ScardyBobScardyBobJoin Date: 2009-11-25Member: 69528Forum Admins, Forum Moderators, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Shadow
Alright, back on my desktop so I wanted to do some additional benching to give an idea of the difference. Also, just got a GTX 670 to replace my HD 6950, so I wanted to do a before and after bench to show the difference (will update with the GTX 670 info when I've swapped them out). Info and specs below.
ScardyBobScardyBobJoin Date: 2009-11-25Member: 69528Forum Admins, Forum Moderators, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Shadow
Got my GTX 670 installed and was able to do some benches before and after the switch (from my HD 6950) to see the difference. Did the benchmark on two resolutions (1920x1080 and 1280x1024) in addition to high and low detail. The demo I used is attached if anyone wants to try it. Also, it looks like vsync got enabled on my HD 6950, 1280x1024, High bench.
^Never mind did not read the entire post. Seems like on low it barely affected the performance levels. Thanks for posting this though most people know the CPU is the bottleneck (except on high settings), I always wanted to see someone actually benchmark the differences.
^Never mind did not read the entire post. Seems like on low it barely affected the performance levels. Thanks for posting this though most people know the CPU is the bottleneck (except on high settings), I always wanted to see someone actually benchmark the differences.
Pretty much. I did notice that my setup was GPU-limited on the high settings (e.g. 'waiting for GPU' was in the >5ms range), so it makes sense that dropping it to low or reducing the resolution would make a difference. However, it doesn't matter how low you drop the resolution if your CPU-limited.
Has anyone tried using nVidia's new Frame Capture Anaylsis Tool to do any frame time benchmarking with NS2? I was going to try it myself but to get the most accurate reading you need a game capture card which I do not have.
HIGH Config: everything turn on and high
CPU: AMD Phenom II X4 955 Black Edition @ 3200.0 MHz / 3.2Ghz
GPU: ATI/AMD Radeon HD 6850 (BARTS PRO) @ 1 GB
OS: Windows 7 Professional, 64 Bit
RAM: 8GB, 666.7 MHz (PC3-10600) (99U5471-020.A00LF)
Minimum FPS: 24
Maximum FPS: 53
Average FPS: 37.141
Build: 248
Resolution: 1920 x 1080
Ping: ~44ms
MAP:
ns2_veil
LOW Config: everything turned off or minimum (except multi-core rendering)
CPU: AMD Phenom II X4 955 Black Edition @ 3200.0 MHz / 3.2Ghz
GPU: ATI/AMD Radeon HD 6850 (BARTS PRO) @ 1 GB
OS: Windows 7 Professional, 64 Bit
RAM: 8GB, 666.7 MHz (PC3-10600) (99U5471-020.A00LF)
Minimum FPS: 26
Maximum FPS: 110
Average FPS: 63.341
Build: 248
Resolution: 1920 x 1080
Ping: ~44 ms
MAP:
ns2_veil
sorry no excel!
Please follow these guidelines while capturing your benchmark:
YES - Run FRAPS Benchmark for 20 Minutes (1200 Seconds)
YES - DO NOT benchmark while in the readyroom. Wait until you have joined a team, and the round has started.
YES - DO NOT benchmark loading screens.
NO, was no server, it was hard enough to keep 2 times fully 20 minutes on veil...concede makes it hard - EITHER benchmark on NS2_SUMMIT or NS2_DESCENT (This will help keep the benchmarks somewhat consistent.)
YES - ONLY benchmark a single round. DO NOT benchmark multiple rounds which return you to the Readyroom or display the Loading screen.
YES - ONLY benchmark while running the game FULLSCREEN.
YES - EITHER benchmark on "High" Visual Detail Settings and ENABLE everything (Bloom, Atmospherics, etc.) or DISABLE everything pretty and benchmark on the lowest possible settings (No Ambient Occlusion, AA, etc.)
The High config makes the game very gummy, i mean like a input delay :P
not playable on high.
