Gtx 480 Unigine Video (and benchmark)

A

Asacolips

Limp Gawd
Joined
Dec 4, 2008
Messages
272
I'm not sure if anyone else noticed this, but apparently Nvidia put up a Unigine demo on their youtube account. From Firing Squad

Most interesting in here to me is the benchmark of the 5870 vs the Gtx 480 at about 3:01. During normal segments of the demo they perform very closely. However, when tessellation kicks in, Nvidia seems to have a pretty big advantage. That's basically what was suggested by their whitepapers a few months ago, but it's neat nonetheless.

Anyways, I'm quite interested in that the tessellation performance looks as good as it does. I personally feel that will eventually be as big of a shift as normal mapping was.

Anyways, thoughts?
 
Last edited:
Correct me if I'm wrong, but that demo wasn't run with AA turned up, probably because of this:

http://www.hardforum.com/showpost.php?p=1035410551&postcount=83

Seems as if a GTX 470 uses up some of its computing power for tessellation which impacts its AA, so that high levels of AA it starts losing to 5870's, at least on averages. As for minimum fps, which I think are more important, it's unclear how much of that edge Fermi retains when AA is turned on. GTX 480 may be fast enough to edge out 5870, but at what price?
 
I do kind of doubt that any AA was used in that video. Seems like performance would have been lower across the board if it had been on. As for the weird AA performance behavior, I kind of wonder if that may be resolved in later driver updates, provided it is an issue in actual games an not just the Unigine benchmark.
 
XtremeSystems also reported they are working a heaven benchmark 1.1, not the 1.0 version that we all have now.

Also, something about nvidia "cheating" by having all its processing shifting towards tesselation in the benchmark, whereas in real games it wouldn't be able to do that because there is much more going on.

I dunno, just hope it brings down prices for us all.
 
is it possible that shaders doing tessellation are taking a hit when you turn on AA, I mean is it possible for ati to do shader based tessellation vs. built in tessellation engine, I am sure they could do it, but it must have its disadvantage as well as advantages.
 
Im sorry, but this GTX 480 just doesnt impress at all. I mean, take the Unigine benchy...30fps avg? Granted its at 1080p...but who doesnt play at this to begin with? What kind of scores is this thing going to get at 1080p doing tessellation in an ACTUAL GAME? Plus, just based on past Nvidia experience, whose to say that their drivers arent modified in any way? I seem to be seeing that all they run to show tessellation is Heaven benchmark...what about DiRT 2 or Battleforge? Umm...Alien vs. Predator? STALKER CoP?

Now Im not saying that modified drivers are the case, but something just dont seem right...theres a lot more demos they could be showing...not only that...but this is supposed to be an AWESOME card right? 30 fps no AA at 1080p and only 30fps avg. in the Heaven benchmark? Even if the 5870 cant do better, thats hardly impressive.

Also, SLI needed for Stereo Vision? Sigh...no comment...

PS - http://www.pcgameshardware.com/aid,...chmarks-Top-article-of-October-2009/Practice/

5870 with tessellation with 4x AA and avg 31fps...again...even if the GTX480 is faster, its no where near impressive (again no 4x AA), at least to me...

Lets not forget to add that the Heaven benchmark seems to be pretty incompetent...the DX10 render is laughably ugly and completely unrealistic of what DX10 is capable of...
 
Last edited:
if it's better then it's better.
SLI or Crossfire IS needed at high resolutions for a select few games, if you want to use high settings/max settings. I agree though that at this point it's pretty pointless estimating performance prior to information other than PR stuff is released.
 
I just loaded the 720 video and froze the frame @ 1:05 minutes and NO AA is being used. The graph does show a considerable lead though. Still if its only performing on par without using tessellation and is going to be 200 dollars more, well that is going to be a hard sell. but let us remember that drivers are going to be rough for a while here and I am sure that this is not going to be the end result
 
Who cares about tesselation? Because games wont benefit of tesselation in a long time.
It's more important how it performs without tesselation, and GTX480 performs equal with HD5870.
So Fermi seems to be a huge fail.
 
