- Jun 4, 2004
- 17,573
- 15,693
- 146
How many times have we heard the following:
"I don't want to upgrade to PCIE yet so I want to know will my processor bottleneck an AGP X1950PRO"
To help out the larger number of ATVF posters who are still interested in an AGP upgrade I've taken a look at whether a decent P4 Prescott system (my main rig
) will bottleneck my new X1950PRO 512.
My system consists of the following:
CPU: P4 3.2E Prescott (3200Mhz 16X multi 800mhz FSB effective)
MOBO: Asus P4P800 E Deluxe (Intel 865PE chipset with PAT)
RAM: 2Gigs dual channel - 2X512MB Corsair XMS DDR 400 2-2-2-5 & 2X512MB Crucial Ballistix DDR 400 2-2-2-5
VC: Diamond Viper X1950PRO 512MB (Core 580 Mem 1400 - effectively built by Sapphire)
PSU: Enermax Liberty 500W - Dual 12V rails 22A max per rail - 32A max combined
CPU Cooler: Zalman 9500 CNPS LED
HD: Boot/App Drive WD 160Gig 7200RPM 8mb cache
LCD: Planar 19in 1280x1024 native 25ms pixel response
Since I don't know how to really benchmark a video game I stuck with using built in benchmarks to check for bottlenecks. My testing consists of the following apps:
1) 3Dmark03
2) 3Dmark05
3) 3Dmark06
4) HL2 Lost Coast Stress Test
5) F.E.A.R. in game test
Each one of these benchmarks was run at 3 different CPU speeds to tease out the point where the CPU becomes the bottleneck:
1) 2800Mhz - FSB 175mhz Ram DDR 350 2-2-2-5 (I expected to run the ram higher with a divider but my higher dividers disappear when I'm below 200Mhz So this speed will be a bit slower than a stock Prescott at 2.8 and definatley slower than a Northy at 2.8)
2)3200Mhz (stock) - FSB 200 Ram DDR 400 2-2-2-5
3)3600Mhz - FSB 225 Ram DDR 450 2.5-3-3-6 (I have not optimized this OC so more performance maybe possible)
All OCs were at stock voltage. It appears I have some headroom left as I almost manged to boot into windows at 3.8 with stock voltage
For F.E.A.R and Lost Coast I tested at the following settings:
1)1024x768 0XAA Trillinear
2)1024x768 4XAA 16X HQAF
3)1280x1024 0XAA Trillinear (960 for FEAR)
4)1280x1024 4XAA 16A HQAF (960 for FEAR)
3Dmark was tested at the default settings
Before we get into the data what we are looking for is following:
1)Large changes in FPS or Score with CPU speed changes at the same settings are indications that the GPU is waiting on the CPU and hence a CPU bottleneck
2)No change in FPS or Score when increasing the resolution and settings at the same CPU speed is also an indication of the GPU not being fully utilized and hence a CPU bottleneck
3)Small changes in FPS or Score when CPU speed changes at the same settings are indications the GPU is fully utilized and is a GPU bottleneck
4) Large drops in FPS or Score when increasing resolution and settings at the same CPU speed is an indication of a GPU bottleneck
Catalyst v7.2 was used for all testing.
Vsync was not enabled
HQAF was checked in CCC
Lost Coast - all settings at Max
FEAR - all settings at Max except no Soft Shadows
So without further ado the data:
3D Mark 03________________________________________________
CPU Speed             Score
2.8:                             14302
3.2:                             14715
3.6:                             15113
Comments: There's not much to talk about here. Obviously a P4/X1950PRO rocks 3Dm03. We see about 400PTS for each 400MHZ increase. We're basically a bit CPU limited but for a benchmark this old and at 10x7 it was to be expected.
3D Mark 05________________________________________________
CPU Speed             Score
2.8:                             6458
3.2:                             7227
3.6:                             7781
Comments: 3Dmark 05 is a bit more interesting. We see almost an 800 pt jump for a 400mhz increase from 2.8-3.2. From 3.2-3.6 we see another 550 pt jump. It looks like we are completely CPU bound at this point. Maybe at 4.0Ghz we would see scores similar to an AMD or low end Core2. While I didn't record it I did run at 3.6 with a core OC of 30mhz on the X1950 and got a whopping 4 extra pts.
3D Mark 06________________________________________________
CPU Speed                 Score       Sm2.0       Sm3.0       CPU
2.8:                                 3963         1753         2058         817
3.2:                                 4157         1801         2065         928
3.6:                                 4328         1845         2081         1038
Comments: Now with 3Dmark 06 we see a roughly 200pt jump for each 400 mhz increase in the CPU speed. This is mostly due to 06 taking your CPU into account in the total score. And in fact looking at the CPU score we see the largest jump. The 3.6 score is low but fairly close to more modern systems.