Update new cpu:
LOW Config: everything turned off or minimum (except multi-core rendering)
CPU: Intel(R) Core(TM) i5-3550 CPU @ 3.30GHz
GPU: ATI/AMD Radeon HD 6850 (BARTS PRO) @ 1 GB
OS: Windows 7 Professional, 64 Bit
RAM: 8GB, 666.7 MHz (PC3-10600) (99U5471-020.A00LF)
Minimum FPS: 44
Maximum FPS: 130
Average FPS: 80.155
Build: 249
Resolution: 1920 x 1080
Ping: ~120 ms
MAP:
ns2_descent
[Frames: 96186 - Time: 1200000ms - Avg: 80.155 - Min: 44 - Max: 130]
Good to know that you can get a cpu powerful enough to put the load back on the gpu.
You certainly can. I have an Intel i7 2600k overclocked to 4.6Ghz and I rarely see any number except 0 before "ms waiting on render thread," in r_stats 1.
My GPU is the bottleneck (I have a GTX 580). It usually has 2/3 ms waiting for GPU.
So for you the ms waiting for gpu is always above 0? I had a 5.0 OC at one point but I cannot for the life of me recall if it put the bottleneck back on the gpu.
Not *all* the time, but the majority of the time. Late game with lots of models and players running all over the place it can put a BIG load on my CPU, but the majority of the time my bottleneck is my GPU (which is not overclocked).
I was running a 2600k with HT off during the 5.0 OC. Since had a fault in the psu that fried the whole pc, now running a 3570k but only at 4.2 because I'm an idiot and my new mobo is too small for my custom cooler
I was running a 2600k with HT off during the 5.0 OC. Since had a fault in the psu that fried the whole pc, now running a 3570k but only at 4.2 because I'm an idiot and my new mobo is too small for my custom cooler
That's a bummer (that you fried your old PC). Sounds like you have a nice new one though, even with the small mobo.
It's interesting that you had yours up to 5.0 and I'm only at 4.6 (HT is on in my rig). But, I guess every chip is a little different. I have been shocked with the performance of my chip at 4.6Ghz. It runs like a champ, cool and stable. I mostly don't push it any further because I use this rig as my daily work / school / everything PC and I'd like the chip to have a good, long life. So, I keep the temps and volts down whenever possible. Plus, I have it set up to throttle down when it isn't in use. So, it doesn't run at 4.6 all the time. I think it throttles down to 1.5Ghz or something when web surfing, using Word, etc.
Turning HT off allows you to use less vcore for higher clocks and less heat is generated. An i7 with HT off is better than an i5 as it's a higher quality chip. That's how I got to 5.0 I assume. Was running at it for about 3 months.
Hmm... I didn't know that, but it makes sense. Maybe I can play around with that sometime.
I am running air—a coolermaster CPU heatsink in push/pull and things stay pretty cold. I'm sure I have plenty more headroom above the 4.6Ghz mark, at least to 4.8 with HT off.
Yeah I'd see what you can get with HT off, be prepared to drop your vcore even going up a few clocks. The reason I decided to turn it off to get more clocks is I don't actually do anything that needs more than 4 threads. So I decided more power per core was more important for things like games (that use 1-2 cores max) than less power and more threads.
