The problem with this review is that none of these cards are common on pricewatch or eBay.
Where would the FX 5200 333 DDR place? How about FX5200 400 DDR, FX5500, FX5600XT, FX5600 (not XT not Ultra assuming that even exists), 5700LE, 5700 (not LE not Ultra).
In short the entire lineup from NVIDIA seems to have 30 shades of FXness and pricewatches categories and this review are both useless now that the names don't match any known specs.
I agree with what the article's author said in regard to the processor used. To compare video cards there should be one variable- the video card. This could be accomplished with a slower processor, but using the fastest setup ensures that nothing on the processor/motherboard/ram "end of things" hampers the test.
"9200 is a junk card, but unless I'm mistaken, it's really a modified Radeon 8500 core."
Radeon 9100 is a rebadged 8500 (LE model?). IIRC the 9000 and 9200 are based on 7500 architecture. BTW a friend of mine just got a barebones P4 2.8C and put a 9600SE in it before I could say anything. I cringed, but if he complains about sluggish games I'll tell him what's what.
"Of course, in looking at the 4200, you have to remember that numbers like we got in halo were obtained using the -use20 option which the 4200 can't do."
No kidding... and neither can the 9200!!! Obviously, the 9200 is a junk card, but unless I'm mistaken, it's really a modified Radeon 8500 core. So it's only DX8.1 capable, right? That's my point in saying the 4200 is a better choice for the budget sector.
And, come on, you have to admit that you hated benchmarking these low end cards. Waiting for the benchmarks to complete must have taken all day, especially on the 9200SE!
Incidentally, I don't think I've ever seen screenshots comparing "Halo -use20" to "Halo -use11". Are the differences really that great? I'm guessing not, but I could be wrong. (A link to such a comparison would be appreciated, incidentally.)
Well, I must say that I disagree with pure video card performance testing being done in any system other than one with maximum performance in all other areas.
Kris has already presented the core reason behind what we do, but maybe this explaination will help shed some light on the subject and add some perspective.
Take Halo for example. There is no current computer system in which you could drop a 9200 and get higher than 11.7fps at 10x7. Likewise, there is no system in which you could drop a 9600se and get higher than 18.7fps. Since these cards are at roughly the same pricepoint, the choice is obvious: the 9600se will be the better card to purchase.
To extend the example, these cards will maintain similar relative performance on lower end systems. Just how close the numbers come varries from game to game, and that is the point of doing CPU scaling test. CPU scaling warrants an entire article in and of itself.
If you just can't stand the wait for us to get to a CPU scaling test, feel free to take a look at the budget CPU roundup, compare the numbers to the FX51 system numbers for the 9800Pro and do a little rough extrapolation (this really isn't scientific and doesn't give you reliable numbers -- just an idea of what things might look like).
I appreciate the feedback, but it just doesn't make sence to test system performance when this type of article is about what the video cards can do. We've got other articles for that.
to change gears ...
As far as the 4200 goes, the test shows that it is definitely not an upgrade to buy a current generation budget card. We generally show previous generation cards in a kind of upgrade path light. If you want a budget card and you can find a 4200, the performance numbers tell the story pretty well.
Of course, in looking at the 4200, you have to remember that numbers like we got in halo were obtained using the -use20 option which the 4200 can't do. the ~18fps at 10x7 isn't bad for getting all the effects the developers wanted to include (the game is very playable on the higher end budget cards at 8x6 and high detail settings on the barton 2500+).
Hope this info helps shed light on why we do things the way we do. Thanks again for all the feedback.
I think I'm in the camp of the majority on this issue. The test setup was flawed. It is more than reasonalbe to assume that anyone who would buy any of these cards being tested would be running them on a system with much more limited capabilities than the fire breathing dragon they threw them on in this test..the results are worthless.
