How 'hardware' is nVidia's "hardware firewall"?

dderidex

Supreme [H]ardness
Joined
Oct 31, 2001
Messages
6,328
So, I've been tinkering with tweaks on my A64 system trying to get the most performance out of it, and using a half dozen or so benchmarks - but Prime95's benchmark mode seems to pretty well reflect the results on other tests.

I came across an interesting dip in performance after doing a driver update I finally tracked down....to enabling nVidia's "hardware firewall".

Summary (average of three tests):
Code:
			nVidia		Microsoft	MS + nV Uninstalled
			=====		=====		=====
384K FFT length: 	20.332 		20.257		20.212
448K FFT length: 	24.336 		24.310		24.231
512K FFT length: 	27.253 		27.245		27.190
640K FFT length: 	34.593 		34.536		34.459
768K FFT length: 	42.080 		42.030		41.943
896K FFT length: 	50.774 		50.750		50.588
1024K FFT length: 	57.366 		57.272		57.155
1280K FFT length: 	76.657 		76.592		76.370
1536K FFT length: 	91.647 		91.498		91.281
1792K FFT length: 	111.554 	111.421		111.108
2048K FFT length: 	126.180 	126.073		125.717


Take a look at these detailed numbers:

Using nVidia's "Hardware" Firewall:
AMD Athlon(tm) 64 Processor 3000+
CPU speed: 2300.20 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 20.389 ms.
Best time for 448K FFT length: 24.372 ms.
Best time for 512K FFT length: 27.289 ms.
Best time for 640K FFT length: 34.599 ms.
Best time for 768K FFT length: 42.117 ms.
Best time for 896K FFT length: 50.832 ms.
Best time for 1024K FFT length: 57.430 ms.
Best time for 1280K FFT length: 76.663 ms.
Best time for 1536K FFT length: 91.609 ms.
Best time for 1792K FFT length: 111.567 ms.
Best time for 2048K FFT length: 126.141 ms.

AMD Athlon(tm) 64 Processor 3000+
CPU speed: 2299.90 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 20.284 ms.
Best time for 448K FFT length: 24.330 ms.
Best time for 512K FFT length: 27.248 ms.
Best time for 640K FFT length: 34.581 ms.
Best time for 768K FFT length: 42.037 ms.
Best time for 896K FFT length: 50.734 ms.
Best time for 1024K FFT length: 57.337 ms.
Best time for 1280K FFT length: 76.673 ms.
Best time for 1536K FFT length: 91.589 ms.
Best time for 1792K FFT length: 111.481 ms.
Best time for 2048K FFT length: 126.197 ms.

AMD Athlon(tm) 64 Processor 3000+
CPU speed: 2299.86 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 20.323 ms.
Best time for 448K FFT length: 24.306 ms.
Best time for 512K FFT length: 27.223 ms.
Best time for 640K FFT length: 34.598 ms.
Best time for 768K FFT length: 42.086 ms.
Best time for 896K FFT length: 50.757 ms.
Best time for 1024K FFT length: 57.332 ms.
Best time for 1280K FFT length: 76.635 ms.
Best time for 1536K FFT length: 91.743 ms.
Best time for 1792K FFT length: 111.613 ms.
Best time for 2048K FFT length: 126.202 ms.

Using Microsoft Windows's "Software" Firewall:
CPU speed: 2300.04 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 20.231 ms.
Best time for 448K FFT length: 24.311 ms.
Best time for 512K FFT length: 27.250 ms.
Best time for 640K FFT length: 34.503 ms.
Best time for 768K FFT length: 42.070 ms.
Best time for 896K FFT length: 50.733 ms.
Best time for 1024K FFT length: 57.291 ms.
Best time for 1280K FFT length: 76.594 ms.
Best time for 1536K FFT length: 91.557 ms.
Best time for 1792K FFT length: 111.424 ms.
Best time for 2048K FFT length: 126.092 ms.

AMD Athlon(tm) 64 Processor 3000+
CPU speed: 2299.92 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 20.281 ms.
Best time for 448K FFT length: 24.272 ms.
Best time for 512K FFT length: 27.276 ms.
Best time for 640K FFT length: 34.556 ms.
Best time for 768K FFT length: 41.983 ms.
Best time for 896K FFT length: 50.773 ms.
Best time for 1024K FFT length: 57.258 ms.
Best time for 1280K FFT length: 76.604 ms.
Best time for 1536K FFT length: 91.467 ms.
Best time for 1792K FFT length: 111.413 ms.
Best time for 2048K FFT length: 126.052 ms.

AMD Athlon(tm) 64 Processor 3000+
CPU speed: 2300.18 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 20.260 ms.
Best time for 448K FFT length: 24.346 ms.
Best time for 512K FFT length: 27.209 ms.
Best time for 640K FFT length: 34.550 ms.
Best time for 768K FFT length: 42.036 ms.
Best time for 896K FFT length: 50.744 ms.
Best time for 1024K FFT length: 57.266 ms.
Best time for 1280K FFT length: 76.579 ms.
Best time for 1536K FFT length: 91.469 ms.
Best time for 1792K FFT length: 111.425 ms.
Best time for 2048K FFT length: 126.075 ms.

