no its not going to be helped. The 9800Gt just too slow basically. You'll need something like a Gtx 460 to use as a dedicated physx cards to see a difference in the fps. Assigning a dedicated card is a good idea but using a high end card with a slow one is not going to help. you'll see no fps improvement
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
9800GT PhysX card with 580 GTX GPU: The heat isn't worth it!
- Thread starter OldM3ta
- Start date
no its not going to be helped. The 9800Gt just too slow basically. You'll need something like a Gtx 460 to use as a dedicated physx cards to see a difference in the fps. Assigning a dedicated card is a good idea but using a high end card with a slow one is not going to help. you'll see no fps improvement
nxcess
2[H]4U
- Joined
- Dec 22, 2004
- Messages
- 2,706
Rarely do trolls make sense.
LGabrielPhoto
2[H]4U
- Joined
- Jan 5, 2006
- Messages
- 3,240
Where exactly did I mention a 9800GT?
I use a gts250 with my 570 running 950 core btw and it helps since when running 3D vision the card is pushed VERY hard.
Last edited:
What's the difference? The 9800 is a rebranded gts 250. If you want to use a card as a physx card use one one step below. Eg a 570 you would use a 460 or a card that won't bottleneck your system. A 250 will bottleneck a 570. using a dedicated physx card can be useful but not using it with a card that will be a bottleneck.
InvisiBill
2[H]4U
- Joined
- Jan 2, 2003
- Messages
- 2,608
If you want to get technical, going from a GTX275 to a GTX285 for PhysX increased my framerate with only a HD5770 for main graphics. So a GTX275 will bottleneck a HD5770. I assume I could've shown the same thing with the GTX285 if I had a higher card to test with. http://hardforum.com/showthread.php?t=1510323
My 9800GT along with my HD5870 (between the GTX560 and GTX570 on Tom's chart) gives me 100fps average, 60fps minimum. With a minimum that's as high as my VSync, I wouldn't even notice anything faster. In my system, a 9800GT is good enough to keep the bottleneck above where I could see it. When I left PhysX set to high but disabled the card, I got an average of ~18fps and a minimum of ~10fps.
With PhysX effects completely disabled, I got nearly twice the framerate. We have a long way to go before any single GPU running PhysX doesn't bottleneck any high-performance GPU at least a little.
A Gtx 285 is faster than a 5770 it can't be bottleneck nor a 275. The post title say Gtx 580 with a 9800gt aka gts 250. It will bottleneck a Gtx 580 plain and simple. If you go look in tomshardware forums there somewhere someone posted a Gtx 460 used a dedicated physx card and the max fps was 20fps. remember you can't go throw ATI cards in the mix coz physx is a nvidia feature. ATI cards physx wouldve normally been done by the cpu where nvidia cards the gpu will do it. So you can't compare the two and say you got a massive fps increase with a ATI card using a dedicated physx you would get the same using a nvidia card coz the nvidia card the gpu will do the physx. That's why people see like 5fps increase with the cost of more heat and power consumption which in the end using a high end gpu ain't worth it as its capable of running the game perfectly well and doing the physx. Show me where you got a massive fps increase with a Gtx 9800Gt running it on a 570 that's got the physx feature of nvidia not ATI cards which the cpu has done the calculations. Off course your going to see big fps increase with it. But this post is a nvidia card with a nvidia card where both have the capabilities of doing physx.
texuspete00
Supreme [H]ardness
- Joined
- Sep 9, 2002
- Messages
- 5,625
They certainly went up with my 480 and a 9800GT. Batman dipped, but the fps was already very high.
EricFX1984
[H]ard|Gawd
- Joined
- May 21, 2010
- Messages
- 1,122
I would get it out of there.. the 9800gt is not worth the space it is using...
LGabrielPhoto
2[H]4U
- Joined
- Jan 5, 2006
- Messages
- 3,240
If you want to try to be technical then it is equal to a 9800GTX+ first of all.
