5870 vs 470 for an EVGA Fanboy

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Which one?

  • 5870

  • 470


Results are only viewable after voting.

Dark Shroud

Golden Member
Mar 26, 2010
1,576
1
0
Just a few questions.

How many people will play 3 displays? 6 displays?
I will probably not do 3 displays for a few years. Even though I could pull it off with a few old CRTs I have from work stations.

Who here say 5870 is better but believe that PhysX will have no future if it doesn't work on a ATI card?
I'll put it this way, PhysX won't have a future if Nvidia keeps blocking hardware accelration in hybrid environments. I would have bought an Nvidia Card to run PhysX for Batman & Dragon Age but I'll just let the Core i7 handle that now.

How many people are going to get a new 120 LCD in the near future?
My gaming rig still uses a 22" 85hz CRT. Right now I'll probably just buy a smaller TV either a true 240hz LED or a Plasma. Nothing bigger than 42." Hopefully a 3D capable model.

Who will likely to check out 3d Gaming in the near future?
Only if it's done by the screen and not some proprietary tech from Nvidia. It doesn't look like ATI will try and do it themselves.

Who hate the sound from PC, going for SLI/xfire with the newest, greatest, hotest video cards, yet not using liquid cooling?
I've never done an SLI/Crossfire set up before. But sound is important to me. I couldn't put up with a GTX 480 and to be fair the same goes for the 4800 X2 cards.
 

MrK6

Diamond Member
Aug 9, 2004
4,458
4
81
You'll get more efficiency & performance with the 5870 over the 470. But the 470 is a smoother more feature rich card that is slightly cheaper than the 5870. Tough call.
That's crap though. No card is "smoother" than the other. For every specific example of an AMD card giving a better gaming experience, I can give one of an NVIDIA card and vice versa. When comparing the two cards the main things one should at least consider are price, performance and overclocking, power usage/heat, fan noise/character, and features. For the average gamer, the 5870 is going to walk all over the GTX470. However, if someone is really into F@H, or uses CUDA, or has some other specific requirements, that's going to change the field.
Just a few questions.

How many people will play 3 displays? 6 displays?
Until they work on the bezels, not for awhile.
Who here say 5870 is better but believe that PhysX will have no future if it doesn't work on a ATI card?
GPU physics in general doesn't have a future until architectures change dramatically. Right now for games we generally have too much CPU power and not enough GPU power. Forcing the GPU to run even more effects is just stupid when there's idle CPU cores that could be used. Considering it's a niche market for gimmicky effects based on a proprietary solution right now, well, it's going to be even more difficult for it to make any headway.

How many people are going to get a new 120 LCD in the near future?
When they come out with a 30" true 120Hz LED backlit LCD at a decent price, I'll be interested.

Who will likely to check out 3d Gaming in the near future?
Already did, it's lame and a gimmick. We need better 3D technology (not the same stuff that's been recycled for the last 50 years) before it'll take off.

Who hate the sound from PC, going for SLI/xfire with the newest, greatest, hotest video cards, yet not using liquid cooling?
That's the thing, 5870's or 5850's in CF are still extremely quiet. From what I've read and heard in reports, the GTX4x0's aren't.
 

Mr. Pedantic

Diamond Member
Feb 14, 2010
5,027
0
76
However, if someone is really into F@H, or uses CUDA, or has some other specific requirements, that's going to change the field.
For F@H it's starting to work back towards CPUs - it's much more power- and money-efficient to get a Core i7 920, OC that to 3.6-3.8 and fold with that rather than buy a GPU. As for CUDA, I would make the argument that if you care that much to solely base the purchase of a $400-500 card on it, wouldn't you go the full monty and actually get a Tesla or a Quadro that has better drivers and better support?

Right now for games we generally have too much CPU power and not enough GPU power.
But then you also have people who run games on their old Athlon X2s, and who put 5850s and 5870s in with these systems, precisely because of what you said. I think games like BC2 are starting to tip the balance back and they're starting to make people realize that the CPU can make a huge difference to gameplay, but it won't happen overnight, and I'm skeptical as to whether GPU physics will ever fade away.

That's the thing, 5870's or 5850's in CF are still extremely quiet. From what I've read and heard in reports, the GTX4x0's aren't.
You're kidding, right? The Cypress cards are quieter than the Fermi ones, but they're still not 'extremely quiet'. In fact, I wouldn't use 'quiet' as a descriptor for any current reference cooler full stop.