Y'all the thing to do is to purchase a Radeon 9500 128MB non pro and mod it to a 9700, and if you feel adventorous, OC the thing up to 9700 Pro speeds. www.allstarshop.com has the L shaped cards in stock (I just got mine), its fantastic. Also, a regular Geforce FX 5600 can be found for under $100 ($92 free shipping 128MB at www.amamax.com) and the 256MB for just over $100 ($100 w/ $4 shipping at house of computers), and a 5600XT for $109 from Newegg.
Ugh... #13, you're missing the point. Derek shouldn't do it all again. He shouldn't have done this 4th part to begin with! What a collossal waste of time! I mean, how long does it take to run a benchmark where you're only pulling 8 fps? WAAAAY too long.
The budget roundup should have skipped all AA/AF testing, since that's not a budget feature. Then it should have basically said, "If you're really looking at a budget graphics card, and you're running less than a 2 GHz PC, you should look at two options. First, if you have a DirectX 7 card or earlier, a GeForce 4 or Radeon 8500/9000 (non-SE!!!) would provide a decent performance increase. If you already have a DX8 or later card, you really need to upgrade your PC. None of the current budget cards are worth getting."
I, too would be interested to see the figures a 9600NP gets. Here in the UK you can get them for just a few pounds more than a 9600SE and they should be a much better option.
I was surprised to see how competitive the 4200 still is. I own one of these cards and was planning on upgrading to a 9600 Pro or XT but seeing as how it will only improve my FPS by 10-15(hardly a big improvement) in my favorite games I think I'll wait intil some real DirectX 9 games start coming out(sorry Tomb Raider).
I have to agree with all the critics. Sorry but this review sux. I'd like to see 9100/8500LE and 9600 and FX5200 thrown in the bunch, since FX5200 Ultra is practicaly non existent over here, and I bet in a lot of other places as well, and a 8500LE is a good reference point for many, and 9600 in much cheaper than 9600Pro and is therefor an interesting buy to consider ... if we had some performance numbers that is. I'd give derek an F for this. Do it again, and listen to what everybody is saying and provide a comparison table of all the cards, so everybody knows what they're looking at. And please don't use an Athlon FX system this time.
Since using AA/AF with budget cards in the vast majority of games just isn't doable with playable frame rates, I have to agree the ti4200 is still the king :) people who bought G4 ti4200 cards back in a day are only 2nd to 9700pro owners in bang for the buck over the long haul IMHO :)
I'm curious as to why the 9800 Pro was included in the results... it's hardly a budget card. I understand it gives an indication of how a budget card performs when compared to the best. But why isn't there any nVidia high end cards? If someone doesn't like ATI and wants to know how a 5700 Ultra compares to a 5900 non-ultra, wouldn't that be a useful tidbit to have here?
Personally, I find it pretty amazing that the 9600SE can beat the 5200Ultra in so many of the benchmarks seeing as it has less than a third of the memory bandwidth in comparison.
It leads me to think that it is a hell of a waste to use this chip on such a low-spec card. I suppose the market is calling for it though.
To everyone above who is griping about the test setup and/or the conclusion, I agree 100%. Who the heck does Derek think he is testing these cards with a PC which won't cause an artificial bottleneck?
Is he trying to provide accurate results or something?
I can't believe I just paid money to read this article... from now on, I'm going to buy every single piece of hardware that is out there and test them all myself on a test PC which I buy so I know EXACTLY which $50 video card to buy.
Or perhaps I'll quit b!tching and just read reviews on websites who offer content for basically free.
Also, the article shows quite clearly that these three budget cards are complete garbage and should all be avoided like the plague. That's useful.
9600SE: $99 or more 9200SE: $46.50 or more 5200 Ultra: $109 or more 5200: $63 or more (but it's even slower than the 5200 Ultra)
GeForce4 Ti4200: $75 or more
I got those prices from Newegg.com, except for the Ti4200 - it's being phased out, so you'd better act fast if you want one! So, if you really want a faster graphics card, and you're looking at a budget, buy a Ti4200 before it's too late.
