Pits Benchmarking Index

Uncle Goy Goes Benchmarking

We always get a lot of questions regarding which part of their Computer people should upgrade first, CPU/MoBo or the video card.......

My take is that it doesn't matter how good your video card is if you don't have the CPU to back it up, and all the fancy screenshots in the world won't mean a thing to you if you still have to run in lesser resolutions and with lots of details switched off just to keep an acceptable frame rate. Personally I don't think "First Person Shooter" (FPS from now on) games like Quake3 make for very good benchmarks. A FSP game doesn't have a demanding physics engine running in the background eating valuable CPU cycles, nor do most bechmarks like 3DMark2000 (www.madonion.com) which I used in these tests because it's a widespread one and it kind of proves my point as you will see when you start looking at the numbers I got.

This is not meant as a "test to end all tests" kind of thing, there's way too much I don't know about video cards to even contemplate such a task. It's just meant as another tool to help you choose what to do when you have some extra cash to throw at your Computer.

One thing it shows though, is that just because one combination seems to be working very well in 3DMark, doesn't mean that it's necessarily faster in your favourite sim. Remember those CPU cycles that are "wasted" on other things besides Gfx.......this will be important when racing online as well, since some CPU cycles will be used for whatever is needed to keep you online. Therefore, the "ideal" video card would be one that does more on less CPU cycles, because we don't have access to all of them while racing our fav sims.

This article was actually started by coinsidence. A customer of ours asked if we could check out his ISDN card over the holidays, as he was going away on vacation and didn't need his PC. The PC in question was a P3-550 that I knew from earlier, so I decided to bring the thing home with me and do some testing, i.e. see how CPU speed vs FSB speed related since I also had access to a P3-700. The P3-700 ran stable at 980 on my old ASUS CUSL2 :-), and the 550 has been rock solid at 770, even at 825 but then only with PCI video cards as the AGP bus is running way too fast by then. It's running on an ASUS CUBX MoBo and the BX chip set doesn't like it when we run at 150 FSB speed.....

So here's the setup:

Common for all

Various video cards :-), release reference drivers only, no "special" tweaks
Quantum 7,5 GB UDMA33 Harddrive
Soundblaster Live! 1024 Value (Liveware 3.0)
Windows 98SE with DirectX 7.0, also installed VIA 4-1 drivers and such for the AMD tests
3DMark2000
Mercedes Benz Truck Racing (German demo), used the TNT2/Geforce settings
Nascar Heat (demo), everything maxed out in the options, ran high detail cars where possible
All tests run in 1024x768, 16bit colours

Cel300A
ASUS P3B-F
128 MB PC133 RAM, generic

P3-550E
ASUS CUBX
128 MB PC133 RAM, generic, which makes the 150 FSB speeds we ran at one point even more amazing...

P3-533B
ASUS P3B-F
128 MB PC133 RAM, generic

P3-700E
ASUS P3B-F
128 MB PC133 RAM, generic

AMD Duron 700
ASUS A7V (see a pattern yet ? :-)
128 MB PC133 CAS2, MemoryMan (Mosel Vitalic, same as Crucial)

AMD T-Bird 1100
ASUS A7V
128 MB PC133 CAS2, MemoryMan (Mosel Vitalic, same as Crucial)

The Video cards at hand at various times have been

3Dfx Voodoo Banshee 16 MB, Creative
3Dfx Voodoo 3-3000 16 MB
3Dfx Voodoo 5-5500 64 MB

ATI Rage Fury 128 32MB
ATI Rage Fury MAXX 64 MB
ATI Radeon 32 SDR

nVIDIA Riva TNT1 16 MB, Creative
nVIDIA Riva TNT2 M64 PRO 32 MB, CardExpert
nVIDIA GefForce 256 32 MB, ASUS 6600
nVIDIA GeForce DDR 32 MB, Creative
nVIDIA GeForce 2MX 32MB, CardExpert, "Golden Sample"
nVIDIA GeForce 2 PRO/400 64MB, CardExpert, "Golden Sample"

More cards will be added to the database when I can get my hands on them, and I'll try to run them on as many CPU combos as I can, but not all of these systems are available to me at all times, so here goes.........

