction on that so much as a comparison with less variables.
 
Thread based on some completely made up FPS numbers with dozens of assumptions. Do you really not have anything else better to do? Sheesh.

Also, you forgot to make up numbers for Crysis.

"But can it play Crysis?"
 
NKD said:
well may be who knows? haven't they always been better than nvidia for the last two gens when it comes to power consumption and performance range, it funny how you pick on one sentence out of the whole thing and quote it, and nitpick on one sentence when I was saying that I could be wrong, may be you couldn't read that, it is a new architecture and the clocks always don't work out the same, I say it again may be I am wrong and I don't care if I am wrong, because we are all making assumptions here, never said I hate nvidia, if that ticked you off. shit I could care less if they run faster or slower, and I hope it is fast enough to keep the competiton going, it is good for all of us.
Click to expand...

Performance? No, they haven't been better. 99% of the time Nvidia has the performance crown, and everyone knows it.

And who really cares about power consumption outside of folders. How many folders are out there, really? Not a lot (comparatively). I mean, it's nice that the 5870 has very low power requirements, and I'm not, by any means knocking it, but that is probably one of the last things most gamers look at when buying a new card.
 
Rossi~ said:
Let's just wait and see aye? We all know they'll get one up on ATi, they always do.

FACT: nVidia can do drivers, ATi can't.
Click to expand...

really? my 8800 GTX / GTX 295 keeps BSOD and that stupid dll error..

while ATI never have this problem for me...
 
psyside said:
No matter if i got flamed by this,or anything idc ill just point the truth in the past i had many cards.Ati & Nvidia (I"M NOT AN FANBOY) when i bought my 3870, i had problems with drivers game crashing etc.sold it,bought 8800GT everything work at once instal,& play no issues at all! My next card Ati 4870 = epic fail awful drivers,rma it bought GTX260 instal and play no issues at all! Next card,GTX295 instal and play the only issue slow control panel,and minor problems.And now if you think i hate Ati or something just look at my rig and my future cards.The point is with Nvidia i had 0 problems and this doesent mean i won't buy Ati card,but generally the drivers problems with Nvidia was minor from my personal experience,with Ati it was impossible to play like 1/3 of the games i played.

And yes i waited 3 months for 4870 drivers to fix my problems guess what that never happend.The problem is i'm an addict and ill get any next top end card no matter what kind of problem i had and keep telling my self hey! "Ati improve the drivers,that was long time ago! Guess ill see in 1 week or so when i get my 5870s.Sorry for offtopic.

BTW GT300 vs 5870..

IMHO Nvidia will be faster, the question is how much and for what price?

500$ vs lets say reduce price on Ati 300$? for +100 you could get +1 5870 and rape the GT300 prooly,but can in the end any of this cards play Crysis ? :D

P.S How long till will play Crysis @ 1920 x 1200 4XAA at constant 60 fps ? :D:D
Click to expand...


If they gt300 is $500, and the 5870 is $300, I'll be getting two 5870's for sure. Unless the gt300 does something amazing that the 5870 doesn't, like give me a handjob for every frag I get.....
 
looks like back to ati......not that im disappointed or anything ive owned ati cards before. it just looks like the gt300 will be to expensive at launch and ati cards will have dropped in price by then.
 
You must log in or register to reply here.
Back
Top