Blackbird said:
Who cares about tesselation? Because games wont benefit of tesselation in a long time.
It's more important how it performs without tesselation, and GTX480 performs equal with HD5870.
So Fermi seems to be a huge fail.
Click to expand...

I wouldnt even go as far as to say its a complete failure...you know as well as I do that itll sell, and if its as fast or slightly faster than the 5870 then I guess Nvidia comes out as a draw/slight win for them.

I can say its a failure price wise...do we know the price? No...will it be most likely ~$500-600? Most definitely, thats a failure in my opinion. Only people who will buy it are Nvidia fans or are just plain crazy about F@H, those are the only two reasons I see it selling, not because of performance increase or price increase over the 5870...
 
Ok, I've seen this earlier when it was posted in the Nvidia forums and it's probably been asked but has anyone with a similar configuration as was tested in the video but with a 5870 have results matching Nvidia's results during the test? With and without AA as well? That would be a better comparison.

Example: The video was a test of the 480GTX at 1080p with no AA, what would be another person's results with a 5870 with no AA. Are they the same as Nvidia has tested it? And, with AA, are the results similar to that video as well?

That would give a probable estimate on how the card performs, though not in a real world situation of course.
 
AndroidVageta said:
I wouldnt even go as far as to say its a complete failure...you know as well as I do that itll sell, and if its as fast or slightly faster than the 5870 then I guess Nvidia comes out as a draw/slight win for them.

I can say its a failure price wise...do we know the price? No...will it be most likely ~$500-600? Most definitely, thats a failure in my opinion. Only people who will buy it are Nvidia fans or are just plain crazy about F@H, those are the only two reasons I see it selling, not because of performance increase or price increase over the 5870...
Click to expand...

I might buy it, but that is because I want to run 3 monitors at 5760x1200. At that resolution, I am going to want either SLI or Crossfire. Now its been 6 months or so and ATI still doesn't have crossfire eyefinity working, except with 3 games.... One of Nvidias stronger areas has always been drivers, so while I might wait awhile for the initial prices to come down, I will probably go with the GTX 480's if their surround vision drivers are better than ATI.
 
octoberasian said:
Ok, I've seen this earlier when it was posted in the Nvidia forums and it's probably been asked but has anyone with a similar configuration as was tested in the video but with a 5870 have results matching Nvidia's results during the test? With and without AA as well? That would be a better comparison.

Example: The video was a test of the 480GTX at 1080p with no AA, what would be another person's results with a 5870 with no AA. Are they the same as Nvidia has tested it? And, with AA, are the results similar to that video as well?

That would give a probable estimate on how the card performs, though not in a real world situation of course.
Click to expand...

This is something that I too am extremely interested in seeing...its not that I have no trust for Nvidia...but whose to say that it wasnt done differently for the ATI card...like using old drivers or something?
 
AndroidVageta said:
This is something that I too am extremely interested in seeing...its not that I have no trust for Nvidia...but whose to say that it wasnt done differently for the ATI card...like using old drivers or something?
Click to expand...

That's my thought too.

And, I know both camps-- AMD and Nvidia-- like to show each other who performs better than the other by "tweaking" certain variables to make it look good on paper.

But, until we get as close to unbiased testing done by Kyle and the rest of [H] and other groups, we'll have to make do with what Nvidia is showing in the video. And, to do that, it would be nice to see the results from another person with similar setup-- 5870, same Core i7 as Nvidia's and same motherboard with CPU clocked at same speed and memory at 6 GB DDR3. This is [H], someone must have a comparable setup. :D
 
We all know that the [H]ard and proper way of doing things is by using real games, not artificial benchmarks that can be optimized for certain routines but would not reflect real world gaming.

All Nvidia has been doing since the very first videos of Fermi have been the Unigen benchmarks. Is that maybe the only thing that it's good at?
 