HL2 Lost Coast_____________________________________________
CPU                 10x7                 10x7                       12x10 [/u]                     12x10
Speed             0XAA-Tri       4XAA-16XHQAF       0XAA-Tri       4XAA-16XHQAF
2.8:                   69.75             67.55                             69.23                     67.32
3.2:                   79.28             78.01                             80.11                     75.76
3.6:                   84.37             82.92                             85.85                     78.76
Comments: Ok now on to a benchmark that actually matters. It's obvious that this game engine is completely CPU bound. Each time we up the OC on the CPU we get a good boost in frame rate. Although the rate of increase is dropping as the CPU speed increases suggesting we are starting to load the GPU. This is reinforced by the fact that at 12X10 with AA an AF there is a measurable drop in FPS and the 3.2 and 3.6 FPS at these settings are closer than the lower rez FPS at the same CPU speeds.
The one thing to point out to the upgrade now crowd is even at 2.8Ghz at 12x10 with 4XAA/16XAF I averaged over 67 FPS if thats not playable I don't know what is. While it IS a CPU bottleneck I have to say who cares at those FPS.
F.E.A.R___________________________________________________
CPU                 10x7                     10x7                   12x9                   12x9
Speed             0XAA-Tri       4XAA-16XHQAF       0XAA-Tri       4XAA-16XHQAF
2.8:                 22/66/200         20/53/162             21/57/155           18/45/129
BrkDwn:   5%/33%/62%   11%/34%/55%       7%/32%/61%   12%/42%46%
3.2:                 28/77/210         25/59/167             25/64/168           22/49/134
BrkDwn:   0%/20%/80%   0%/38%/62%       0%/31%/69%   5%/42%/53%
3.6:                 30/77/203         25/60/173             26/66/180           22/50/135
BrkDwn:   0%/15%/85%   0%/36%/64%      0%/26%/74%   1%/48%/51%
Comments: FEAR is the opposite example from Source. We are mostly GPU bound here. The scores are very close across the settings at both 3.6 and 3.2. At 2.8 and at lower rez we are a little CPU bound but once you up the rez to 12x9 with AF and AA it's basically the GPU holding you back. FEAR could actually use a faster GPU
________________________________________________________________
So is the X1950PRO worth it? For me - Hell Yes!
If you have an older rig that still has decent specs (P4 800mhz FSB/HT, AXP 2.0Ghz and up, or any A64 with at least 1Gig of ram) a good PSU or are willing to get one and you fit one of the following then do yourself a favor and consider an X1950PRO:
1) Don't have the time for a complete rebuild of your rig just to game.
2) Don't have the money to buy a new mobo and/or CPU and/or RAM AND a new GPU
3) Don't want to upgrade to a cheap mobo just for PCIE.
4) Are generally satisfied with you rig except for games.
So there it is. I hope you guys find this useful! It took forever to format!
"I don't want to upgrade to PCIE yet so I want to know will my processor bottleneck an AGP X1950PRO"
To help out the larger number of ATVF posters who are still interested in an AGP upgrade I've taken a look at whether a decent P4 Prescott system (my main rig
) will bottleneck my new X1950PRO 512.My system consists of the following:
CPU: P4 3.2E Prescott (3200Mhz 16X multi 800mhz FSB effective)
MOBO: Asus P4P800 E Deluxe (Intel 865PE chipset with PAT)
RAM: 2Gigs dual channel - 2X512MB Corsair XMS DDR 400 2-2-2-5 & 2X512MB Crucial Ballistix DDR 400 2-2-2-5
VC: Diamond Viper X1950PRO 512MB (Core 580 Mem 1400 - effectively built by Sapphire)
PSU: Enermax Liberty 500W - Dual 12V rails 22A max per rail - 32A max combined
CPU Cooler: Zalman 9500 CNPS LED
HD: Boot/App Drive WD 160Gig 7200RPM 8mb cache
LCD: Planar 19in 1280x1024 native 25ms pixel response
Since I don't know how to really benchmark a video game I stuck with using built in benchmarks to check for bottlenecks. My testing consists of the following apps:
1) 3Dmark03
2) 3Dmark05
3) 3Dmark06
4) HL2 Lost Coast Stress Test
5) F.E.A.R. in game test
Each one of these benchmarks was run at 3 different CPU speeds to tease out the point where the CPU becomes the bottleneck:
1) 2800Mhz - FSB 175mhz Ram DDR 350 2-2-2-5 (I expected to run the ram higher with a divider but my higher dividers disappear when I'm below 200Mhz
So this speed will be a bit slower than a stock Prescott at 2.8 and definatley slower than a Northy at 2.8)2)3200Mhz (stock) - FSB 200 Ram DDR 400 2-2-2-5
3)3600Mhz - FSB 225 Ram DDR 450 2.5-3-3-6 (I have not optimized this OC so more performance maybe possible)
All OCs were at stock voltage. It appears I have some headroom left as I almost manged to boot into windows at 3.8 with stock voltage
For F.E.A.R and Lost Coast I tested at the following settings:
1)1024x768 0XAA Trillinear
2)1024x768 4XAA 16X HQAF
3)1280x1024 0XAA Trillinear (960 for FEAR)
4)1280x1024 4XAA 16A HQAF (960 for FEAR)
3Dmark was tested at the default settings
Before we get into the data what we are looking for is following:
1)Large changes in FPS or Score with CPU speed changes at the same settings are indications that the GPU is waiting on the CPU and hence a CPU bottleneck
2)No change in FPS or Score when increasing the resolution and settings at the same CPU speed is also an indication of the GPU not being fully utilized and hence a CPU bottleneck
3)Small changes in FPS or Score when CPU speed changes at the same settings are indications the GPU is fully utilized and is a GPU bottleneck
4) Large drops in FPS or Score when increasing resolution and settings at the same CPU speed is an indication of a GPU bottleneck
Catalyst v7.2 was used for all testing.
