Upgrade! ATI or nvidia, DX 11 tessellation???

[opethdisciple]

It is coming across like that.





Allot of people do not change their gpu's every generation and who knows what the future may hold in terms of games. Future games could be released which use heavy Tessellation and this is where fermi will shine over the 5800 series. Its as simple as that.

I do not change my gpu every generation either, however, I think if you buy a top gpu (in the top category I nclude things like (460 and the 5850 and up) I think its fair to say 2 years is a long time in the gpu market, by which time your card will start to struggle with games two years down the line.

Fine, in two years the 400 series will be better abled to cope with heavely tesselated games than the 5000 series, but in two years we will also have new gpus from ATI and Nvidia....

So its a moot point. You do not buy a gpu for benchmarks... at least I dont.

 

[opethdisciple]

Anyway, Im not here to argue with any one. Nvidia/ATI there is no difference.

Just comes down to price and personal prefernce. Jus for the op, he should not fear buying an ATI gpu, thats all.

 

[Razor Time]

I do not change my gpu every generation either, however, I think if you buy a top gpu (in the top category I nclude things like (460 and the 5850 and up) I think its fair to say 2 years is a long time in the gpu market, by which time your card will start to struggle with games two years down the line.

Fine, in two years the 400 series will be better abled to cope with heavely tesselated games than the 5000 series, but in two years we will also have new gpus from ATI and Nvidia....


So its a moot point. You do not buy a gpu for benchmarks... at least I dont.

A 460 and a 5850 are not top end cards and are clearly mid range gpu solutions.

As for the gpu's lifespan I still have a 8800gtx in one of my rigs and that card is 4 years old and is still gaming capable with most games.

I wouldnt of thought anyone buys a gpu to run a benchmark. However benchmarks like DX11 Heaven can be a useful indication on how a gpu will perform with future DX11 games.

 

[opethdisciple]

460 and 5850 mid range solution, but for the general none enthusats they are still a hefty gpu with a big price tag, hence why I included them. So I say again, a gpu has 'generally' a two year life span.

Of course.. you can still run your gpu for longerthan that, Im not saying that every two years you must upgrade, thats stupid.

Im just saying that to stay current I personally think a gpu upgrade every two years is a good length of time, even if every two years you get a mid range solution, thats all thats needed and probably a very good way to do it, price to perfomance....

 

[Razor Time]

Of course.. you can still run your gpu for longerthan that, Im not saying that every two years you must upgrade, thats stupid.

Im just saying that to stay current I personally think a gpu upgrade every two years is a good length of time, even if every two years you get a mid range solution, thats all thats needed and probably a very good way to do it, price to perfomance....

Ok that is not in dispute. I would also say that a top end gpu has a two year life span maxing out games. So fast forward 12 months and a dx11 heavily tessellated game comes out which gpu do you think will handle it better? My money would be on fermi it does have the better tech with regards to tessellation, which is clearly shown in DX11 benchmarks like Heaven and Stone Giant.

 

[opethdisciple]

Ok that is not in dispute. I would also say that a top end gpu has a two year life span maxing out games. So fast forward 12 months and a dx11 heavily tessellated game comes out which gpu do you think will handle it better? My money would be on fermi it does have the better tech with regards to tessellation, which is clearly shown in DX11 benchmarks like Heaven and Stone Giant.

Yes, there you are correct.

However at the time I purchased my gpu, the choice was a £40 cheaper 470 or a £100 + expensive 480.... so I choose a 5870.

The 470 which was the other option for me, ok, yea, in two years time it will probably handle the tesselation better, but tesselation is only one component of the game, the 470 would perform worse in ever other area.

I just think, you buy tech based on what is in the market and the forseeable future now, and I think thats why ATI was able to come to the market quicker than Nvidia.

There are no games based on the heaven or stone giant benchmarks... they are just benchmarks.

Aditionally, as we have seen with Crysis and warhead, similar grpahics can be improved through coding.

It may just be the case that currently the coding behind tesselation is whats at the root of the problem, improvments to the coding will no doubt also allow ATI 5000 series to perform slightly better. Of course the Nvidia 400 series will still be better, but in one area.

