- Sep 5, 2003
- 19,458
- 765
- 126
Just wanted to shed some more light the subject of "Should I buy a new videocard or a CPU?"
I see many people trying to build gaming systems and continuing to believe the myth that you need to have a fast cpu for gaming. Of course, in theory this is the optimal setup, however many are unable to match the best components together since not everyone can afford A64 3800+ and 6800Ultra videocard.
However, if gaming is of a great importance for your basic computer needs, and you are bound to compromise, then which component should you compromise on. Or rather, do you really need a fast cpu?
Let's look more into the matter...by comparing a P4 3.2ghz and and a lower end P4 2.4 but matched with a much faster videocard.
Call of Duty - 1600x1200 4AA/8AF
9800xt + P4 3.2C = 60.7 frames
x800pro + P4 2.4C = 81.5
LOMAC - 1600x1200 4AA
9800xt + P4 3.2 = 22.5
x800Pro + P4 2.4 = 34.9
IL-2 Sturmovik - 1280x1024 4AA/8AF
9800xt + P4 3.2 = 29.8
X800Pro + P4 2.4 = 37.3
Splinter Cell - 1600x1200
9800xt + P4 3.2 = 44
X800Pro + P4 2.4 = 68.2
Tomb Raider:AOD - 1600x1200 4AA
9800xt + P4 3.2 = 24.1
X800Pro + P4 2.4 = 51.5
UT2004 - 1600x1200 4AA/8AF
9800xt + P4 3.2 = 31.1
X800Pro + P4 2.4 = 52.1
Far Cry - 1280x1024 4AA/8AF
9800xt + P4 3.2 = 30.6
X800Pro + P4 2.4 = 49
Halo - 1600x1200
9800xt + P4 3.2 = 28.1
X800Pro + P4 2.4 = 44.1
It appears that the videocard upgrade is a much better option for you if you are looking to increase your framerates for gaming as your primary objective.
In every single game, increasing the cpu speed from lets say your 2.6C to 3.2C speeds had a minute effect on gaming performance.
Considering that even on AXP 2500+ X800Pro often doubles the performance in Far Cry over a 9800pro (same cpu) solution as seen Here, the videocard upgrade is what you should seek if you are willing to part with $400US right now. Keep in mind that 6800GT is an overall faster card compared to x800pro (even in Far Cry and esp. Doom 3) and overclocks to 6800ultra speeds (while even if you overclock x800pro it will not reach the fillrate of the x800xt due to the 12 pipelines). Therefore, the above benchmarks, will only get better by substituting a 6800Gt in place of the X800Pro.
A lot of people bring about the concern that if they purchase a high end videocard like 6800Gt and pair it with AXP 2500+, their videocard will be cpu-limited. Sure to some extent this is true. However, others try to solve this problem by overclocking the cpu. But considering that even without a cpu overclock, a faster videocard delivers significantly better gaming experience, the cpu should be the least of your worries when it comes to gaming right now. If you can only afford to spend money on one or other as an upgrade, or if you plan on building a new system and can't figure out the compromise betweeen A64 2800+ and 6800GT vs. A64 3400+ and 6800nu, for gaming, go with a faster videocard.
Let's investigate further. It is reasonable to believe that in the future as games become more shader intensive, even after taking into considerations dependency on AI and physics calculations, the videocard will be subject to more and more stress % wise when it comes to providing the playability factor for gaming. Games that will be more demanding than doom will make things even worse. Let's see what kind of things we can learn from Doom 3:
Resolution Scaling - Videocard Dependency Here we see that 6800Ultra achieves around 75 frames @1280x1024 and 60 frames @1600x1200 and 9800Pro gets 48 frames @1024x768 and barely above 30@ 1280x1024. This is simply unacceptable and the CPU is 3.4 Extreme Edition.
Resolution Scaling - CPU Dependency Here we see that CPU speed does indeed contribute to significant performance boost. But wait a second, even P4 2.4C achieves above 60Frames per Second Average, making it sufficient as long as you have 6800Ultra. In fact even with a lowly XP2000+ and 6800Ultra you are getting 46.1 FPS @ 1280x1024 vs. 31 or so FPS with 9800Pro and P4 3.4 Extreme Edition.
I would like to come back to the constant buzz on the internet with the cpu being not fast enough for the newest videocards. Primarily, there have been a lot of misconception suggesting that you shouldn't invest into the more expensive videocards if you dont have P4 3.0 or equivalent. And furthermore, that today's cpus aren't fast enough for the best videocards.