I used to rag on Nvidia about their "proprietary features". But now I'm starting to respect them somewhat for this. It is their strong point. Gpu accelerated physics, in-game MSAA in Batman, their ray tracing, ocean, hair, & rocket sled demos, twimtbp titles, the advanced game profiles menu in their driver control panel.
I think the Batman AA thing is a bit cheap and underhand. But NVidia deserve respect for all the other software features that they have been working on. They poured a lot of money into PhysX, GPU ray tracing and things like that, and they should get props for it. Now if only AMD would do the same...it's not all about making the best GPU anymore, it's about the software environment as well.
 
Last edited:

MrK6

Diamond Member
Aug 9, 2004
4,458
4
81
For F@H it's starting to work back towards CPUs - it's much more power- and money-efficient to get a Core i7 920, OC that to 3.6-3.8 and fold with that rather than buy a GPU. As for CUDA, I would make the argument that if you care that much to solely base the purchase of a $400-500 card on it, wouldn't you go the full monty and actually get a Tesla or a Quadro that has better drivers and better support?
The average user isn't going to spend $3000 on a graphics card to run CUDA, that's just unrealistic. NVIDIA's top end GPUs are very capable in CUDA and can be used to run all sorts of neat stuff. For instance, a friend of mine is a physics post-doc and uses CUDA to run his matrices - great non-video game usage for top-end hardware.
But then you also have people who run games on their old Athlon X2s, and who put 5850s and 5870s in with these systems, precisely because of what you said. I think games like BC2 are starting to tip the balance back and they're starting to make people realize that the CPU can make a huge difference to gameplay, but it won't happen overnight, and I'm skeptical as to whether GPU physics will ever fade away.
If you're going to argue everything from the lowest common denominator, how about just stop replying here and we can all go buy Dells in our favorite colors. BC2 is an example of exactly what will be happening, especially with the success it's had so far. There is no doubt that architecturally, GPU's are better designed for some types of physics calculations (fluids, etc.), but as I said, if the GPU can't even handle the rendering, doing physics calculations is just going to hurt it more. All I'm saying is GPU physics in its current manifestation isn't going to go far.

You're kidding, right? The Cypress cards are quieter than the Fermi ones, but they're still not 'extremely quiet'. In fact, I wouldn't use 'quiet' as a descriptor for any current reference cooler full stop.
I had a 5870 and currently own two 5850's in CF, as well as installing 5850's and 5870's on a few other systems. How many 58xx cards have you worked with that you think you know better? And since when does a 58xx card run at "full stop" in any game? The fans go up to 30-33%, at most.

I think the Batman AA thing is a bit cheap and underhand. But NVidia deserve respect for all the other software features that they have been working on. They poured a lot of money into PhysX, GPU ray tracing and things like that, and they should get props for it. Now if only AMD would do the same...it's not all about making the best GPU anymore, it's about the software environment as well.
I agree. I think AMD needs to step up it's software game in drivers, program development, game developer relations, etc. I think they're getting more aggressive, but indeed they are no where near NVIDIA in this realm (understandably so, as they are a "newer" company, stemming back to ATI being bought by AMD).
 

BababooeyHTJ

Senior member
Nov 25, 2009
283
0
0
The average user isn't going to spend $3000 on a graphics card to run CUDA, that's just unrealistic. NVIDIA's top end GPUs are very capable in CUDA and can be used to run all sorts of neat stuff. For instance, a friend of mine is a physics post-doc and uses CUDA to run his matrices - great non-video game usage for top-end hardware.
If you're going to argue everything from the lowest common denominator, how about just stop replying here and we can all go buy Dells in our favorite colors. BC2 is an example of exactly what will be happening, especially with the success it's had so far. There is no doubt that architecturally, GPU's are better designed for some types of physics calculations (fluids, etc.), but as I said, if the GPU can't even handle the rendering, doing physics calculations is just going to hurt it more. All I'm saying is GPU physics in its current manifestation isn't going to go far.

I completely agree with you there. On top of that I have never been impressed with any physics effects in a game with Physix.