And if you're thinking of saving $25 and getting the 9200SE, you're a crazy lunatic! At least pay the $10 to $25 extra and get a DX8 card with a 128 bit interface (the nicer 9000/9100/9200 cards).
It'd be good if the cards we are focusing on in the review are shown as a different colour in the tables (eg orange). At the moment it is a little confusing.
I understand that there is a desire to make the results comparable with the other parts of the article, but Sauron is absolutely right. So this is the maximum performance you'll get out of these budget cards? Big deal - you'll NEVER see these numbers, because you'll never actually run this configuration.
The conclusion is also screwed up, since the GeForce 4 Ti4200 kicks the crap out of all three budget cards in virtually every test. Current cost is about $75 I think, which would also make it a "budget" card. The only tests it didn't win had 4xAA and 8xAF enabled, which would also never be done with any of the lower end cards. Budget-minded shoppers don't give a damn about eye candy features. I should know - I'm one of them!
Anyway, personally, I think that Derek Wilson was given a pointless task in running benchmarks on the fastest current PC available paired with several low end cards. I feel sorry for him. If this budget information was going to be truly useful, we would need to see numbers on a 2.0 GHz/2500+ system (or less).
Of course, on a positive note, these "maximum performance" figures for the budget cards are probably easily attainable with anything over 1.4 GHz. Well, except for the really high scores in some of the less complex games. I know my Ti4200 never scores anywhere near what is listed in this review, and on games like Quake 3 and Unreal Tournament 2003, it probably would take at least a 2.4 GHz system to max it out.
The ultra high end setup is to eliminate bottlenecks. Its not really expected for you to go buy an FX51 and all; it just shows you the highest* thoroughput of the cards. We have proven in the past this scales very nicely.
When we do our graphics and CPU benchmarks, we try to use the exact same components with the exception of the video card/CPU/Memory in question. This is simple scientific method: we want one changing variable rather than several.
I question the validity of a budget video card shootout when the test platform is and AMD FX51 processor running on a Nforce3 platform and utilizing 1 gig of ram( a setup noone who is shopping for a budget video card has). I mean, how many people out there who play games will put together an Athlon FX system with 1 gig of top of the line RAM and then stick a radeon 9200 in it? Nobody.
If you're trying to help people on a budget decide which card to buy I believe you should test the cards on a "budget" or mid-range system to show how the cards perform on the type of systems which prospective purchasers of these cards are likely to own. I would at least show one or two comparative benchmarks on a system with a mid-range althon, similar to your CPU scaling tests in past reviews. In it's current form, this article is misleading at best, and is going to convince someone who has a 1 gig athlon to go out and buy a radeon 9200 and try to play C&C Generals or Halo on it which is not going to be a very fun experience at 5-12fps.
PS - I just want to say that I generally love the content of this site and consider it to be the top tech review site on the net. This article, however, really got under my skin because people who don't have alot of cash to spend are going to rely on it without noticing the fact that the cards were tested on an Athlon FX51.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
22 Comments
Back to Article
Mana211 - Wednesday, May 19, 2004 - link
The problem with this review is that none of these cards are common on pricewatch or eBay.Where would the FX 5200 333 DDR place? How about FX5200 400 DDR, FX5500, FX5600XT, FX5600 (not XT not Ultra assuming that even exists), 5700LE, 5700 (not LE not Ultra).
In short the entire lineup from NVIDIA seems to have 30 shades of FXness and pricewatches categories and this review are both useless now that the names don't match any known specs.
Zephyr106 - Saturday, December 20, 2003 - link
I agree with what the article's author said in regard to the processor used. To compare video cards there should be one variable- the video card. This could be accomplished with a slower processor, but using the fastest setup ensures that nothing on the processor/motherboard/ram "end of things" hampers the test."9200 is a junk card, but unless I'm mistaken, it's really a modified Radeon 8500 core."