Comments
Type: Sager NP9130/P151EM1
CPU: Intel i7 3720QM (2.6GHz; Turbo 3.6GHz)
GPU: Nvidia GeForce GTX 670M 1.5GB
Operating System: Windows 7 Ultimate 64-bit
RAM: 16GB DDR3 1600
Build: 242
Resolution: 1366x768
Ping: ~ 60-80
Visual Detail: Low (everything off)
ns2_refinery
Minimum FPS: 3
Maximum FPS: 65
Average FPS: 43.563
ns2_docking
Minimum FPS: 25
Maximum FPS: 65
Average FPS: 43.055
ns2_summit
Minimum FPS: 25
Maximum FPS: 67
Average FPS: 46.173
DxDiag: http://pastebin.com/vYxvj23R
CPU: Intel i7 2600k OCd to 4.7GHz
GPU: AMD Radeon HD 6950 2GB GDDR5
Operating System: Windows 7 Ultimate 64-bit
RAM: 16GB DDR3 1333
Build: 243
Resolution: 1920x1080
Ping: ~ 60-80
ns2_docking
Visual Detail: Low (everything off)
Minimum FPS: 0
Maximum FPS: 130
Average FPS: 82.055
ns2_summit (first bench)
Visual Detail: Low (everything off)
Minimum FPS: 15
Maximum FPS: 138
Average FPS: 101.089
ns2_summit (second bench)
Visual Detail: Low (everything off)
Minimum FPS: 16
Maximum FPS: 132
Average FPS: 92.424
ns2_summit (third bench)
Visual Detail: High (everything on)
Minimum FPS: 14
Maximum FPS: 79
Average FPS: 41.015
DxDiag: http://pastebin.com/gTH77q9K
CPU: Intel i7 2600k OCd to 4.7GHz
Operating System: Windows 7 Ultimate 64-bit
RAM: 16GB DDR3 1333
Build: 243
Map: ns2_summit
Players: 24
Ping: ~ 60-80
Demo: http://dl.dropbox.com/u/5490563/Bench_B243_Summit_24.zip
GPU: AMD Radeon HD 6950 2GB GDDR5
Resolution: 1920x1080
Visual Detail: Low
Minimum FPS: 63
Maximum FPS: 166
Average FPS: 98.348
Resolution: 1920x1080
Visual Detail: High
Minimum FPS: 35
Maximum FPS: 76
Average FPS: 49.267
Resolution: 1280x1024
Visual Detail: Low
Minimum FPS: 70
Maximum FPS: 174
Average FPS: 111.375
Resolution: 1280x1024
Visual Detail: High
Minimum FPS: 55
Maximum FPS: 64
Average FPS: 60.006
GPU: Nvidia Geforce GTX 670 2GB GDDR5
Resolution: 1920x1080
Visual Detail: Low
Minimum FPS: 61
Maximum FPS: 162
Average FPS: 102.353
Resolution: 1920x1080
Visual Detail: High
Minimum FPS: 49
Maximum FPS: 121
Average FPS: 77.386
Resolution: 1280x1024
Visual Detail: Low
Minimum FPS: 70
Maximum FPS: 178
Average FPS: 109.031
Resolution: 1280x1024
Visual Detail: High
Minimum FPS: 67
Maximum FPS: 158
Average FPS: 105.431
http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-Dissected-Full-Details-Capture-based-Graphics-Performance-Testin
^ download link and how it works.
lookin to see if anyones got a 7950 + 3770k, and what it's benching in current build.
NOTE: Map was ns2_refinery, first 20 minutes of match, no late-game performance.
The following in-game settings were used:
CPU: Intel Core i5 3570K @ 4.2Ghz
GPU: AMD/ATI Radeon HD 4870 X2
OS: Windows 8 64 Bit
RAM: 16GB DDR3 1600mhz
Minimum FPS: 0
Maximum FPS: 132
Average FPS: 69
Build: 243
Resolution: 1600 x 1200
Ping: ~45 ms
GPU: GTX 660 2 GB
OS: Windows 7 64-bit
RAM: 8 GB DDR3-1600
Build: 244
Resolution: 1920 x 1080
Visual Detail: High
Ping: ~150 ms
ns2_docking:
Minimum FPS: 35
Maximum FPS: 103
Average FPS: 58.120
ns2_mineshaft:
Minimum FPS: 33
Maximum FPS: 74
Average FPS: 52.089
CPU: AMD Phenom II X4 955 Black Edition @ 3200.0 MHz / 3.2Ghz
GPU: ATI/AMD Radeon HD 6850 (BARTS PRO) @ 1 GB
OS: Windows 7 Professional, 64 Bit
RAM: 8GB, 666.7 MHz (PC3-10600) (99U5471-020.A00LF)
Minimum FPS: 24
Maximum FPS: 53
Average FPS: 37.141
Build: 248
Resolution: 1920 x 1080
Ping: ~44ms
MAP:
ns2_veil
LOW Config: everything turned off or minimum (except multi-core rendering)
CPU: AMD Phenom II X4 955 Black Edition @ 3200.0 MHz / 3.2Ghz
GPU: ATI/AMD Radeon HD 6850 (BARTS PRO) @ 1 GB
OS: Windows 7 Professional, 64 Bit
RAM: 8GB, 666.7 MHz (PC3-10600) (99U5471-020.A00LF)
Minimum FPS: 26
Maximum FPS: 110
Average FPS: 63.341
Build: 248
Resolution: 1920 x 1080
Ping: ~44 ms
MAP:
ns2_veil
sorry no excel!