I had a 5870 and currently own two 5850's in CF, as well as installing 5850's and 5870's on a few other systems. How many 58xx cards have you worked with that you think you know better? And since when does a 58xx card run at "full stop" in any game? The fans go up to 30-33%, at most.

Thanks, I was wondering about that. To me sound is a major factor. I couldn't put up with the 4870 reference cooler. I don't feel like messing with aftermarket cooling either. It took a lot of effort to get the results that I have now with my gtx280. If I had the money I would love to try out a 5870.
 

Madcatatlas

Golden Member
Feb 22, 2010
1,155
0
0
I agree. I think AMD needs to step up it's software game in drivers, program development, game developer relations, etc. I think they're getting more aggressive, but indeed they are no where near NVIDIA in this realm (understandably so, as they are a "newer" company, stemming back to ATI being bought by AMD).

This is the problem ATI/AMD face:

Does greater marketshare come first or does greater service/support?

AMD/ATI, much like nVIDIA, have a limited pool of resources (unlike, it seems, Intel). If AMD/ATi can ever manage to fix their finances so that they can for a time focus on service/support (by this i mean heavily investing itself in making their mark on 3rd party applications etc etc), they will surely obliterate nVidia.

Do we want nVidia obliterated? no. I would personally like a 50/50% marketshare between these two giants in the discreet market. And i would want them to openly share the current feature sets (eyefinity,phycx,cuda,stream) they sport as modular attachments to whatever new GPU they launch.
 

Qbah

Diamond Member
Oct 18, 2005
3,754
10
81
I'd go with the HD5870 here - it's faster. For me idle power draw is also very important so the HD5870 is an obvious choice.
 

BenSkywalker

Diamond Member
Oct 9, 1999
9,140
67
91
GPU physics in general doesn't have a future until architectures change dramatically.

Using the old GPU metrics, the 5830 should be faster then the 480, the 5870 should be over 60% faster(GTX 285 should also be faster). There has been a dramatic architectural shift, unfortunately too much loyalist rhetoric to have that conversation on these forums.
 

Mr. Pedantic

Diamond Member
Feb 14, 2010
5,027
0
76
The average user isn't going to spend $3000 on a graphics card to run CUDA, that's just unrealistic. NVIDIA's top end GPUs are very capable in CUDA and can be used to run all sorts of neat stuff. For instance, a friend of mine is a physics post-doc and uses CUDA to run his matrices - great non-video game usage for top-end hardware.
It's also unrealistic to buy a $500 GPU solely for the purposes of running Badaboom.

If you're going to argue everything from the lowest common denominator, how about just stop replying here and we can all go buy Dells in our favorite colors. BC2 is an example of exactly what will be happening, especially with the success it's had so far. There is no doubt that architecturally, GPU's are better designed for some types of physics calculations (fluids, etc.), but as I said, if the GPU can't even handle the rendering, doing physics calculations is just going to hurt it more. All I'm saying is GPU physics in its current manifestation isn't going to go far.
I'm just saying what I think. Stop being so aggressive and stop taking bits of my posts out of context.

I had a 5870 and currently own two 5850's in CF, as well as installing 5850's and 5870's on a few other systems. How many 58xx cards have you worked with that you think you know better? And since when does a 58xx card run at "full stop" in any game? The fans go up to 30-33%, at most.
Thanks to the wonders of internet video, I don't think it's all that necessary. Plus, acoustics are different for different people.
 

Madcatatlas

Golden Member
Feb 22, 2010
1,155
0
0
Using the old GPU metrics, the 5830 should be faster then the 480, the 5870 should be over 60% faster(GTX 285 should also be faster). There has been a dramatic architectural shift, unfortunately too much loyalist rhetoric to have that conversation on these forums.

Yup, i definetly would blame on that too if people didnt agree with me. Good one.
 

BenSkywalker

Diamond Member
Oct 9, 1999
9,140
67
91
Yup, i definetly would blame on that too if people didnt agree with me. Good one.

Agree with what? What am I even saying? You must be able to tell me, or you simply prove my point.
 