Radeon 9100 is a rebadged 8500 (LE model?). IIRC the 9000 and 9200 are based on 7500 architecture. BTW a friend of mine just got a barebones P4 2.8C and put a 9600SE in it before I could say anything. I cringed, but if he complains about sluggish games I'll tell him what's what.
TrogdorJW - Friday, December 19, 2003 - link
"Of course, in looking at the 4200, you have to remember that numbers like we got in halo were obtained using the -use20 option which the 4200 can't do."No kidding... and neither can the 9200!!! Obviously, the 9200 is a junk card, but unless I'm mistaken, it's really a modified Radeon 8500 core. So it's only DX8.1 capable, right? That's my point in saying the 4200 is a better choice for the budget sector.
And, come on, you have to admit that you hated benchmarking these low end cards. Waiting for the benchmarks to complete must have taken all day, especially on the 9200SE!
Incidentally, I don't think I've ever seen screenshots comparing "Halo -use20" to "Halo -use11". Are the differences really that great? I'm guessing not, but I could be wrong. (A link to such a comparison would be appreciated, incidentally.)
DerekWilson - Thursday, December 18, 2003 - link
Well, I must say that I disagree with pure video card performance testing being done in any system other than one with maximum performance in all other areas.Kris has already presented the core reason behind what we do, but maybe this explaination will help shed some light on the subject and add some perspective.
Take Halo for example. There is no current computer system in which you could drop a 9200 and get higher than 11.7fps at 10x7. Likewise, there is no system in which you could drop a 9600se and get higher than 18.7fps. Since these cards are at roughly the same pricepoint, the choice is obvious: the 9600se will be the better card to purchase.
To extend the example, these cards will maintain similar relative performance on lower end systems. Just how close the numbers come varries from game to game, and that is the point of doing CPU scaling test. CPU scaling warrants an entire article in and of itself.
If you just can't stand the wait for us to get to a CPU scaling test, feel free to take a look at the budget CPU roundup, compare the numbers to the FX51 system numbers for the 9800Pro and do a little rough extrapolation (this really isn't scientific and doesn't give you reliable numbers -- just an idea of what things might look like).
I appreciate the feedback, but it just doesn't make sence to test system performance when this type of article is about what the video cards can do. We've got other articles for that.
to change gears ...
As far as the 4200 goes, the test shows that it is definitely not an upgrade to buy a current generation budget card. We generally show previous generation cards in a kind of upgrade path light. If you want a budget card and you can find a 4200, the performance numbers tell the story pretty well.
Of course, in looking at the 4200, you have to remember that numbers like we got in halo were obtained using the -use20 option which the 4200 can't do. the ~18fps at 10x7 isn't bad for getting all the effects the developers wanted to include (the game is very playable on the higher end budget cards at 8x6 and high detail settings on the barton 2500+).
Hope this info helps shed light on why we do things the way we do. Thanks again for all the feedback.
deathwalker - Wednesday, December 17, 2003 - link
I think I'm in the camp of the majority on this issue. The test setup was flawed. It is more than reasonalbe to assume that anyone who would buy any of these cards being tested would be running them on a system with much more limited capabilities than the fire breathing dragon they threw them on in this test..the results are worthless.Myrandex - Wednesday, December 17, 2003 - link
Y'all the thing to do is to purchase a Radeon 9500 128MB non pro and mod it to a 9700, and if you feel adventorous, OC the thing up to 9700 Pro speeds. www.allstarshop.com has the L shaped cards in stock (I just got mine), its fantastic. Also, a regular Geforce FX 5600 can be found for under $100 ($92 free shipping 128MB at www.amamax.com) and the 256MB for just over $100 ($100 w/ $4 shipping at house of computers), and a 5600XT for $109 from Newegg.TrogdorJW - Tuesday, December 16, 2003 - link
Ugh... #13, you're missing the point. Derek shouldn't do it all again. He shouldn't have done this 4th part to begin with! What a collossal waste of time! I mean, how long does it take to run a benchmark where you're only pulling 8 fps? WAAAAY too long.The budget roundup should have skipped all AA/AF testing, since that's not a budget feature. Then it should have basically said, "If you're really looking at a budget graphics card, and you're running less than a 2 GHz PC, you should look at two options. First, if you have a DirectX 7 card or earlier, a GeForce 4 or Radeon 8500/9000 (non-SE!!!) would provide a decent performance increase. If you already have a DX8 or later card, you really need to upgrade your PC. None of the current budget cards are worth getting."