Let's see
FX 3.0 ghz + X800 PE vs. FX 2.4ghz X800PE
I don't know about the majority of users who ENJOY gaming and intend on spending $300+US on a videocard and will want to play at 640x480 and 800x600 and without AA/AF. In other words, you dont buy a Ferrari Enzo to have it parked in your garage or your driveway unless you are the type of person who cares about your penis size. There is no point in buying high end videocards and not playing them at 1280x1024 with AA/AF enabled. From the review, as you can tell, going with a 3.0ghz FX makes almost no difference for gaming. If you are thinking there are 15-25% increases at low resolutions, I don't think it matters if you are getting 100-400 FPS already. Now last time I checked, FX-53 @ 2.4ghz 1mb cache on Socket 939 beat A64 3800+. So I am just going to assume that on average per clock speed, A64 FX = 1.5 P4. This would place FX3.0 at the very least close to 4.5ghz P4. Not only did going to a much faster CPU prove to be useless for videocard intensive tasks, the cpu also happened to be FX 3.0ghz paired with arguably the fastest videocard right now. This only helps to highlight that users with A64 2.0+ should have no problem taking advantage of future videocards following R420 and NV45, since it seems that today's X800xt and 6800U can be found struggling in some sitations already, and where cpu speed makes no difference towards a performance boost.
So to conclude:
1) Next time you have money to spend towards an upgrade and your primary need revolves around gaming, spend 100% of that money towards a videocard *unless you have P2 233mhz MMX.
2) Dont hesitate to get AXP 2500+ ($70) and 6800GT ($370) because your friend next door with A64 3400+ ($270) and 9800Pro is kicking himself.
3) Disregard my point of view if you never play with AA/AF enabled and gaming isn't your primary concern.
Other ways to save money in order to get a better videocard include buying higher latency ram as opposed to 2-2-2-5 for $150 per stick of 512mb; a motherboard that suits your needs and not XXXXPlatinum Edition that has 8 Usb ports and 8 SATA and RAid 1/0/etc (if you do need those features sure get it, but from my own experience (or lack thereof) when I bought top end motherboards, I found myself not needing half of the features they had. It is important to know that lower priced brands can be equally as good and you don't need to spend $150 or above for a good motherboard).
Now if you are thinking of upgrading your slow videocard, check this review out just to see how slow it has become at low resolutions:
101 Top-End Videocards (1999-2004)
So if you are still not convinced whether or not you want to purchase that new videocard you want so bad, please refer here:
Fastest Cards of Summer 2004
I hope this was helpful. Let me know what your thoughts on this are and feel free to post benchmarks that help to back up or contrast my opinion.
I see many people trying to build gaming systems and continuing to believe the myth that you need to have a fast cpu for gaming. Of course, in theory this is the optimal setup, however many are unable to match the best components together since not everyone can afford A64 3800+ and 6800Ultra videocard.
However, if gaming is of a great importance for your basic computer needs, and you are bound to compromise, then which component should you compromise on. Or rather, do you really need a fast cpu?
Let's look more into the matter...by comparing a P4 3.2ghz and and a lower end P4 2.4 but matched with a much faster videocard.
Call of Duty - 1600x1200 4AA/8AF
9800xt + P4 3.2C = 60.7 frames
x800pro + P4 2.4C = 81.5
LOMAC - 1600x1200 4AA
9800xt + P4 3.2 = 22.5
x800Pro + P4 2.4 = 34.9
IL-2 Sturmovik - 1280x1024 4AA/8AF
9800xt + P4 3.2 = 29.8
X800Pro + P4 2.4 = 37.3
Splinter Cell - 1600x1200
9800xt + P4 3.2 = 44
X800Pro + P4 2.4 = 68.2
Tomb Raider:AOD - 1600x1200 4AA
9800xt + P4 3.2 = 24.1
X800Pro + P4 2.4 = 51.5
UT2004 - 1600x1200 4AA/8AF
9800xt + P4 3.2 = 31.1
X800Pro + P4 2.4 = 52.1
Far Cry - 1280x1024 4AA/8AF
9800xt + P4 3.2 = 30.6
X800Pro + P4 2.4 = 49
Halo - 1600x1200
9800xt + P4 3.2 = 28.1
X800Pro + P4 2.4 = 44.1
It appears that the videocard upgrade is a much better option for you if you are looking to increase your framerates for gaming as your primary objective.
In every single game, increasing the cpu speed from lets say your 2.6C to 3.2C speeds had a minute effect on gaming performance.
Considering that even on AXP 2500+ X800Pro often doubles the performance in Far Cry over a 9800pro (same cpu) solution as seen Here, the videocard upgrade is what you should seek if you are willing to part with $400US right now. Keep in mind that 6800GT is an overall faster card compared to x800pro (even in Far Cry and esp. Doom 3) and overclocks to 6800ultra speeds (while even if you overclock x800pro it will not reach the fillrate of the x800xt due to the 12 pipelines). Therefore, the above benchmarks, will only get better by substituting a 6800Gt in place of the X800Pro.