MrK6

Diamond Member
Aug 9, 2004
4,458
4
81
Thanks, I was wondering about that. To me sound is a major factor. I couldn't put up with the 4870 reference cooler. I don't feel like messing with aftermarket cooling either. It took a lot of effort to get the results that I have now with my gtx280. If I had the money I would love to try out a 5870.
Yeah, those my sentiments. Hopefully prices on the 5870's come down again, as they really are a treat to play with. Maybe I'm past that "phase" but I no longer want the fastest performance at any cost. I remember cranking my case fans to get my Athlon 64 3000+ to 2.5GHz (stock 2GHz), and well, those days are behind me I guess. It's the same reason now I'm running my 5850's at 900MHz and the default profile instead of 1000MHz with the fans at 50%.

This is the problem ATI/AMD face:

Does greater marketshare come first or does greater service/support?

AMD/ATI, much like nVIDIA, have a limited pool of resources (unlike, it seems, Intel). If AMD/ATi can ever manage to fix their finances so that they can for a time focus on service/support (by this i mean heavily investing itself in making their mark on 3rd party applications etc etc), they will surely obliterate nVidia.

Do we want nVidia obliterated? no. I would personally like a 50/50% marketshare between these two giants in the discreet market. And i would want them to openly share the current feature sets (eyefinity,phycx,cuda,stream) they sport as modular attachments to whatever new GPU they launch.
I agree. I think both companies have areas they need to work on. If AMD can get some more talent as well as resources allocated to software and development, they'll really be rounding out their presence in the market.

Using the old GPU metrics, the 5830 should be faster then the 480, the 5870 should be over 60% faster(GTX 285 should also be faster). There has been a dramatic architectural shift, unfortunately too much loyalist rhetoric to have that conversation on these forums.
What metrics are you using that makes you say it "should" be faster?

It's also unrealistic to buy a $500 GPU solely for the purposes of running Badaboom.
If you encode a lot of videos and it cuts your rendering time by 2/3 or more, it can definitely be worth the money. Going back to my friend, some of the math he's running is now 100x faster (I'm not exaggerating). Now hold that to the fact that if someone who's buying video cards has two similarly priced (+/- $50) cards, and one has this functionality and the other doesn't, it definitely tilts the decision.

I'm just saying what I think. Stop being so aggressive and stop taking bits of my posts out of context.
Sorry, didn't mean to come of as a dick, but it seems that you missed my point. I'm not negating the importance of CPU's in games, in fact, if you look at my posting history you'll see I've been arguing their importance for a long time. What I'm saying is that CPU's gaming performance is largely based on three factors - architecture, speed (GHz), and cores. We're only just starting to take advantage of the "cores" part and there still aren't many games out there (I can't think of one off the top of my head anyway) that keep a quad core pegged at 100% all the time. What I'm pointing out is that there's plenty of power there to do physics, and such power isn't readily available in graphics cards.

Thanks to the wonders of internet video, I don't think it's all that necessary. Plus, acoustics are different for different people.
That's fine, and what I'm saying is that you've been misled somewhere, somehow. The 5870's and 5850's are QUIET cards in normal usage. I would have no problem putting them in an HTPC. Now the GTX 480 could be much quieter than initial reviews made it out to be, but further testing ([H]'s follow up articles were great) and user reviews have not negated what early reviews have said so far.
 

Voo

Golden Member
Feb 27, 2009
1,684
0
76
If you encode a lot of videos and it cuts your rendering time by 2/3 or more, it can definitely be worth the money. Going back to my friend, some of the math he's running is now 100x faster (I'm not exaggerating). Now hold that to the fact that if someone who's buying video cards has two similarly priced (+/- $50) cards, and one has this functionality and the other doesn't, it definitely tilts the decision
You can't argue with that. The only thing you've got to keep in mind is that GPUs are still a SIMD architecture and therefore have some inherent disadvantages (well their divergent threads help a bit with that, but that just offsets the problems a bit) that limit their usefullness to a specific class of algorithms.

For those algorithms the speedup can be anywhere between 10 and 500 times, but these are all HPC things.. I have a hard time thinking about consumer space applications that would really profit from CUDA (except the obvious). So while GPGPU can be a great thing (and CUDA programming is fun too), it's not as if it will magically speed up most applications by a factor of 10. But if you're doing a lot of video rendering? There's nothing better than GPGPU out there.
 

Seero

Golden Member
Nov 4, 2009
1,456
0
0
So, it is clear that those who got, or wanna get an ATI card know that it won't accelerate PhysX, meaning that their CPU will either by super fast, or simply disable its effect. They knew PhysX is widely used, but hope that it either dies off or will one day work for non-nvidia video cards.