Maringer - Tuesday, December 16, 2003 - link
I, too would be interested to see the figures a 9600NP gets. Here in the UK you can get them for just a few pounds more than a 9600SE and they should be a much better option.Medellon - Tuesday, December 16, 2003 - link
I was surprised to see how competitive the 4200 still is. I own one of these cards and was planning on upgrading to a 9600 Pro or XT but seeing as how it will only improve my FPS by 10-15(hardly a big improvement) in my favorite games I think I'll wait intil some real DirectX 9 games start coming out(sorry Tomb Raider).AtaStrumf - Tuesday, December 16, 2003 - link
I have to agree with all the critics. Sorry but this review sux. I'd like to see 9100/8500LE and 9600 and FX5200 thrown in the bunch, since FX5200 Ultra is practicaly non existent over here, and I bet in a lot of other places as well, and a 8500LE is a good reference point for many, and 9600 in much cheaper than 9600Pro and is therefor an interesting buy to consider ... if we had some performance numbers that is. I'd give derek an F for this. Do it again, and listen to what everybody is saying and provide a comparison table of all the cards, so everybody knows what they're looking at. And please don't use an Athlon FX system this time.cristic - Tuesday, December 16, 2003 - link
for #10: yes it would be, also like a 9800non-pro against a 9600XT and 9700 and 9700proIf I should buy a low-end card right now I would go with a 4200... heck I own one, since god know when! (:
DAPUNISHER - Tuesday, December 16, 2003 - link
Since using AA/AF with budget cards in the vast majority of games just isn't doable with playable frame rates, I have to agree the ti4200 is still the king :) people who bought G4 ti4200 cards back in a day are only 2nd to 9700pro owners in bang for the buck over the long haul IMHO :)Jeff7181 - Tuesday, December 16, 2003 - link
I'm curious as to why the 9800 Pro was included in the results... it's hardly a budget card. I understand it gives an indication of how a budget card performs when compared to the best. But why isn't there any nVidia high end cards? If someone doesn't like ATI and wants to know how a 5700 Ultra compares to a 5900 non-ultra, wouldn't that be a useful tidbit to have here?nastyemu25 - Tuesday, December 16, 2003 - link
lol@#7i concur.
Maringer - Tuesday, December 16, 2003 - link
Personally, I find it pretty amazing that the 9600SE can beat the 5200Ultra in so many of the benchmarks seeing as it has less than a third of the memory bandwidth in comparison.It leads me to think that it is a hell of a waste to use this chip on such a low-spec card. I suppose the market is calling for it though.
merlocka - Tuesday, December 16, 2003 - link
To everyone above who is griping about the test setup and/or the conclusion, I agree 100%. Who the heck does Derek think he is testing these cards with a PC which won't cause an artificial bottleneck?Is he trying to provide accurate results or something?
I can't believe I just paid money to read this article... from now on, I'm going to buy every single piece of hardware that is out there and test them all myself on a test PC which I buy so I know EXACTLY which $50 video card to buy.
Or perhaps I'll quit b!tching and just read reviews on websites who offer content for basically free.