Second I own both a GTS 250 and a 570 and IT DOES IMPROVE the FPS because my GPU is already working extremely hard doing 3d vision so I have it, I tested it , and it helps so stop misinforming people with your theories when I , and many others, have actually done it instead of just taking based on zero experience
If you want to try to be technical then it is equal to a 9800GTX+ first of all.
Second I own both a GTS 250 and a 570 and IT DOES IMPROVE the FPS because my GPU is already working extremely hard doing 3d vision so I have it, I tested it , and it helps so stop misinforming people with your theories when I , and many others, have actually done it instead of just taking based on zero experience
Hey can you tell me how many FPS increase you get on any games with that setup. Because I'm probably going to get a 560 Ti 2Gb card , and use my 260GTX for Phsyix. From what I have read I can expect 5-7 FPS increase, that doesn't sound like alot, but then again you get all those particles in games that support it and slightly higher fps so that's cool. Again how much performance increase have you noticed? Any hard numbers or just like average numbers? Hopefully both. What about Metro if you've tried it?
Last edited:
elm669
2[H]4U
- Joined
- Jul 8, 2003
- Messages
- 2,463
Yup, I still after 25 years on this planet don't understand the imperial system. wth is 5/64 of a inch or 13/16 anyways of a inch, fractions make no sense for measurements.. if you told me about 2mm or about 20mm I would understand without having to look it up. Someone tells me to get a 1/4 wrench or socket, first thing I do it convert to metric and find the nearest size wrench. lol The whole of the Imperial system makes no sense I tell ya, just like water freezes at 32F and boils at 212F.. why no boil at 100C or freeze at 0C.. THAT makes sense.
/signed metric convert
update: Just tried my old 9800GT (G92) with my 560Ti.. I LOST fps
Last edited:
LGabrielPhoto
2[H]4U
- Joined
- Jan 5, 2006
- Messages
- 3,240
Again this is going to depend on your resolution, graphic settings and even game.
In my case a 570/480 both get help by ,my 250 because of how hard 3D vision is at 1080p with some games.
I will try to post some scores when I get back home.
YamahaAlex37
[H]ard|Gawd
- Joined
- Jun 23, 2005
- Messages
- 1,775
Where the hell are you getting this information. Before I popped a GTS 250 with my 570, UT3 Tornado was UNPLAYABLE, now it plays just fine 30+ fps. I'm pretty sure a 9800GT would also help just as much, and if you want me to prove it I'll put my brothers in here.
I remember when researching Phys-X to know if it would be worth it I found plenty of information that proved a 9800gt is the choice Phys-X card.
Also, for some of us with high end cards like the 570 and dual monitors, the GPU will refuse to leave performance mode while powering multiple monitors. The only way to fix this was to stick another card in there and it DECREASED the temps by at least 20.
cannondale06
[H]F Junkie
- Joined
- Nov 27, 2007
- Messages
- 16,180
well he is sort of right. the 9800gt would be an okay physx card with a gtx260/275/280/285 level of gpu but would not help nearly as much with a very high end gpu like a gtx570/580. a gtx570 would probably still get close to the same fps handling both graphics and physx as opposed to letting a 9800gt handle physx. it really just depends on the game and settings being used.
pigwalk
[H]F Junkie
- Joined
- Dec 14, 2004
- Messages
- 8,396
I just take out the 9600gt I use for PhysX when I'm not playing a game that makes use of it. I'll drop it back in when I start Batman:AA and then out again.
cannondale06
[H]F Junkie
- Joined
- Nov 27, 2007
- Messages
- 16,180
you only have a gtx260 so of course it will help quite a bit. using a 9600gt with a gtx570 or 580 would likely be slower then letting a gpu like that handle both graphics and physx though.
spacing guild
[H]ard|Gawd
- Joined
- Dec 9, 2010
- Messages
- 1,356
+1 Just added a low-power 9800 GT to my new Sandy Bridge build with GTX 460. Lost 3000 points in 3D Vantage. It sucks bc I was saving the 9800 for that purpose. Not worried about heat. Antec one-hundred on the way...
Besides the obvious, anyone have any ideas about why the score dropped?
Went from P17,000 to P14,000.
Using Core 2500k stock, P67 motherboard.