It is also very interesting that they believe that Nvidia 470 is a lot louder than 5870. To them, 470 is jet engine, and 5870 uses passive cooling. The fact is, both are loud, very loud, and 470 is a bit louder. 5870 is about 42-60 dba, and 470 is about 44-62 dba. If you think 5870 is silent and 470 is OMG loud, then it is probably your problem.

The real down side of Fermi is its power consumption. You need to ask yourself, how often is your video card on full load? and how often is it on idle? I am not doing math here, but I say 470 roughly uses 30 more watts if you are a heavy gamer. If you really think it is a OMG big difference, then it is probably your problem too.

If you think 3D gaming is a gimmick, then you will probably found yourself outdated. Avatar is not the first 3D movie, but it is the one that make 3D movie attractive. If 3D is 50 years old tech, then tessellation is about 400 years, as the word "tessellation" was found in 1660, and the tech was found in 1600.

I am not saying 470 is a must buy, but I seriously don't see how 5870 is that much better by any aspect. IMO, since Dx11 is not the norm and games really are still mostly Dx9, those are just like fans with lights. With those v-cards at 1600x1200, games will probably runs over 60 FPS. Now with a 60hz monitor, you won't even see more than 60FPS. If 3D and PhysX is not your cup of tea, than there is actually no way to use up those extra horsepower. So why do you getting 5870 if you don't have a super hi rez display or multi display? If it is in the name of DirectX 11, then you really should go for 470 as it is better at tessellation.

Some claimed that games bottlenecked by video card, while this is true for a few games that are designed to utilize high-end cards like 2033, it is false on most games. The fact is that Dx9 is single threaded and Dx10 is generally 2 threaded, I7 really is a waste. What really bottlenecks FPS is the FSB design, which the QPI kills it hands down. On top of that, I/O wait also plays a big role on MMOs, which SSD is the upgrade that you should aim for, not video card.

IMO if you really wanna go for ATI, 5850 is the best. If Nvidia is your cup of tea, go for 470. 5850 don't support PhysX, but cheaper and max out almost all games. If it doesn't, buy another one and go for xfire. If 3D is your cup of tea and you really want to burn some money, go for 3 120 projector and SLI some 480s. Otherwise, 470 will handle most games in 3D load on a single display. If you are using 48xx or 280/295, save your money for the next gen.
 

Jacky60

Golden Member
Jan 3, 2010
1,123
0
0
I run 2048 x 1152 (I usually compare performance at 1920x1200) as far as resolution.
CPU is a Core i5 750 - so CPU bottleneck shouldn't be too much of an issue.
Not interested in SLI - had a dual-gpu 7950 and was not impressed.
Also, not interested in overclocking.


The more I think about it the more I like just going with nvidia. In the end its only a couple degrees hotter than many cards I wouldn't be a fanboy if I didn't (though those poll results are pretty epic). :D

Feel free to convince me otherwise. I've probably changed my mind 4 times in the past 24 hours.

Well if you want to pay more -in UK 470 is same price as 5870 -for less performance then go ahead with Nvidia-they need people like you so they can keep making money regardless of performance. I'd go 5870 myself but I've bought from whichever manufacturer offered best performance i.e.Both Nvidia and ATI whichever best
 

Voo

Golden Member
Feb 27, 2009
1,684
0
76
It is also very interesting that they believe that Nvidia 470 is a lot louder than 5870. To them, 470 is jet engine, and 5870 uses passive cooling. The fact is, both are loud, very loud, and 470 is a bit louder. 5870 is about 42-60 dba, and 470 is about 44-62 dba. If you think 5870 is silent and 470 is OMG loud, then it is probably your problem.
You do know that dB has a logarithmic scale? Other than that enthusasts hopefully don't have a problem changing fan profiles, so if the 470 has 61.5dB @93° under load (Crysis, AT review) and the 5870 has 59.3dB @77° I think one easy solution should be obvious.
 

Lonyo

Lifer
Aug 10, 2002
21,938
6
81
So, it is clear that those who got, or wanna get an ATI card know that it won't accelerate PhysX, meaning that their CPU will either by super fast, or simply disable its effect. They knew PhysX is widely used, but hope that it either dies off or will one day work for non-nvidia video cards.
PhysX is indeed widely used.
PhysX which can run on the GPU? Not so much. In fact, it's barely used at all.

