VOGONS


First post, by st31276a

User metadata
Rank Member
Rank
Member

I remembered Intel’s ridiculous marketing campaign about the SSE instructions of the P3 making your internet faster. Of course this was an absurd claim in the days of 56k modems, but I just experienced some late arriving truth to that statement.

I have this Mendocino 400 system on a GA-6VA with 128 ram and windows 98. I happened to come upon a Katmai 550 out of a Dell GX1, which I promptly tried to install.

Although its funny proprietary heatsink is a little bit too wide and bends the retention clips outwards by about 5 mm at the top, it plugged into the slot and fired right up. I added another 128 ram for good measure, although I doubt this made a huge difference since it was not swapping excessively. However, unreal and ut99 now run very smooth on the voodoo2, where they were rather choppy before. What surprised me the most though, is how much faster browsing has become in firefox 2.0.0.20. When I loaded vogons threads in the past, I could see it rendering the tables on the page, taking about two seconds to complete. Now the threads just snap open as I click on them. Pages witb lots of images can scroll quickly without sticking and stuttering. I was expecting an incremental increase, this is huge! Subjectively this is about a 500% improvement.

The system speed increasing from 66 to 100 MHz plus the extra cache and SSE really plays its part.

The win98 ip stack unfortunately did not hurry up much. I now download at 440 kilobytes per second instead of 420. 3c905B, fast internet, fast site, https with aes256.

I wanted to post this on the system, but the old firefox does not want to connect to Vogons with ssl, so I am tapping this out on my phone.

I’ll figure something out to mount the cpu properly.

Reply 1 of 3, by the3dfxdude

User metadata
Rank Oldbie
Rank
Oldbie

I suspect it's mainly the FSB and the cache.

Firefox 2.0.0.20 came out in Dec 18, 2008. Granted, the 2.x series began on Oct 14, 2006. The P3 came out Feb '99, with that marketing campaign, which is over 7 years earlier. I'm not sure if the first releases of Firefox always was optimized with SSE, or optimized with anything else at all, but it I suspect there is a strong possibility it never had been optimized for any older processor. But when firefox came out in 2002, people were already moving on from the P3 to the P4 and Athlon XP, which also has SSE.

So one thing to do to isolate the SSE portion, would be to compare the P2 to P3 at 100mhz FSB at the same speed roughly, but I don't know if Firefox would be a fair comparison. It would have to be a browser that supported MMX or SSE. I don't know enough about Firefox in this case. One way to find out could be to check on an Athlon against the P3 of the same speed on Firefox, and then check both the P3 and Athlon on IE 5.0. How to measure performance in a browser accurately... I don't know. But it would be fun to see! It would be awesome also optimize a build of Firefox with 3dNow and try that on the Athlon.

I'd suspect that SSE was not that impactful then, but hard to believe it would have been so that much later. We're well past the P3 era in 2006.

Reply 2 of 3, by auron

User metadata
Rank Oldbie
Rank
Oldbie

yeah, not to mention going from a 400 to a 550. one would really be testing a p2 450 vs. a p3 450 and not put any extra RAM in either to make any kind of solid judgment about SSE. even then IIRC the katmai had a tweaked l1 cache controller that could account for a little bit of difference.

i think in one of the threads about processors with SSE running under windows 95 a method was mentioned for seeing whether a program uses SSE instructions...

Reply 3 of 3, by swaaye

User metadata
Rank l33t++
Rank
l33t++

Intel may have gotten some plugin vendors to optimize for SSE. I'm thinking things like Quicktime.

One of the things SSE targeted was MPEG2 decoding and it is very good at that.