20100305heaven.jpg

X-2010012014384714655.jpg

Note 60-120 seconds...

Wow... cherry picking anyone?
 
Valset said:
I just loaded the 720 video and froze the frame @ 1:05 minutes and NO AA is being used. The graph does show a considerable lead though. Still if its only performing on par without using tessellation and is going to be 200 dollars more, well that is going to be a hard sell. but let us remember that drivers are going to be rough for a while here and I am sure that this is not going to be the end result
Click to expand...

I noticed that anistropy was set to 1 on the video. Default is 4.
 
Hmm... this does not appear to be the 20% more performance they were talking of... Unless that's something for just games using tesselation and without AA. It seems pretty on-par with the 5870 for normal gaming so far. The only reason I'm strongly considering purchasing one is Folding@Home. I'm really hoping a GTX 470 will net around 16k PPD. If it does, I can turn the heat off :)
 
ZerazaX said:
Note 60-120 seconds...

Wow... cherry picking anyone?
Click to expand...

Yep, for anyone who doesn't see it right away:

cherrypickingfermi.jpg


But the entire benchmark (and their overreliance on Unigen on practically everything) is a big fat cherry pick. It's probably the only thing they show because it's the only thing they can show any appreciable gain (every 60 seconds) in during heavy tesselation in an artificial benchmark. If Fermi was kicking ass in Crysis Warhead or leaping ahead of 5870 by leaps and bounds in Dirt 2, you'd think they would show those. Also interesting how all the comparisons are to the 5870 which isn't even ATI's halo card whereas Fermi is purporting to be the king of the hill.
 
Last edited:
ZerazaX said:
Note 60-120 seconds...

Wow... cherry picking anyone?
Click to expand...
Duh, it was noted at the time it was probably a PR cherry pick based on the sequence they used. Seriously though, PR numbers are as useful as the back of the box for picking a graphics card.
 
The guy in the video mentioned at the beginning of the video there was no tessellation and both the GTX 480 and 5870 seemed to have very similar FPS in the 50 to 60 range during the first 30 seconds or so.

I'm curious if this means that in older DX9 or DX10 games not using tessellation Nvidia just showed us that the 480 is almost no faster than the 5870? Or am I'm over simplifying it?
 
Ravynmagi said:
The guy in the video mentioned at the beginning of the video there was no tessellation and both the GTX 480 and 5870 seemed to have very similar FPS in the 50 to 60 range during the first 30 seconds or so.

I'm curious if this means that in older DX9 or DX10 games not using tessellation Nvidia just showed us that the 480 is almost no faster than the 5870? Or am I'm over simplifying it?
Click to expand...

I would say that's correct. I think the 480 will only be faster in specific applications. I'm really disappointed because I need to upgrade my dying GTS 512, and I planned on buying their flagship. But, not if it doesn't preform any better than that. I'm 90% sure that my next card will be my first ATI card (5870) in a few weeks.
 
Ravynmagi said:
I'm curious if this means that in older DX9 or DX10 games not using tessellation Nvidia just showed us that the 480 is almost no faster than the 5870? Or am I'm over simplifying it?
Click to expand...

I believe that's what we will see in real reviews come March 26:th.
 
Ravynmagi said:
The guy in the video mentioned at the beginning of the video there was no tessellation and both the GTX 480 and 5870 seemed to have very similar FPS in the 50 to 60 range during the first 30 seconds or so.

I'm curious if this means that in older DX9 or DX10 games not using tessellation Nvidia just showed us that the 480 is almost no faster than the 5870? Or am I'm over simplifying it?
Click to expand...

You aren't over simplifying it. However, how many DX9 games are there that you need more frames on anyways? If the 480 or the 5870 gets more frames in TF2 it isn't going to affect my buying decision as both will be getting 200+ FPS in that game.

At the same time, I hope Nvidia doesn't think that a lead in tesselation is enough to justify a huge difference in price. The closer we get, the more I'm thinking we might see a 400$ 480 GTX.
 