I think you guys should be trying this on a GAME. Like Batman AA, Mafia 2, or Metro 2033. I have personally tried the Mafia 2 demo and it had the Physic Benchmark, same with Batman AA. I haven't tried Metro 2033 yet though.
I don't believe in any Gaming Benchmarks (3D Vantage etc) that don't run on an actual game.
That's just a bunch of e-peen baloney, b.s.-marketing, lame gimmicks to try and get all the math nerds excited, and waste their dough on a graphics card that doesn't even perform in my eyes.
InvisiBill
2[H]4U
- Joined
- Jan 2, 2003
- Messages
- 2,608
I'm not familiar with the details of how they do PhysX testing, but a GTX460 doing just PhysX will be faster than a 9800GT doing just PhysX. However, if you're playing a game where the graphics are using a lot of the GTX460's power, then adding a 9800GT dedicated to PhysX may be faster overall.
In other words...
PhysX test: GTX460 > 9800GT
Gaming: GTX460 < GTX460+9800GT (possibly - I haven't done any testing on that)
InvisiBill
2[H]4U
- Joined
- Jan 2, 2003
- Messages
- 2,608
Doing only graphics, a GTX275/285 is faster than the HD5770. However, the HD5770 doing just graphics is faster than the GTX275/285 doing just PhysX. Without PhysX, the HD5770 averaged 125fps. With the GTX285 doing PhysX also, it dropped to 82fps. Adding the GTX285 dropped my framerate, therefore it is by definition a bottleneck. It was doing a lot more and making the game look a lot better, but it did drop the framerate in doing so. For the GTX285 to not be a bottleneck, I would have to get that same 125fps that I got without it. The GTX285 doing PhysX is slowing down the HD5770 doing graphics.
I mentioned my HD5870 because that's what I have. My point was only that a GTX560/570-level card will not be horribly bottlenecked by a 9800GT PhysX card. My 9800GT does PhysX well enough that I got 100fps average and 60fps minimum. With VSync on a 60Hz panel, I wouldn't be able to see any difference at all going with a faster PhysX card. This has nothing to do with whether you should or shouldn't add a 9800GT as a PhysX card to go along with your GTX570, only whether or not it will be a huge bottleneck (it won't, according to the numbers). If someone only got 20fps with a GTX460 dedicated to PhysX, their setup and/or settings are massively different from mine, or they did something wrong.
I only brought the HD5870 into the discussion because it's comparable to the GTX570 (especially if part of the GTX570's resources are going toward PhysX rather than just graphics). I've never compared the boost gained by adding a dedicated PhysX card to an ATI GPU vs. its performance doing PhysX on the CPU. My comparison is only between a 9800GT's PhysX capabilities and a GTX570-level card's graphical capabilities. In my testing, the 9800GT can do PhysX well enough to maintain 60fps minimum, and therefore will not visibly bottleneck my HD5870 setup (which should be similar to a GTX570). As shaolin95 has said repeatedly in this thread, exact results will depend on a lot of variables.
I never stated that you'd see a massive improvement by adding a 9800GT as a dedicated PhysX card along with a GTX570. In fact, I've been saying for a long time with every post in this section that you should get the most powerful PhysX card you're willing to spend money on. Physics calculations scale well, so basically more GPU == more frames, with almost no limit on that. For example, there's still room for improvement even with a HD5770 for graphics and a GTX285 for PhysX (but you're paying a lot of money for a tiny bit of improvement).
Based on my HD5870+9800GT experience, I'd say that you'd probably get almost exactly the same results using a GTX570 vs. GTX570+9800GT. I decided that my 9800GT was "good enough" for my HD5870 and sold my GTX285 (I wanted the $200 more than the 1-7fps increase), but I wouldn't go any lower nor would I specifically purchase a 9800GT to use for PhysX with any modern high-end GPU.
cannondale06
[H]F Junkie
- Joined
- Nov 27, 2007
- Messages
- 16,180
no those are completely different scenarios. with the AMD card your only other option was the cpu so even a low end gpu would give you a massive boost. a gtx570 is capable of doing BOTH physx and graphics where again the AMD card was not. so the faster the Nvidia card you have for graphics then the faster your dedicated physx card needs to be. a 9800gt for physx would be a massive help if using a gtx275 or so for graphics. if using a gtx570 for graphics though the 9800gt is not going to perform much better than the the gtx570 would doing both graphics physx. I have already mentioned this earlier.