At present the vast majority of games are still dx9 then dx 10 and then a really small amount of dx11.

In two years time, when the dx11 architecture has evolved a lot more (dx12, no doubt), this little debate we are having here will be a moot point.

Becuase it will be upgrade time.

 

[JeffyB]

Ok that is not in dispute. I would also say that a top end gpu has a two year life span maxing out games. So fast forward 12 months and a dx11 heavily tessellated game comes out which gpu do you think will handle it better? My money would be on fermi it does have the better tech with regards to tessellation, which is clearly shown in DX11 benchmarks like Heaven and Stone Giant.

The 5 series will be 2 years old by then and you'd be in your upgrade window anyway (assuming you bought at launch) and newer cards will be far superior at tesselation, should tesselation actually become that important a feature.

 

[Rroff]

Given how close in general the 470 runs to the 5870, how well they overclock and the gains seen with the latest 260 drivers I don't think tessellation is the only area the 470 has going for it.

 

[Razor Time]

Yes, there you are correct.

However at the time I purchased my gpu, the choice was a £40 cheaper 470 or a £100 + expensive 480.... so I choose a 5870.

The 470 which was the other option for me, ok, yea, in two years time it will probably handle the tesselation better, but tesselation is only one component of the game, the 470 would perform worse in ever other area.

The OP is looking now and not what prices were a few months ago. The 470 is a fair better deal than a 5870 is at the moment. Also the 470 performs very well when you overclock and up the AA. A 5870 as good as it is cant do tessellation as well nor can it handle high levels of AA as well as fermi. This is a plain and simple fact. No amount of cool running or low power usage will change this. The 5800 series is also lacking in TRSAA, 32xAA, physx and 3d. Its not always a simple matter of comparing benchmarks running at 2xAA on a DX9 game at stock clocks.

I just think, you buy tech based on what is in the market and the forseeable future now, and I think thats why ATI was able to come to the market quicker than Nvidia.

ATI came to the market quicker as the 5000 series is a beefed up 4000. Basically the 5000/4000/3000/2000 are all the same architecture but the 5000 series has DX11 bolted on. Nvidia's cards are totally new architecture and they totally dropped a bollock on getting it out of the door on time.

This is why ATI came to the market quicker as their path was easier than bringing out a whole new architecture. This helped them regain allot of market share. At the time the 5000 series was released it was pointless buying anything other than a 5000 series as nvidia were still stuck with their dx10 cards.

There are no games based on the heaven or stone giant benchmarks... they are just benchmarks.

Thats correct no game currently uses heavy tessellation but who is to say that will not change. As I mentioned before DX11 benchmarks are a good indication of how future games may perform. Stating that you may have a different card by then is not the point as who honestly knows how long someone will keep their hardware for. I have multiple pc's in my house and I cycle the tech from my main rig down to other rigs. This means that I may keep a gpu for many years. We also dont know when the next heavy tessellated game will emerge.

The plain and simple fact is fermi handles heavy tessellation better than the current crop of ATI cards. No doubt this will change and see saw back to ATI with the up and coming 6000 series. And so on.....

 

[opethdisciple]

Yes, 470 very good at its current price.

 

[Carlos_will]

Right, this thread has been very interesting to follow. And again, I am chuffed I have started off such a popular thread waaaaaaaaaaaay.

Anyway, so we have come to the conclusion the 470gtx is probably the best at value for performance and has technologies that ati lacks. But what about power consumption? I have seen on a website of a dude comparing the 4 card, 5850, 5870, 470 and 480. The 5850 and 5870 wattage usage is similar to the 8800gtx, which is quite impressive. Now this is straying away a bit from the topic, but in terms of electricty bills (haha) would I notice a lot of difference between I don't know...300watts (load) and 450-ish watts (load) ? I'm going to uni and would like to save on the bills.