At the same time, I hope Nvidia doesn't think that a lead in tesselation is enough to justify a huge difference in price. The closer we get, the more I'm thinking we might see a 400$ 480 GTX.
Click to expand...

They would have to be $400 to sell, but there's no fucking way. Not with yields as bad as they reportedly are.
 
So I haven't really been paying attention to the fermi posts since most are just opinions, but if what im seeing is right, might fermi be the way to go for DX 11? Seems like it to me.

Guess we will have to wait and see.
 
Casper312 said:
So I haven't really been paying attention to the fermi posts since most are just opinions, but if what im seeing is right, might fermi be the way to go for DX 11? Seems like it to me.

Guess we will have to wait and see.
Click to expand...

no that's not right, Fermi will be the way to go if you play games that do heavy tessellation
 
dook43 said:
They would have to be $400 to sell, but there's no fucking way. Not with yields as bad as they reportedly are.
Click to expand...

You are assuming that they are going to make a profit on the 480 at launch, that the 5870 is making a reasonable profit (as opposed to an outlandish one), and the yields are as bad as reported by Charlie.

If I were Nvidia and I only had 5000 to launch as some places are reporting, I'd say screw it and launch them at 399$. If they eat 100$ per card it would be 0.05% of one quater's revenue that is lost. Well worth it for a PR move.

Of course for the 5000 cards at launch rumors to be true, their yields have to be sub 1% as they supposedly hot lotted 9000 wafers.

The rumors just don't add up.
 
its to early to tell if fermi can equal or better the 5870. a valid 3rd party benchmark is all it would take. till then this is just more PR garbage i hate nvidia for.
 
Casper312 said:
So I haven't really been paying attention to the fermi posts since most are just opinions, but if what im seeing is right, might fermi be the way to go for DX 11? Seems like it to me.

Guess we will have to wait and see.
Click to expand...

From everything we've seen from Nvidia. Fermi is the way to go only if the only games you play are heavily tesselated artificial benchmarks instead of actual games. Wait and see...
 
So it performs similar to a HD5870. Meh I'm ok with that honestly the only thing I care about is that it actually supports Multi-monitor 3Dvision I realize I'm the only guy on the forums that cares about this but HOLYSHIT! OMG OMG OMG WANT. :( I actually like 3Dvision.
 
vengence said:
You are assuming that they are going to make a profit on the 480 at launch, that the 5870 is making a reasonable profit (as opposed to an outlandish one), and the yields are as bad as reported by Charlie.

If I were Nvidia and I only had 5000 to launch as some places are reporting, I'd say screw it and launch them at 399$. If they eat 100$ per card it would be 0.05% of one quater's revenue that is lost. Well worth it for a PR move.

Of course for the 5000 cards at launch rumors to be true, their yields have to be sub 1% as they supposedly hot lotted 9000 wafers.

The rumors just don't add up.
Click to expand...

I'm hoping this is true. I mean 5 months ago $600 sure but they know it's late to the market and they'll account for that I'm hoping.
 
piscian18 said:
So it performs similar to a HD5870. Meh I'm ok with that honestly the only thing I care about is that it actually supports Multi-monitor 3Dvision I realize I'm the only guy on the forums that cares about this but HOLYSHIT! OMG OMG OMG WANT. :( I actually like 3Dvision.
Click to expand...

That's the only thing Nvidia has over AMD aside from whatever cuda/physx stuff that doesn't really do much for gaming.

3DVision is actually a very fun thing and is sort of the forgotten facet of gaming that never really caught on while Eyefinity is all the rage.

The problem is cost. Multimonitor 3D vision requires basically TWO fermi cards (as Fermi does not support more than 2 monitors per card), and at minimum THREE 120Hz monitors. Each of those individual components is basically $600-700.

$700 x 5 = $3500!!!
 