InvisiBill
2[H]4U
- Joined
- Jan 2, 2003
- Messages
- 2,608
I'm well aware that the GTX570 can do both graphics and PhysX. I ran PhysX on my 8600GT, 9800GT, and GTX285 before getting my HD5870.
The same principles apply whether you're doing graphics on the same Nvidia GPU, a separate Nvidia GPU, or an ATI GPU. If you don't have enough PhysX processing power, it will bottleneck your graphics. If your main GPU is ATI, then yes, just about any Nvidia GPU dedicated to PhysX will be faster than CPU PhysX. However, GPU PhysX with a low-end card will still be much slower than the ATI alone with no PhysX. You're getting pretty fog and breakable stuff, but you're also lowering your framerate. For example, the HD5770 with PhysX turned off averaged 125fps. With the 9800GT for PhysX, it averaged 76fps. Going up to the GTX275 for PhysX, it went up to 79fps. With the GTX285, it was up to 82fps.
PhysX is very intensive. Even a GTX285 can't process PhysX fast enough to keep up with the HD5770 doing just graphics (82fps vs. 125fps). If your PhysX card isn't fast enough, it'll bottleneck your graphics, regardless of what GPU you're using for the graphics.
That's exactly what I said in my last post.
cannondale06
[H]F Junkie
- Joined
- Nov 27, 2007
- Messages
- 16,180
NO the same principles do NOT apply to what I am saying. when you have an AMD card as your gpu and try to run hardware physx it will run on the cpu which is really slow in most cases. adding even an 8600gt would help in that case since it can handle hardware physx better than a cpu. a 9800gt would of course handle it a little better but even the 8600gt would give a large increase over letting the cpu handle it.
now IF you had a gtx570 it can already handle physx while also handling graphics. it will handle both of those better than letting an 8600gt handle physx. something like a 9800gt would be the very very bare minimum to use for a physx card with a gtx570 or you might as well let the gtx570 do both. and really the 9800gt as a physx card may only just match or in some rare case not even match what the gtx570 could do while handling both graphics and physx.
those are NOT the same overall scenarios because the AMD card could never run physx in the first place so any card capable of running physx would help. it does not matter if its a 4870 or 5870 for graphics because its the cpu that is handling the physx so even something like an 8600gt would help tremendously over using the cpu. with an Nvidia card for graphics, the faster it is then the faster the gpu for handling physx needs to be.
now IF you had a gtx570 it can already handle physx while also handling graphics. it will handle both of those better than letting an 8600gt handle physx. something like a 9800gt would be the very very bare minimum to use for a physx card with a gtx570 or you might as well let the gtx570 do both. and really the 9800gt as a physx card may only just match or in some rare case not even match what the gtx570 could do while handling both graphics and physx.
those are NOT the same overall scenarios because the AMD card could never run physx in the first place so any card capable of running physx would help. it does not matter if its a 4870 or 5870 for graphics because its the cpu that is handling the physx so even something like an 8600gt would help tremendously over using the cpu. with an Nvidia card for graphics, the faster it is then the faster the gpu for handling physx needs to be.
Last edited:
orangeperoxide
n00b
- Joined
- Mar 29, 2011
- Messages
- 25
60 degrees on load isnt that hot.
my gpu along with a pound of dust goes to 100degrees at load.
my gpu along with a pound of dust goes to 100degrees at load.
InvisiBill
2[H]4U
- Joined
- Jan 2, 2003
- Messages
- 2,608
I agree completely. I've included my quotes agreeing with those points:
- Any dedicated PhysX GPU will be faster than doing PhysX on the CPU with your ATI GPU. "If your main GPU is ATI, then yes, just about any Nvidia GPU dedicated to PhysX will be faster than CPU PhysX."