CPUSpeed/FSBspeed Video card 3DMark MBTR NheatDemo
Cel300A RIVA TNT1 16MB 1010 15,8 20,9 LD
Cel300A@450/100 RIVA TNT1 16MB 1440 19,6 31,5 LD
Cel300A@472/105 RIVA TNT1 16MB 1490 20,0 32,2 LD
Cel300A RIVA TNT2 M64PRO 1020 18,1 20,8 LD
Cel300A@450/100 RIVA TNT2 M64PRO 1460 23,0 28,9 LD
Cel300A@472/105 RIVA TNT2 M64PRO 1515 23,8 29,5 LD
Cel300A GeForce2 MX 2390 19,1 20,6 LD
Cel300A@450/100 GeForce2 MX 3370 29,8 32,6 LD
Cel300A@472/105 GeForce2 MX 3650 31,6 34,2 LD
Cel300A GeForce256 2345 19,4 20,2 LD
Cel300A@450/100 GeForce256 3310 30,1 32,7 LD
Cel300A@472/105 GeForce256 3300 31,1 34,1 LD
Cel300A GeForceDDR 2370 19,5 20,4 LD
Cel300A@450/100 GeForceDDR 3465 30,3 32,8 LD
Cel300A@472/105 GeForceDDR 3510 31,1 34,3 LD
Cel300A GeForce2 PRO 400 2340 19,2 20,3 LD
Cel300A@450/100 GeForce2 PRO 400 3530 29,6 32,6 LD
Cel300A@472/105 GeForce2 PRO 400 3860 31,2 34,2 LD
Cel300A RageFury128 32MB 945 16,3 17,7 LD
Cel300A@450/100 RageFury128 32MB 1375 22,3 27,3 LD
Cel300A@472/105 RageFury128 32MB 1425 22,9 28,4 LD
Cel300A RageFury MAXX 64MB 1010 14,1 18,9 LD
Cel300A@450/100 RageFury MAXX 64MB 1535 17,1 30,1 LD
Cel300A@472/105 RageFury MAXX 64MB 1605 17,4 31,7 LD
Cel300A Radeon 32 MB DDR 2315 16,5 18,7 LD
Cel300A@450/100 Radeon 32 MB DDR 3320 25,5 29,8 LD
Cel300A@472/105 Radeon 32 MB DDR 3335 26,7 31,2 LD
Cel300A 3Dfx Banshee 16MB 790 Failed 19,1 LD
Cel300A@450/100 3Dfx Banshee 16MB 1135 Failed 30,2 LD
Cel300A@472/105 3Dfx Banshee 16MB 1170 Failed 31,1 LD
Cel300A 3Dfx V5 5500 64MB 1010 19,6 20,0 LD
Cel300A@450/100 3Dfx V5 5500 64MB 1525 29,7 31,7 LD
Cel300A@472/105 3Dfx V5 5500 64MB 1605 30,8 33,0 LD
------------------------
P3-550@366/66 3Dfx Banshee 16MB 1340 Failed 28,6 LD
P3-550 3Dfx Banshee 16MB 1415 Failed 37,3 LD
P3-550@733/133 3Dfx Banshee 16MB 1435 Failed 34,9 HD/38,6 LD
P3-550@770/140 3Dfx Banshee 16MB 1440 Failed 34,9 HD/39,0 LD
P3-550@366/66 3Dfx V5 5500 64MB 2250 25,3 24,9 HD/32,1 LD
P3-550 3Dfx V5 5500 64MB 3280 36,7 38,0 HD/47,0 LD
P3-550@733/133 3Dfx V5 5500 64MB 4170 42,1 48,8 HD/54,8 LD
P3-550@770/140 3Dfx V5 5500 64MB 4370 43,0 49,9 HD/56,0 LD
P3-550@366/66 RIVA TNT1 16MB 1800 18,1 30,1 LD
P3-550 RIVA TNT1 16MB 2015 20,6 36,1 LD
P3-550@733/133 RIVA TNT1 16MB 2080 21,8 36,3 LD
P3-550@770/140 RIVA TNT1 16MB Failed Failed Failed
P3-550@366/66 RIVA TNT2 M64PRO 1805 21,2 28,3 LD
P3-550 RIVA TNT2 M64PRO 2015 24,1 31,3 LD
P3-550@733/133 RIVA TNT2 M64PRO 2075 24,8 31,2 LD
P3-550@770/140 RIVA TNT2 M64PRO 2085 24,9 31,2 LD
P3-550@366/66 GeForce2 MX 3350 24,1 30,8 LD
P3-550 GeForce2 MX 4325 36,4 37,4 HD/47,2 LD
P3-550@733/133 GeForce2 MX 4945 44,4 48,4 HD/53,6 LD
P3-550@770/140 GeForce2 MX 5010 45,9 49,6 HD/55,9 LD
P3-550@366/66 RageFury128 32MB 1780 20,6 25,2 HD/31,6 LD
P3-550 RageFury128 32MB 1995 24,1 34,1 HD/39,1 LD
P3-550@733/133 RageFury128 32MB 2035 24,8 35,8 HD/39,4 LD
P3-550@770/140 RageFury128 32MB Failed Failed Failed
P3-550@366/66 RageFury MAXX 64MB 2250 15,8 24,9 HD/31,9 LD
P3-550 RageFury MAXX 64MB 3300 18,3/22,7 39,9 HD/47,9 LD
P3-550@733/133 RageFury MAXX 64MB 3935 19,5/24,1 49,9 HD/55,3 LD