Infact, is there even a calculator online or anything to tell me how much electrcity my rig uses

 

[Rroff]

IF it works properly the GTX470 will drop down to 50MHz clocks when idle or low desktop useage which saves a ton on power consumption especially if you spend more time at the desktop than gaming i.e. leave the PC on overnight... however it seems a bit hit and miss if this actually works

 

[Broken Hope]

So fast forward 12 months and a dx11 heavily tessellated game comes out which gpu do you think will handle it better? My money would be on fermi it does have the better tech with regards to tessellation, which is clearly shown in DX11 benchmarks like Heaven and Stone Giant.

Considering ATI have something like 90% of the DX11 market according to the steam survey no developer is going to release a game with heavy levels of tessellation that would run like crap on 90% of the DX11 cards available.

In the end it doesn't really mean much for games how fast fermi is for tessellation until ATI increase their tessellation performance to equal or beat fermi's.

 

[Alexrose1uk]

IF it works properly the GTX470 will drop down to 50MHz clocks when idle or low desktop useage which saves a ton on power consumption especially if you spend more time at the desktop than gaming i.e. leave the PC on overnight... however it seems a bit hit and miss if this actually works

Yeah, ULPS has been a bit of a headache overall, for both Nv and AMD. Some people have a great, quiet, cool time with it, others just end up with frustration, which is a shame, as tis a really good idea overall, especially in this day and age.

I was pleasantly surprised when the UPLS on my 5870 mobile cards actually worked haha

 

[arknor]

Does that mean, that the future of gaming is tesselation? Maybe not at all. Considering its already been with us, for so long. Its marketing. Nvidia, say buy our gpu's becuase we do tesselation better than ATI. (As I mentioned before, some game developers still code in dx9!)

it may have been with us for years but in its original basic form when ati called it "truform" ati werent a very popular graphics card maker (8500 days) and they never encouraged anyone to use it , i think only counter-strike and NWN used it and it was almost spot the difference.

modern tesselation can be really awesome without much of a performance hit on fermi cards and no doubt nvidia will give the devs the encouragement they need to use it

 

[Razor Time]

Considering ATI have something like 90% of the DX11 market according to the steam survey no developer is going to release a game with heavy levels of tessellation that would run like crap on 90% of the DX11 cards available.

In the end it doesn't really mean much for games how fast fermi is for tessellation until ATI increase their tessellation performance to equal or beat fermi's.

Which will be in a few months time when ati release the 6000 series. Also do not forget all the nvidia backed games ie the way it's meant to be played games.
Nvidia are likely to ramp up the tessellation to suit their cards more.

 

[Broken Hope]

modern tesselation can be really awesome without much of a performance hit on fermi cards and no doubt nvidia will give the devs the encouragement they need to use it

To be honest modern tessellation so far has been crap, when you look at the wire frames of the models they just tessellate without any sort of intelligence, there is no gain or benefit from tessellating a flat surface into thousands of triangles you gain no extra model detail but incur a huge performance hit.

Until developers start using tessellation properly I'd prefer they don't use it much at all.

 

[REAPERUK]

Go Nvidia Fermi 400 series 470/480 or wait for the refresh 400 cards if you want the best at DX11/tessellation, and not forgetting 32AA ect.

if you do wait for Refresh Fermi look closely at the ATI 6000 release but i doubt this will be a bit hitter in new tech, hope i am wrong due to we need Competition.

 

[Broken Hope]

Go Nvidia Fermi 400 series 470/480 or wait for the refresh 400 cards if you want the best at DX11/tessellation, and not forgetting 32AA ect.

if you do wait for Refresh Fermi look closely at the ATI 6000 release but i doubt this will be a bit hitter in new tech, hope i am wrong due to we need Competition.

Nvidia haven't even got their full fermi line out yet and you seem to think they are going to be able to refresh fermi on the 40nm process when they are already maxed out on it? I'd love to see how they are going to refresh when they are already too big and too hot on the current process. Wishful thinking.

 

[REAPERUK]

Nvidia haven't even got their full fermi line out yet and you seem to think they are going to be able to refresh fermi on the 40nm process when they are already maxed out on it? I'd love to see how they are going to refresh when they are already too big and too hot on the current process. Wishful thinking.

yea you probably got a point, looks like the 470/480 is the only way forward if you want current best fastest tech.

As for heat on load the Fermi 480's run as hot as a 280 at load but run cooler on Idle than the 280..