- A low end PhysX card will not improve on having only a single GTX570 doing both graphics and PhysX. "I'd say that you'd probably get almost exactly the same results using a GTX570 vs. GTX570+9800GT."
What you're missing is that the ATI GPU is a certain speed without using PhysX at all. Not on a GPU or on the CPU. Turned off. No PhysX. None. For example, my HD5870 with no PhysX effects in Batman:AA averaged 217fps. The speed of my HD5870 doing graphics is 217fps. This number has absolutely nothing to do with PhysX whatsoever.
When I add the 9800GT for PhysX, I get fancy PhysX effects and average 98fps. That's much higher than the 18fps I get when I turn PhysX on high without an Nvidia GPU, but it's still much slower than the 217fps that my HD5870 is capable of. If I replace the 9800GT with a GTX275, my average jumps up to 105fps. With the GTX285, I get 108fps. As I get faster PhysX GPUs, I get higher framerates. The PhysX card is bottlenecking the system (the drop from 217fps to 108fps), and getting a better PhysX card reduces that bottleneck.
The Nvidia GPU is much faster than doing PhysX on the CPU. However, a weak PhysX GPU will cause your ATI card to be much slower than your ATI alone with no PhysX effects. If you're used to 200fps on your ATI GPU without PhysX, you can't throw in an 8600GT for PhysX and expect to still get 200fps. Since it only has 32 shaders compared to the 9800GT's 112, I'd probably get around 28fps (112:98 == 32:28). I'd test this, but the friend that bought my 8600GT moved a thousand miles away.
Going from 18fps with CPU PhysX to 28fps with the 8600GT doing PhysX is an increase of over 50%. That's a lot of improvement from tossing in a cheap old card. However, most people probably wouldn't consider 28fps with PhysX a "tremendous" improvement compared to 217fps without PhysX.
If you want to improve on a single Nvidia card, you need a dedicated PhysX GPU that is more powerful than what the single card is currently providing for PhysX. Similarly, if you want decent framerates in your hybrid PhysX setup (comparable to your no-PhysX framerates), you need a PhysX GPU that can keep up with your main ATI GPU.
cannondale06
[H]F Junkie
- Joined
- Nov 27, 2007
- Messages
- 16,180
I am not missing the fact that you can run the game without physx at all. my whole point was about getting the most out of running hardware physx. and how would a weak gpu being used for physx cause your AMD card to be slower when not even running physx? that makes zero sense because if you are not using hardware physx then the card for dedicated physx is doing nothing at all. in other words you will still average the same fps in a game with hardware physx off whether or not there is a dedicated gpu for physx in your system.
Last edited:
Well lets have a look. A Gtx 480 with a 9800Gt.
Look at the fps increase with the Gtx 480 without the 9800Gt. But look at the fps increase when a faster card is used with it. Clearly shows that a slow card can become a bottleneck. a dedicated physx is a good idea but you got to match it up at least. Not any old card will give a boost and if you get a boost you got to compare performance increase vs power consumption and extra heat. In some cases is it worth it for a extra 2-3fps?
Btw physx will be done by the cpu on Amd gpu. Not nvidia physx but physics
Look at the fps increase with the Gtx 480 without the 9800Gt. But look at the fps increase when a faster card is used with it. Clearly shows that a slow card can become a bottleneck. a dedicated physx is a good idea but you got to match it up at least. Not any old card will give a boost and if you get a boost you got to compare performance increase vs power consumption and extra heat. In some cases is it worth it for a extra 2-3fps?
Btw physx will be done by the cpu on Amd gpu. Not nvidia physx but physics
the 275 have physx features that the 9800 don't. Using a Amd card with a nvidia card you'll off course see more fps because the cpu is awful with X87 and that's what basically going to happen.