A lot of people bring about the concern that if they purchase a high end videocard like 6800Gt and pair it with AXP 2500+, their videocard will be cpu-limited. Sure to some extent this is true. However, others try to solve this problem by overclocking the cpu. But considering that even without a cpu overclock, a faster videocard delivers significantly better gaming experience, the cpu should be the least of your worries when it comes to gaming right now. If you can only afford to spend money on one or other as an upgrade, or if you plan on building a new system and can't figure out the compromise betweeen A64 2800+ and 6800GT vs. A64 3400+ and 6800nu, for gaming, go with a faster videocard.
Let's investigate further. It is reasonable to believe that in the future as games become more shader intensive, even after taking into considerations dependency on AI and physics calculations, the videocard will be subject to more and more stress % wise when it comes to providing the playability factor for gaming. Games that will be more demanding than doom will make things even worse. Let's see what kind of things we can learn from Doom 3:
Resolution Scaling - Videocard Dependency Here we see that 6800Ultra achieves around 75 frames @1280x1024 and 60 frames @1600x1200 and 9800Pro gets 48 frames @1024x768 and barely above 30@ 1280x1024. This is simply unacceptable and the CPU is 3.4 Extreme Edition.
Resolution Scaling - CPU Dependency Here we see that CPU speed does indeed contribute to significant performance boost. But wait a second, even P4 2.4C achieves above 60Frames per Second Average, making it sufficient as long as you have 6800Ultra. In fact even with a lowly XP2000+ and 6800Ultra you are getting 46.1 FPS @ 1280x1024 vs. 31 or so FPS with 9800Pro and P4 3.4 Extreme Edition.
I would like to come back to the constant buzz on the internet with the cpu being not fast enough for the newest videocards. Primarily, there have been a lot of misconception suggesting that you shouldn't invest into the more expensive videocards if you dont have P4 3.0 or equivalent. And furthermore, that today's cpus aren't fast enough for the best videocards.
Let's see
FX 3.0 ghz + X800 PE vs. FX 2.4ghz X800PE
I don't know about the majority of users who ENJOY gaming and intend on spending $300+US on a videocard and will want to play at 640x480 and 800x600 and without AA/AF. In other words, you dont buy a Ferrari Enzo to have it parked in your garage or your driveway unless you are the type of person who cares about your penis size. There is no point in buying high end videocards and not playing them at 1280x1024 with AA/AF enabled. From the review, as you can tell, going with a 3.0ghz FX makes almost no difference for gaming. If you are thinking there are 15-25% increases at low resolutions, I don't think it matters if you are getting 100-400 FPS already. Now last time I checked, FX-53 @ 2.4ghz 1mb cache on Socket 939 beat A64 3800+. So I am just going to assume that on average per clock speed, A64 FX = 1.5 P4. This would place FX3.0 at the very least close to 4.5ghz P4. Not only did going to a much faster CPU prove to be useless for videocard intensive tasks, the cpu also happened to be FX 3.0ghz paired with arguably the fastest videocard right now. This only helps to highlight that users with A64 2.0+ should have no problem taking advantage of future videocards following R420 and NV45, since it seems that today's X800xt and 6800U can be found struggling in some sitations already, and where cpu speed makes no difference towards a performance boost.
So to conclude:
1) Next time you have money to spend towards an upgrade and your primary need revolves around gaming, spend 100% of that money towards a videocard *unless you have P2 233mhz MMX.
2) Dont hesitate to get AXP 2500+ ($70) and 6800GT ($370) because your friend next door with A64 3400+ ($270) and 9800Pro is kicking himself.
3) Disregard my point of view if you never play with AA/AF enabled and gaming isn't your primary concern.
Other ways to save money in order to get a better videocard include buying higher latency ram as opposed to 2-2-2-5 for $150 per stick of 512mb; a motherboard that suits your needs and not XXXXPlatinum Edition that has 8 Usb ports and 8 SATA and RAid 1/0/etc (if you do need those features sure get it, but from my own experience (or lack thereof) when I bought top end motherboards, I found myself not needing half of the features they had. It is important to know that lower priced brands can be equally as good and you don't need to spend $150 or above for a good motherboard).
Now if you are thinking of upgrading your slow videocard, check this review out just to see how slow it has become at low resolutions:
101 Top-End Videocards (1999-2004)
So if you are still not convinced whether or not you want to purchase that new videocard you want so bad, please refer here:
Fastest Cards of Summer 2004
I hope this was helpful. Let me know what your thoughts on this are and feel free to post benchmarks that help to back up or contrast my opinion.
Facebook
Twitter