It is also very interesting that they believe that Nvidia 470 is a lot louder than 5870. To them, 470 is jet engine, and 5870 uses passive cooling. The fact is, both are loud, very loud, and 470 is a bit louder. 5870 is about 42-60 dba, and 470 is about 44-62 dba. If you think 5870 is silent and 470 is OMG loud, then it is probably your problem.
Most people seem to think "GTX470, well the GTX480 is loud so the GTX470 is loud too!" It's a family issue, people not differentiating between the cards enough, like people saying the GTX480 has better minimum frame rates than the HD5870 so the GTX470 must as well. They are two different cards which need to be considered as different cards because they have many different characteristics like performance, power and noise.
Of course it is louder, but not as loud as the GTX480 which people associate with it.

The real down side of Fermi is its power consumption. You need to ask yourself, how often is your video card on full load? and how often is it on idle? I am not doing math here, but I say 470 roughly uses 30 more watts if you are a heavy gamer. If you really think it is a OMG big difference, then it is probably your problem too.
Again, a family issue. The GTX480 uses a LOT of power, the GTX470 still uses a not inconsiderable amount more, but is nowhere near as bad as the GTX480, but people don't differentiate enough between the two cards.

If you think 3D gaming is a gimmick, then you will probably found yourself outdated. Avatar is not the first 3D movie, but it is the one that make 3D movie attractive. If 3D is 50 years old tech, then tessellation is about 400 years, as the word "tessellation" was found in 1660, and the tech was found in 1600.
Err, ok...
But it still is a gimmick. And it's also not an NV-only thing. In fact, that Avatar you reference has a game out on multiple platforms, consoles and PC. And it does indeed support 3D as well. And you know what? It supports 3D on ATI hardware, NV hardware, and on those consoles. So while I think 3D is indeed still a gimmick, even if you don't think 3D is a gimmick, it's not a vendor-specific thing (although NV has their own implementation
which currently seems to be superior to ATIs in terms of developer support, the freedom of hardware you need to buy with it etc), but it's not an ATI vs NV thing, since both can support it, and sometimes do (but it depends on your hardware... in both cases)

I am not saying 470 is a must buy, but I seriously don't see how 5870 is that much better by any aspect. IMO, since Dx11 is not the norm and games really are still mostly Dx9, those are just like fans with lights. With those v-cards at 1600x1200, games will probably runs over 60 FPS. Now with a 60hz monitor, you won't even see more than 60FPS. If 3D and PhysX is not your cup of tea, than there is actually no way to use up those extra horsepower. So why do you getting 5870 if you don't have a super hi rez display or multi display? If it is in the name of DirectX 11, then you really should go for 470 as it is better at tessellation.
You can have a way to use up the extra horsepower, such as Eyefinity, but that's even more expense and reduces the "it's faster and a little more expensive" because it becomes "it's faster and to have a great experience it's a lot more expensive", but on the other hand not all games can run with everything cranked on either card at some resolutions, so in many ways you are buying for the future.
Unfortunately superior raw performance now might not translate into superior performance in the future when things like tessellation do become the norm, so it becomes a life issue (how long will the life of the card be in your machine), and will tessellation be the thing which kills the HD5870 in future games, and let the GTX470 show itself as a superior future proof product or not, but unfortunately no-one has a crystal ball. In current DX11 games the HD5870 is (IIRC) typically faster. BFBC2, Stalker, Dirt 2 are ATI wins according to AT, while NV wins in Battleforge. So currently the superior tessellation performance doesn't mean wins in DX11 in general.
These cards are also not getting even 60fps at 1920x1200 with everything cranked in some of these games, so the extra horsepower won't be wasted at all, in fact it will be quite useful for maintaining playable frame-rates.
Not even going to go into the "you can't see over60gps" thing.

Some claimed that games bottlenecked by video card, while this is true for a few games that are designed to utilize high-end cards like 2033, it is false on most games. The fact is that Dx9 is single threaded and Dx10 is generally 2 threaded, I7 really is a waste. What really bottlenecks FPS is the FSB design, which the QPI kills it hands down. On top of that, I/O wait also plays a big role on MMOs, which SSD is the upgrade that you should aim for, not video card.
As you increase the resolution and details, you're getting more and more graphics bottlenecked, so you are fairly wrong there. When a faster card gets higher frame rates, you are video card bottlenecked. While there is an issue with most/many sites testing with a 4GHz Core i7 (which is not what everyone has), at high resolutions like the OP is using, a more powerful card is useful. Of course, it's all a balance game, but at high resolutions the balance shifts to the GPU.