Please follow these guidelines while capturing your benchmark:
YES - Run FRAPS Benchmark for 20 Minutes (1200 Seconds)
YES - DO NOT benchmark while in the readyroom. Wait until you have joined a team, and the round has started.
YES - DO NOT benchmark loading screens.
NO, was no server, it was hard enough to keep 2 times fully 20 minutes on veil...concede makes it hard - EITHER benchmark on NS2_SUMMIT or NS2_DESCENT (This will help keep the benchmarks somewhat consistent.)
YES - ONLY benchmark a single round. DO NOT benchmark multiple rounds which return you to the Readyroom or display the Loading screen.
YES - ONLY benchmark while running the game FULLSCREEN.
YES - EITHER benchmark on "High" Visual Detail Settings and ENABLE everything (Bloom, Atmospherics, etc.) or DISABLE everything pretty and benchmark on the lowest possible settings (No Ambient Occlusion, AA, etc.)
The High config makes the game very gummy, i mean like a input delay :P
not playable on high.
Update new cpu:
LOW Config: everything turned off or minimum (except multi-core rendering)
CPU: Intel(R) Core(TM) i5-3550 CPU @ 3.30GHz
GPU: ATI/AMD Radeon HD 6850 (BARTS PRO) @ 1 GB
OS: Windows 7 Professional, 64 Bit
RAM: 8GB, 666.7 MHz (PC3-10600) (99U5471-020.A00LF)
Minimum FPS: 44
Maximum FPS: 130
Average FPS: 80.155
Build: 249
Resolution: 1920 x 1080
Ping: ~120 ms
MAP:
ns2_descent
[Frames: 96186 - Time: 1200000ms - Avg: 80.155 - Min: 44 - Max: 130]
You certainly can. I have an Intel i7 2600k overclocked to 4.6Ghz and I rarely see any number except 0 before "ms waiting on render thread," in r_stats 1.
My GPU is the bottleneck (I have a GTX 580). It usually has 2/3 ms waiting for GPU.
Not *all* the time, but the majority of the time. Late game with lots of models and players running all over the place it can put a BIG load on my CPU, but the majority of the time my bottleneck is my GPU (which is not overclocked).
What CPU are you running?
That's a bummer (that you fried your old PC). Sounds like you have a nice new one though, even with the small mobo.
It's interesting that you had yours up to 5.0 and I'm only at 4.6 (HT is on in my rig). But, I guess every chip is a little different. I have been shocked with the performance of my chip at 4.6Ghz. It runs like a champ, cool and stable. I mostly don't push it any further because I use this rig as my daily work / school / everything PC and I'd like the chip to have a good, long life. So, I keep the temps and volts down whenever possible. Plus, I have it set up to throttle down when it isn't in use. So, it doesn't run at 4.6 all the time. I think it throttles down to 1.5Ghz or something when web surfing, using Word, etc.
I am running air—a coolermaster CPU heatsink in push/pull and things stay pretty cold. I'm sure I have plenty more headroom above the 4.6Ghz mark, at least to 4.8 with HT off.
- GPU: Amd Radeon HD 6850 1Gb
- Operating System: Windows 7 64bit
- RAM: 12 Gb DDR3 1333Mhz
- Minimum FPS: 27
- Maximum FPS: 132
- Average FPS: 72
- Build of NS2: 261
- Resolution: 1680x1050
- Ping: 70-90 ms
- Map: Summit
- Visual Detail: All High except Shadows=off and Infestation=low