But again this post is about a Gtx 580 with a 9800Gt which is absurd to do such a thing.
that is because when you turn physx on you are asking the cpu to do X87 physx calculations without the nvidia card present. Remember the assign physx to cpu/gpu feature mentioned in the post a couple of times? You know why that feature is there? Physx are X87. Normal physics in games all of them use physics. But that's in X86. In X87 a cpu are awful that's why nvidia can say a gpu is 17x times faster than a cpu. But with X86 a cpu does it actually just as fast and even faster in some occasions than a gpu. That's why when your not using a game with nvidia physx assign it over to the cpu not the nvidia gpu and check the fps again. Remember I'm talking about a game thats not nvidia physx branded.
No the 5770 is slowing down the 285 lol take out the 5770 and use the 275 as your physx card. Sell the 5770 on ebay Lmoa
InvisiBill
2[H]4U
- Joined
- Jan 2, 2003
- Messages
- 2,608
I never said that a weak PhysX GPU would slow you down when not running PhysX.
With an AMD GPU, CPU PhysX is not your only other option - there's also the option of no PhysX at all. My point was that someone who was used to getting 200fps without PhysX is not going to consider it "a massive boost" to get 28fps with PhysX by adding an 8600GT. The 8600GT will probably be faster than CPU PhysX, but it's nowhere near what they were getting before with no PhysX at all. If they want a decent PhysX experience with their AMD GPU, they still need to get an appropriate-level PhysX card.
Like what? I've never heard this before.
No, the extra PhysX load (even when being handled by a GTX285) is slowing down the HD5770 from 125fps to 82fps. The GTX285 doing PhysX is bottlenecking the HD5770 doing graphics. This should stress the point that it's nearly impossible to have too much PhysX power.
FYI, I borrowed the HD5770 and GTX275 specifically for testing, and sold the GTX285 a year ago.
this is what I meant. That's why games go into crawling mode fps wise when you don't have a nvidia gpu with a game that's got a physx patch
http://techreport.com/discussions.x/19216
one gpu waiting for another gpu? What do you call that. Physx processing done 285 waiting on the 5770 to get done? Probably why the decrease. Bet you wouldve gotten higher fps using a faster Amd gpu. If you did then all fingers point to the 5770
InvisiBill
2[H]4U
- Joined
- Jan 2, 2003
- Messages
- 2,608
I'm aware of that. In the followup stuff, Nvidia basically admitted that their CPU PhysX code sucked. Their reasoning was that the code is basically 10 years old, from before SSE was ubiquitous. They're supposedly in the process of rewriting it with modern CPUs in mind. Believe them or don't, whatever.
My test results when enabling High PhysX in-game but disabling the PhysX card also support this. Whereas the HD5870 was much faster than the HD5770 in the regular testing, they were almost exactly the same (both sucked horribly) when trying to do tons of PhysX on the CPU.
When one part has to wait on another slower part, I call the slower part a bottleneck.
The HD5770 was waiting on the GTX275 to finish its PhysX calculations. I proved this by upgrading the PhysX card to a GTX285 and getting higher framerates. The GTX275 dedicated to PhysX was a bottleneck, since replacing it with a faster PhysX card increased framerates.
I repeated this with the HD5870. Across the board, the HD5870's scores were higher than the HD5770's. However, I repeated the dedicated PhysX card test. Once again, replacing the GTX275 with the GTX285, with no other changes, increased the framerate. The GTX275 was bottlenecking the system.
I didn't have a card higher than the GTX285 to test with. I can't prove that going to a GTX470/480 would've increased my framerates more (which would indicate that the GTX285 was also still bottlenecking the main GPU). However, based on my experiences testing 9800GT -> GTX275 -> GTX285 and the fact that the framerates were still much lower than without any PhysX, I would expect there to also be an improvement going up to an even faster PhysX card.
I'm not saying that you should buy a $100 video card and a $500 PhysX card. I'm not even saying you should buy a $500 PhysX card to go with a $500 video card. What I am saying is that even with a crappy $100 video card, you'll probably still see a slight improvement going from a $450 PhysX card to a $500 PhysX card. If you want the absolute best PhysX performance, buy Nvidia's top of the line single-GPU card, regardless of what your main GPU is. And when Nvidia releases their newer, faster card, you'd probably see a slight improvement if you switched to that card for PhysX, even if your main GPU isn't all that great.