IMO if you really wanna go for ATI, 5850 is the best. If Nvidia is your cup of tea, go for 470. 5850 don't support PhysX, but cheaper and max out almost all games. If it doesn't, buy another one and go for xfire. If 3D is your cup of tea and you really want to burn some money, go for 3 120 projector and SLI some 480s. Otherwise, 470 will handle most games in 3D load on a single display. If you are using 48xx or 280/295, save your money for the next gen.
The GTX, due to its pricing, is a damned good buy compared to the HD5870 (which is still selling for over $400 at most places it seems, needs to go back to that $380 mark), since it's only a bit slower and quite a margin cheaper, but then, that's only if you can actually buy one.

While you make many (IMO poor) points, your conclusion is mostly right, depending on the games being played.
If the OP was a hardcore BFBC2 player, the HD5870 would be the only choice. Equally if he wanted to play something where PhysX actually helps as a main game, then the GTX470 would be more worthwhile.
If the OP wants to buy right now, the HD5870 is the easier choice (due to higher availability, although that might change next week).
But generally the difference between the two cards is that the GTX470 loses, but not enough (in most situations) to be a difference between playable and unplayable so the extra $60 might not be worth it for the most part, but in some games where the HD5870 does seem to win big, if those are what's being played, then it might definitely be worth considering. Or if the aim is an eventual multiple display setup, same as a GTX470 might be the choice if 3D was a goal before the next upgrade was planned.

While the GTX470 does have some added features, they either aren't really used (PhysX) or they aren't actually NV features (3D) and require hardware the OP might not have and therefore wouldn't care for. I know I wouldn't care about 3D because I have a nice enough 24" monitor that I don't feel a need to replace and which would double my budget if I did want to upgrade graphics and get a 3D supporting monitor. But when the performance difference is so slim in many games, it's all a game specific thing, but if I had a budget of $x, and I could easily spend $x, why wouldn't I do that and maximise my performance? Especially when I'm likely to be GPU limited due to my screen resolution.
 

solofly

Banned
May 25, 2003
1,421
0
0
I would personally like a 50/50% marketshare between these two giants in the discreet market.

Not me, I would rather see 33/33/33 as in Intel joining the party...

PS.
I didn't vote but we already know where I stand...:)
 
Last edited:

evolucion8

Platinum Member
Jun 17, 2005
2,867
3
81
Chill dude. It's graphics cards, not organ transplants.

My new liver doesn't overclock that well compared to my previous liver, what can I do? :D

Just so nobody calls you out on it I want to clarify, The gap closes until you get someone running AA to walk the tightrope between the 5870 and 480 memory amounts. In other words, you can run out of mem on the 5870 and force swapping, which slows it dramatically, while the 480 is still chugging along. These are typically at unrealistic settings where neither card is giving playable frame rates, but % will swing dramatically to the 480. At playable settings the performance gap does indeed close between them.

That said this is about 5870 vs. 470 and that's a no brainer IMO, unless you need/want the proprietary nVidia features. The 5870 is just a better gaming card.

I agree with you.

IMO, since Dx11 is not the norm and games really are still mostly Dx9, those are just like fans with lights.

By your metric, would mean that there's no reason to upgrade to anything faster than a GTS 250 because most games are DX9.

With those v-cards at 1600x1200, games will probably runs over 60 FPS. Now with a 60hz monitor, you won't even see more than 60FPS. If 3D and PhysX is not your cup of tea, than there is actually no way to use up those extra horsepower. So why do you getting 5870 if you don't have a super hi rez display or multi display? If it is in the name of DirectX 11, then you really should go for 470 as it is better at tessellation.

There's way to use that horsepower, by telling the developers to stop doing console ports and adding more graphic details. PhysX may tax the GPU harder, but the eye candy gains are very hard to perceive (Except Batman AA). While the GF100 architecture is faster at pure tessellation, once you put everything in equation in real game situations like pixel fillrate, shader power, texture power, etc, that gap will close a lot, and probably there will be no gap at all. Do you remember the shading power of the ATi's X1K architecture compared to the G70 architecture? It was almost twice as powerful, and yet, never found a game until 4 years later that ran twice faster on the ATi's X1K architecture compared to nVidia's G70.