TrogdorJW - Tuesday, December 16, 2003 - link
Also, the article shows quite clearly that these three budget cards are complete garbage and should all be avoided like the plague. That's useful.9600SE: $99 or more
9200SE: $46.50 or more
5200 Ultra: $109 or more
5200: $63 or more (but it's even slower than the 5200 Ultra)
GeForce4 Ti4200: $75 or more
I got those prices from Newegg.com, except for the Ti4200 - it's being phased out, so you'd better act fast if you want one! So, if you really want a faster graphics card, and you're looking at a budget, buy a Ti4200 before it's too late.
And if you're thinking of saving $25 and getting the 9200SE, you're a crazy lunatic! At least pay the $10 to $25 extra and get a DX8 card with a 128 bit interface (the nicer 9000/9100/9200 cards).
Emma - Tuesday, December 16, 2003 - link
It'd be good if the cards we are focusing on in the review are shown as a different colour in the tables (eg orange). At the moment it is a little confusing.Thanks.
TrogdorJW - Tuesday, December 16, 2003 - link
I understand that there is a desire to make the results comparable with the other parts of the article, but Sauron is absolutely right. So this is the maximum performance you'll get out of these budget cards? Big deal - you'll NEVER see these numbers, because you'll never actually run this configuration.The conclusion is also screwed up, since the GeForce 4 Ti4200 kicks the crap out of all three budget cards in virtually every test. Current cost is about $75 I think, which would also make it a "budget" card. The only tests it didn't win had 4xAA and 8xAF enabled, which would also never be done with any of the lower end cards. Budget-minded shoppers don't give a damn about eye candy features. I should know - I'm one of them!
Anyway, personally, I think that Derek Wilson was given a pointless task in running benchmarks on the fastest current PC available paired with several low end cards. I feel sorry for him. If this budget information was going to be truly useful, we would need to see numbers on a 2.0 GHz/2500+ system (or less).
Of course, on a positive note, these "maximum performance" figures for the budget cards are probably easily attainable with anything over 1.4 GHz. Well, except for the really high scores in some of the less complex games. I know my Ti4200 never scores anywhere near what is listed in this review, and on games like Quake 3 and Unreal Tournament 2003, it probably would take at least a 2.4 GHz system to max it out.
KristopherKubicki - Monday, December 15, 2003 - link
Sauron,The ultra high end setup is to eliminate bottlenecks. Its not really expected for you to go buy an FX51 and all; it just shows you the highest* thoroughput of the cards. We have proven in the past this scales very nicely.
When we do our graphics and CPU benchmarks, we try to use the exact same components with the exception of the video card/CPU/Memory in question. This is simple scientific method: we want one changing variable rather than several.
Hope that helps!
Kristopher
KristopherKubicki - Monday, December 15, 2003 - link
http://www.anandtech.com/video/showdoc.html?i=1933...Radeon 9700 Pro outperforming the 9800 XT :)
Nice.
Kristopher
sauron - Monday, December 15, 2003 - link
I question the validity of a budget video card shootout when the test platform is and AMD FX51 processor running on a Nforce3 platform and utilizing 1 gig of ram( a setup noone who is shopping for a budget video card has). I mean, how many people out there who play games will put together an Athlon FX system with 1 gig of top of the line RAM and then stick a radeon 9200 in it? Nobody.If you're trying to help people on a budget decide which card to buy I believe you should test the cards on a "budget" or mid-range system to show how the cards perform on the type of systems which prospective purchasers of these cards are likely to own. I would at least show one or two comparative benchmarks on a system with a mid-range althon, similar to your CPU scaling tests in past reviews. In it's current form, this article is misleading at best, and is going to convince someone who has a 1 gig athlon to go out and buy a radeon 9200 and try to play C&C Generals or Halo on it which is not going to be a very fun experience at 5-12fps.
PS - I just want to say that I generally love the content of this site and consider it to be the top tech review site on the net. This article, however, really got under my skin because people who don't have alot of cash to spend are going to rely on it without noticing the fact that the cards were tested on an Athlon FX51.