Vsync was not enabled
HQAF was checked in CCC
Lost Coast - all settings at Max
FEAR - all settings at Max except no Soft Shadows
So without further ado the data:
3D Mark 03________________________________________________
CPU Speed             Score
2.8:                             14302
3.2:                             14715
3.6:                             15113
Comments: There's not much to talk about here. Obviously a P4/X1950PRO rocks 3Dm03. We see about 400PTS for each 400MHZ increase. We're basically a bit CPU limited but for a benchmark this old and at 10x7 it was to be expected.
3D Mark 05________________________________________________
CPU Speed             Score
2.8:                             6458
3.2:                             7227
3.6:                             7781
Comments: 3Dmark 05 is a bit more interesting. We see almost an 800 pt jump for a 400mhz increase from 2.8-3.2. From 3.2-3.6 we see another 550 pt jump. It looks like we are completely CPU bound at this point. Maybe at 4.0Ghz we would see scores similar to an AMD or low end Core2. While I didn't record it I did run at 3.6 with a core OC of 30mhz on the X1950 and got a whopping 4 extra pts.
3D Mark 06________________________________________________
CPU Speed                 Score       Sm2.0       Sm3.0       CPU
2.8:                                 3963         1753         2058         817
3.2:                                 4157         1801         2065         928
3.6:                                 4328         1845         2081         1038
Comments: Now with 3Dmark 06 we see a roughly 200pt jump for each 400 mhz increase in the CPU speed. This is mostly due to 06 taking your CPU into account in the total score. And in fact looking at the CPU score we see the largest jump. The 3.6 score is low but fairly close to more modern systems.
HL2 Lost Coast_____________________________________________
CPU                 10x7                 10x7                       12x10 [/u]                     12x10
Speed             0XAA-Tri       4XAA-16XHQAF       0XAA-Tri       4XAA-16XHQAF
2.8:                   69.75             67.55                             69.23                     67.32
3.2:                   79.28             78.01                             80.11                     75.76
3.6:                   84.37             82.92                             85.85                     78.76
Comments: Ok now on to a benchmark that actually matters. It's obvious that this game engine is completely CPU bound. Each time we up the OC on the CPU we get a good boost in frame rate. Although the rate of increase is dropping as the CPU speed increases suggesting we are starting to load the GPU. This is reinforced by the fact that at 12X10 with AA an AF there is a measurable drop in FPS and the 3.2 and 3.6 FPS at these settings are closer than the lower rez FPS at the same CPU speeds.
The one thing to point out to the upgrade now crowd is even at 2.8Ghz at 12x10 with 4XAA/16XAF I averaged over 67 FPS if thats not playable I don't know what is. While it IS a CPU bottleneck I have to say who cares at those FPS.
F.E.A.R___________________________________________________
CPU                 10x7                     10x7                   12x9                   12x9
Speed             0XAA-Tri       4XAA-16XHQAF       0XAA-Tri       4XAA-16XHQAF
2.8:                 22/66/200         20/53/162             21/57/155           18/45/129
BrkDwn:   5%/33%/62%   11%/34%/55%       7%/32%/61%   12%/42%46%
3.2:                 28/77/210         25/59/167             25/64/168           22/49/134
BrkDwn:   0%/20%/80%   0%/38%/62%       0%/31%/69%   5%/42%/53%
3.6:                 30/77/203         25/60/173             26/66/180           22/50/135
BrkDwn:   0%/15%/85%   0%/36%/64%      0%/26%/74%   1%/48%/51%
Comments: FEAR is the opposite example from Source. We are mostly GPU bound here. The scores are very close across the settings at both 3.6 and 3.2. At 2.8 and at lower rez we are a little CPU bound but once you up the rez to 12x9 with AF and AA it's basically the GPU holding you back. FEAR could actually use a faster GPU
________________________________________________________________
So is the X1950PRO worth it? For me - Hell Yes!
If you have an older rig that still has decent specs (P4 800mhz FSB/HT, AXP 2.0Ghz and up, or any A64 with at least 1Gig of ram) a good PSU or are willing to get one and you fit one of the following then do yourself a favor and consider an X1950PRO:
1) Don't have the time for a complete rebuild of your rig just to game.
2) Don't have the money to buy a new mobo and/or CPU and/or RAM AND a new GPU
3) Don't want to upgrade to a cheap mobo just for PCIE.
4) Are generally satisfied with you rig except for games.
So there it is. I hope you guys find this useful! It took forever to format!
Facebook
Twitter