Interesting results, no? Granted, incredibly minor difference in performance - BUT...it's consistent across multiple tests. And it's an *improvement* in disabling a HARDWARE firewall and enabling a SOFTWARE firewall.

Odd.
 
I had forgotten about the runtime Apache service.

Uninstalling the firewall completely gave me even BETTER numbers!

Updated summary:
Code:
			nVidia		Microsoft	MS + nV Uninstalled
			=====		=====		=====
384K FFT length: 	20.332 		20.257		20.212
448K FFT length: 	24.336 		24.310		24.231
512K FFT length: 	27.253 		27.245		27.190
640K FFT length: 	34.593 		34.536		34.459
768K FFT length: 	42.080 		42.030		41.943
896K FFT length: 	50.774 		50.750		50.588
1024K FFT length: 	57.366 		57.272		57.155
1280K FFT length: 	76.657 		76.592		76.370
1536K FFT length: 	91.647 		91.498		91.281
1792K FFT length: 	111.554 	111.421		111.108
2048K FFT length: 	126.180 	126.073		125.717


Detailed Results for MS Firewall + nV firewall uninstalled
AMD Athlon(tm) 64 Processor 3000+
CPU speed: 2299.73 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 20.152 ms.
Best time for 448K FFT length: 24.196 ms.
Best time for 512K FFT length: 27.214 ms.
Best time for 640K FFT length: 34.476 ms.
Best time for 768K FFT length: 41.958 ms.
Best time for 896K FFT length: 50.603 ms.
Best time for 1024K FFT length: 57.159 ms.
Best time for 1280K FFT length: 76.331 ms.
Best time for 1536K FFT length: 91.293 ms.
Best time for 1792K FFT length: 111.077 ms.
Best time for 2048K FFT length: 125.715 ms.

AMD Athlon(tm) 64 Processor 3000+
CPU speed: 2299.82 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 20.269 ms.
Best time for 448K FFT length: 24.251 ms.
Best time for 512K FFT length: 27.199 ms.
Best time for 640K FFT length: 34.475 ms.
Best time for 768K FFT length: 41.910 ms.
Best time for 896K FFT length: 50.589 ms.
Best time for 1024K FFT length: 57.145 ms.
Best time for 1280K FFT length: 76.381 ms.
Best time for 1536K FFT length: 91.267 ms.
Best time for 1792K FFT length: 111.143 ms.
Best time for 2048K FFT length: 125.662 ms.

AMD Athlon(tm) 64 Processor 3000+
CPU speed: 2300.07 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 20.216 ms.
Best time for 448K FFT length: 24.246 ms.
Best time for 512K FFT length: 27.158 ms.
Best time for 640K FFT length: 34.427 ms.
Best time for 768K FFT length: 41.962 ms.
Best time for 896K FFT length: 50.572 ms.
Best time for 1024K FFT length: 57.162 ms.
Best time for 1280K FFT length: 76.397 ms.
Best time for 1536K FFT length: 91.284 ms.
Best time for 1792K FFT length: 111.104 ms.
Best time for 2048K FFT length: 125.773 ms.
 
baldrik said:
im sorry, but are we comparing 10ths of mili-seconds?
Granted, but we are talking about *increases* by disabling something in hardware and enabling it in software. We SHOULD be see *decreases* in performance, but it goes up, instead.

Begging the question of - what's the point? Why enable the "hardware firewall" in the first place?
 
dderidex said:
Granted, but we are talking about *increases* by disabling something in hardware and enabling it in software. We SHOULD be see *decreases* in performance, but it goes up, instead.

Begging the question of - what's the point? Why enable the "hardware firewall" in the first place?

It makes you feel safe, like locking the door and leaving the windows open :p
 
dderidex said:
Granted, but we are talking about *increases* by disabling something in hardware and enabling it in software. We SHOULD be see *decreases* in performance, but it goes up, instead.

Begging the question of - what's the point? Why enable the "hardware firewall" in the first place?

Perhaps because it does a better job of protection than the MS one?

I've got nVidia's firewall installed, and I really like its configuration and how it works. I think the MS firewall is great for the home user, but that nVidia's is better if you know a little about configuring systems.

That said, interesting results. I'd like to see them repeated with an nForce 4 Ultra board as well someday, with the ActiveArmor firewall.
 
I hate to split hairs, but even "hardware" firewalls use software (IOS anyone?)

The real test of the performance of a hardware vs software firewall would be when stressing the network connection.

If you are hell-bent-for-leather performance enthusiast, disable anything that sucks up CPU cycles and offload it to another device if possible.

Run the same tests with multiple machine portscanning/64KB floodpings etc to judge the value.
 
Back
Top