P3-550@770/140 RageFury MAXX 64MB 4040 19,8/24,5 51,2 HD/56,4 LD
P3-550@366/66 GeForce256 3330 24,0 30,8 LD
P3-550 GeForce256 3970 36,7 37,8 HD/46,2 LD
P3-550@733/133 GeForce256 4530 44,4 46,6 HD/53,3 LD
P3-550@770/140 GeForce256 4580 45,2 48,3 HD/55,2 LD
P3-550@366/66 GeForceDDR 3360 23,8 31,1 LD
P3-550 GeForceDDR 4395 36,9 38,2 HD/45,9 LD
P3-550@733/133 GeForceDDR 5165 46,2 48,1 HD/53,8 LD
P3-550@770/140 GeForceDDR 5175 47,1 49,0 HD/55,2 LD
------------------------
P3-533B@400/100 RIVA TNT1 16MB 1805 19,2 25,3 HD/30,9 LD
P3-533B RIVA TNT1 16MB 1975 21,1 31,6 HD/35,5 LD
P3-533B@400/100 RIVA TNT2 M64PRO 1810 21,9 28,8 LD
P3-533B RIVA TNT2 M64PRO 1975 23,8 30,9 LD
P3-533B@400/100 GeForce2 MX 3265 27,2 33,3 LD
P3-533B GeForce2 MX 4025 35,6 34,8 HD/45,1 LD
P3-533B@400/100 GeForce256 3120 27,1 32,3 LD
P3-533B GeForce256 3720 36,1 35,0 HD/45,2 LD
P3-533B@400/100 GeForceDDR 3245 27,3 33,4 LD
P3-533B GeForceDDR 3980 36,3 34,8 HD/45,0 LD
P3-533B@400/100 GeForce2 PRO 400 3400 27,8 33,3 LD
P3-533B GeForce2 PRO 400 4505 36,2 35,0 HD/45,1 LD
P3-533B@400/100 RageFury128 32MB 1825 22,3 27,0 HD/33,8 LD
P3-533B RageFury128 32MB 24,1 33,5 HD/39,5 LD
P3-533B@400/100 RageFury MAXX 64MB 2275 16,1 (?) 27,0 HD/34,9 LD
P3-533B RageFury MAXX 64MB 2980 18,3 (?) 36,6 HD/46,0 LD
P3-533B@400/100 3Dfx Banshee 16MB 1350 Failed 30,1 LD
P3-533B 3Dfx Banshee 16MB 1420 Failed 36,2 LD
P3-533B@400/100 3Dfx Voodoo3 300016MB 2205 Failed 25,3 HD/33,2 LD
P3-533B 3Dfx Voodoo3 300016MB 2730 Failed 34,4 HD/43,9 LD
P3-533B@400/100 3Dfx V5 5500 64MB 2250 27,5 26,0 HD/34,3 LD
P3-533B 3Dfx V5 5500 64MB 2980 35,7 35,3 HD/45,7 LD
------------------------
AMD Duron 700 Radeon 32 MB SDR 4140 38,3 50,7/46,7 LD (no HD ?)
AMD Duron 700 Rage Fury MAXX 64 MB 3320 19,2 (?) 45,7 HD/ 53,6 LD
AMD Duron 700 RageFury128 32MB 2030 24,7 31,3 HD/ 34,4 LD
AMD Duron 700 3Dfx Banshee 16MB 1440 Failed* 35,1 HD/ 39,0 LD
AMD Duron 700 3Dfx Voodoo3 300016MB 2805 Failed* 40,5 HD/ 49,9 LD
AMD Duron 700 3Dfx V5 5500 64MB 3295 39,8 41,7 HD/ 51,6 LD
AMD Duron 700 RIVA TNT1 16MB 2035 21,2 33,9 HD/ 36,2 LD
AMD Duron 700 RIVA TNT2 M64PRO 2010 25,0 29,5 HD/ 30,9 LD
AMD Duron 700 GeForce2 MX 4735 42,9 44,1 HD/ 52,8 LD
AMD Duron 700 GeForce256 4344 43,5 44,2 HD/ 52,8 LD
AMD Duron 700 GeForceDDR 4915 44,1 44,2 HD/ 53,2 LD
------------------------
AMD T-Bird 1100 Radeon 32 MB SDR 4395 45,1 60,7LD (no HD ?)
AMD T-Bird 1100 Rage Fury MAXX 64 MB 4255 20,8 (?) 59,9 HD/ 61,4 LD
AMD T-Bird 1100 RageFury128 32MB 2080 24,8 31,1 HD/ 34,3 LD
AMD T-Bird 1100 3Dfx Banshee 16MB 1440 Failed* 35,1 HD/ 39,0 LD
AMD T-Bird 1100 3Dfx Voodoo3 300016MB 3165 Failed* 50,7 HD/ 56,6 LD
AMD T-Bird 1100 3Dfx V5 5500 64MB 4435 43,9 53,2 HD/ 59,8 LD
AMD T-Bird 1100 RIVA TNT1 16MB 2110 21,3 33,9 HD/ 36,2 LD
AMD T-Bird 1100 RIVA TNT2 M64PRO 2075 25,0 29,6 HD/ 30,9 LD
AMD T-Bird 1100 GeForce2 MX 5180 50,8 58,4 HD/ 60,9 LD
AMD T-Bird 1100 GeForce256 4660 52,1 58,5 HD/ 60,9 LD
AMD T-Bird 1100 GeForceDDR 5400 54,6 58,9 HD/ 61,3 LD
AMD T-Bird 1100 GeForce2 PRO 400 7760 52,1 59,2 HD/ 61,5 LD