3DVision is actually a very fun thing and is sort of the forgotten facet of gaming that never really caught on while Eyefinity is all the rage.
Click to expand...
Well 3D is getting FULL rage now from pretty much every company out there so 3d vision is sure to pick up.
The only advantage flatfinity has is you can see it in youtube and get an idea of how it looks as opposed to 3d vision ;)


First I do not care about all the non sense silly threads going around. I will wait and see how it actually performs instead of going by all the "previews" that people post every day.
But let's say it performs just a bit better than a 5870? Are ATI fanboys forgetting that the 5870 came AFTER the GTX295 and was beaten by it in many games? Yeah dual vs single, whatever, but the fact still remains.
I do not care about ATI until they open their eyes and join the 3d revolution. If they do then I will look at their cards and select which ever brand offers me the best deal. In the mean time, I cannot care less if an ATI card can give me 1000fps in Crysis if it does not offer me REAL 3D gaming not just multi monitor stitching aka Flatfinity.
Regards
 
vengence said:
If I were Nvidia and I only had 5000 to launch as some places are reporting, I'd say screw it and launch them at 399$. If they eat 100$ per card it would be 0.05% of one quater's revenue that is lost. Well worth it for a PR move.
Click to expand...
If there are only 5000 cards at launch, NV would be silly to price them at $399. At $399, flippers will buy all of them up in the first 5 minutes, and then put them on eBay for $800. I think there are more than enough NV fans and f@h enthusiasts for NV to sell 5000 GF100 cards at any sub-$1000 price.

Assuming that there are only 5000 cards at launch, pricing them at $399 be a fat gift to eBay flippers, with hardly any positive PR value (because potential buyers would be mad that flippers got all the cards).

vengence said:
Of course for the 5000 cards at launch rumors to be true, their yields have to be sub 1% as they supposedly hot lotted 9000 wafers.
Click to expand...
My guess is that not all 9000 ordered wafers are produced at the same time, so the 5000 card rumor, if true, may be only the launch shipment quantity, with more to come in the future. Another guess is that most of the good chips are allotted to the HPC products, which bring in much more profit than gaming products.

But of course these are only guesses.
 
shaolin95 said:
Well 3D is getting FULL rage now from pretty much every company out there so 3d vision is sure to pick up.
The only advantage flatfinity has is you can see it in youtube and get an idea of how it looks as opposed to 3d vision ;)


First I do not care about all the non sense silly threads going around. I will wait and see how it actually performs instead of going by all the "previews" that people post every day.
But let's say it performs just a bit better than a 5870? Are ATI fanboys forgetting that the 5870 came AFTER the GTX295 and was beaten by it in many games? Yeah dual vs single, whatever, but the fact still remains.
I do not care about ATI until they open their eyes and join the 3d revolution. If they do then I will look at their cards and select which ever brand offers me the best deal. In the mean time, I cannot care less if an ATI card can give me 1000fps in Crysis if it does not offer me REAL 3D gaming not just multi monitor stitching aka Flatfinity.
Regards
Click to expand...
You are intitled to your opinon. And we listened... for a while. You've just become annoying now. You get off on 3D, we get it. If you'd like to have a circle jerk about 3D vision please go start a thread about it and discuss with your self to your hearts content.
 
chklin said:
If there are only 5000 cards at launch, NV would be silly to price them at $399. At $399, flippers will buy all of them up in the first 5 minutes, and then put them on eBay for $800. I think there are more than enough NV fans and f@h enthusiasts for NV to sell 5000 GF100 cards at any sub-$1000 price.

Assuming that there are only 5000 cards at launch, pricing them at $399 be a fat gift to eBay flippers, with hardly any positive PR value (because potential buyers would be mad that flippers got all the cards).
Click to expand...
People are allways pissed about scalpers. But they are mad at scalpers, not at the company selling the cards. 4870x2s were scalped for a while but people kept praising ATI. Nvidia wouldn't care as long as it can say it has a faster cheaper card even if it doesn't have any on the shelves.
 
You must log in or register to reply here.
Back
Top