Some claimed that games bottlenecked by video card, while this is true for a few games that are designed to utilize high-end cards like 2033, it is false on most games. The fact is that Dx9 is single threaded and Dx10 is generally 2 threaded, I7 really is a waste. What really bottlenecks FPS is the FSB design, which the QPI kills it hands down. On top of that, I/O wait also plays a big role on MMOs, which SSD is the upgrade that you should aim for, not video card.

FSB design doesn't really kill the fps by much, that's why the i7 isn't much faster than the C2Q in gaming performance, heck even sometimes the C2Q architecture matches it and outperforms it thanks to its faster cache. I7 might be a waste for gaming, it is certainly useful if you game and do lots of background stuff.

If you are using 48xx or 280/295, save your money for the next gen.

I'm following that suggestion.
 

Seero

Golden Member
Nov 4, 2009
1,456
0
0
There's way to use that horsepower, by telling the developers to stop doing console ports and adding more graphic details. PhysX may tax the GPU harder, but the eye candy gains are very hard to perceive (Except Batman AA). While the GF100 architecture is faster at pure tessellation, once you put everything in equation in real game situations like pixel fillrate, shader power, texture power, etc, that gap will close a lot, and probably there will be no gap at all. Do you remember the shading power of the ATi's X1K architecture compared to the G70 architecture? It was almost twice as powerful, and yet, never found a game until 4 years later that ran twice faster on the ATi's X1K architecture compared to nVidia's G70.
See for yourself

FSB design doesn't really kill the fps by much, that's why the i7 isn't much faster than the C2Q in gaming performance, heck even sometimes the C2Q architecture matches it and outperforms it thanks to its faster cache. I7 might be a waste for gaming, it is certainly useful if you game and do lots of background stuff.
If you have a C2Q and I7 in the same room, playing the same game, then you will see the difference. The Frame per second is close, but the actual smoothness is far apart, especially on MMOs.
 

evolucion8

Platinum Member
Jun 17, 2005
2,867
3
81

That doesn't change much, that's another pure tessellation test. I stand by my point, when you are rendering more stuff than just pure tessellation which is even overdone in Unigine, the gap will be much closer than you think, specially when the resolution is increased. Look at Metro 2033 for example, the performance gap is much smaller compared to the Unigine results.

If you have a C2Q and I7 in the same room, playing the same game, then you will see the difference. The Frame per second is close, but the actual smoothness is far apart, especially on MMOs.

GPU's play far more importance in gaming performance than just CPU alone, MMO's doesn't push technology, they just suck in terms of graphics. I'm a FPS player, so I doubt that going from my heavily overclocked Quad to an i7 920 will bring me a lot of improvement in gaming performance, perhaps it might have an edge once I crank the clocks beyond 3.6GHz. With my current CPU, I was able to get almost 100% scaling going with Crossfire, so I'm far from being CPU limited.

Metro 2033 which is a brand new game showed the worst scaling, jumped from 26fps in the scene 2 with a single GPU to 40fps with Crossfire with everything maxed in a CPU limited native resolution of 1280x1024. Crysis and Warhead framerate almost doubled, the same happened with Bad Company 2 which stuttered a bit when 8xFSAA was used and never went beyond 45fps in average, now is in the realm of 90's in average, higher than 100% scaling (There's some scenarios which that happens, very sparingly though)
 
Last edited:

Nemesis 1

Lifer
Dec 30, 2006
11,366
2
0
Not me, I would rather see 33/33/33 as in Intel joining the party...

PS.
I didn't vote but we already know where I stand...:)


Same here I didn't vote either. Not much sense to it . I make my own decisions . I buy what I want. I really don't care who sells what. I think its pretty clear what this poll was all about . It backfired.
 

SlowSpyder

Lifer
Jan 12, 2005
17,305
1,002
126
I ended up going with a 470. Thanks for the feedback.

No offense meant, but I don't understand the point of this thread then... the poll shows overwhelmingly what this forum thinks. The 5870 is faster, and probably pretty close in price. But, whatever, enjoy your new card.