*A few notes*

The 3Dfx Banshee and V3-3000 cards refused to run MBTR properly in the settings I had chosen, but rather than "compromising" with the settings, I chose to skip them altogether, because the whole point here was to run in 1024x768 and 16 bit colors. This resolution/color depth is pretty much what we'd expect from our systems these days.

The ATI Radeon card was a strange one, it scored low and refused to run NHeat with High detail cars. I only had access to that card for one night so I didn't have time to sort it out properly, but chose to include it anyway as it ran MBTR decently at least and NHeat with LD cars too at least gives us a hint to it's performance.

The big surprise of the bunch was the ATI Rage Fury MAXX, and although it didn't run well in MBTR, it's 3Dmark scores are pretty good and it ran NHeat very very good. The same can be said for the V3-3000, which scales very nicely with the available CPU power.

No need to e-mail me and tell me that your particular card will run faster on the same CPU with only a few minor tweaks, I already know that :-), like I said earlier, these are all "default" settings as provided by the latest "release" drivers as of 12/20/00. I'll try and cover some aspects of "tweaking" to get the most out of your system at another time, this is only meant as a reference database to show you what you should expect from your CPU/vid card combo or help you make an informed decision when you spend your hard earned cash .......and in my honest opinion, unless you have a really crappy old video card, upgrade you MoBo and CPU first, here's a few examples........

CPUSpeed/FSBspeed Video card 3DMark MBTR NheatDemo
P3-533B@400/100 3Dfx Voodoo3 300016MB 2205 Failed 25,3 HD/ 33,2 LD
P3-533B@400/100 GeForce2 PRO 400 3400 27,8 33,3 LD
AMD Duron 700 3Dfx Voodoo3 300016MB 2805 Failed* 40,5 HD/ 49,9 LD
P3-533B@400/100 RIVA TNT1 16MB 1805 19,2 25,3 HD/ 30,9 LD
AMD T-Bird 1100 RIVA TNT1 16MB 2110 21,3 33,9 HD/ 36,2 LD
AMD Duron 700 GeForce2 MX 4735 42,9 44,1 HD/ 52,8 LD

As you can see, you get better perfomance spending less money on an AMD Duron with MoBo than you'd (most likely) get from bying the latest all singing and dancing video card to go with an "older" CPU, with or without cup holder. On the other hand, spending money on a really hot CPU/Mobo and then have it crippled by an ancient video card doesn't make too much sense either, better to spend a little less on the CPU and get a decent vid card to go with it, unless of course you are getting a very good deal on a certain CPU/MoBo combo right now and plan on getting the vid card next... just don't get too disappointed if you don't get all you hoped for right away....:-)