rasz_pl wrote:

SPBHM wrote:

well

https://images.anandtech.com/old/cpu/intel-pe … um3/Image94.gif

so I found the source https://www.anandtech.com/show/254/10, but I still have no clue what exactly "Rage's Dispatch demo" is.
I was right there in the thick of it in 1999, both gaming and working in the industry, and I cant recall any Rage's Dispatch game. Is this it https://www.mobygames.com/game/rage ?

well, that would be some game by Rage Software at the time (the ones that made Incoming), so I guess it was renamed or cancelled.

dionb wrote:

SSE allows you to manipulate multiple floating-point data with a single instruction. If you have an application that is bottlene […]
Show full quote

NostalgicAslinger wrote:

[...]

So much performance increase with SSE? 😲

SSE allows you to manipulate multiple floating-point data with a single instruction. If you have an application that is bottlenecked on doing exactly that and supports these instructions, you will save multiple instructions so multiple cycles and get a performance increase in the order of magnitude you see here.

Question is: how realistic is this workload.
Answer: not, at least not for the vast majority of applications you might throw at one of these CPUs.

yes, I posted because it's interesting to say the least, but don't expect anything like that in other games...
still, it would be interesting to see actual benchmarks with thunderbird vs palomino at the same clock on newer titles (thinking up to mid 2000s) I guess!

question, didn't GTA 3 use SSE to a good extent? I remember it recommended Pentium 3 even with low clock (on katmai range) and said specifically to avoid Pentium II and k6, or maybe I'm wrong...

but, I remember when I used some patch on early Skyrim from x87 to SSE2 (not the same as SSE obviously) it bumping performance from 24 to over 40fps in one bad spot with my E5200

The latest 3dfx Voodoo 4/5 Win 9x/ME reference drivers allow the "Geometry Assist" Option in the driver. This is a T&L emulated option, because 3dfx cards don't have a Hardware T&L unit. It will use the SSE or 3Dnow! unit of the CPU, so if you use a Pentium II or Celeron Mendocino, the option doesen't make any differences.

If the option is enabled, and you have the right CPU, 3DMark 2000 (for example) allows the D3D Hardware T&L mode. It gives a little bit more performance on a PIII with SSE, on the Athlon with 3DNow! is not so much difference with this option (I have tested it)! A comparison with on and off would be interesting. I will test it on my PIII-S 1400 with V5 5500 AGP and Athlon 1 GHz Slot A with V4 4500 AGP. My two Athlon XP machines have Win 2000 installed, so I can not test it here.

The Windows 2000 drivers don't support this option. I think, because Win 9x allows a direct hardware access and the NT kernel doesen't allow it.

because Win 9x allows a direct hardware access

No, it doesn't.

The Serpent Rider wrote:

because Win 9x allows a direct hardware access

No, it doesn't.

I have read this in another forum (just do not find it...), why a voodoo 1 or 2, doesen't work in D3D mode under NT systems.

Here another find: https://boards.straightdope.com/sdmb/showpost … 499&postcount=4

95, 98, and ME are all designed to be backwards compatible to MS-DOS. They allow DOS programs to do what DOS programs normally do, which is access all of the hardware themselves (since DOS didn't have drivers, there wasn't much of a choice, the software had to access the hardware itself).

NostalgicAslinger wrote:

The latest 3dfx Voodoo 4/5 Win 9x/ME reference drivers allow the "Geometry Assist" Option in the driver. This is a T&L emulated option, because 3dfx cards don't have a Hardware T&L unit. It will use the SSE or 3Dnow! unit of the CPU, so if you use a Pentium II or Celeron Mendocino, the option doesen't make any differences.

If the option is enabled, and you have the right CPU, 3DMark 2000 (for example) allows the D3D Hardware T&L mode. It gives a little bit more performance on a PIII with SSE, on the Athlon with 3DNow! is not so much difference with this option (I have tested it)! A comparison with on and off would be interesting. I will test it on my PIII-S 1400 with V5 5500 AGP and Athlon 1 GHz Slot A with V4 4500 AGP. My two Athlon XP machines have Win 2000 installed, so I can not test it here.

The Windows 2000 drivers don't support this option. I think, because Win 9x allows a direct hardware access and the NT kernel doesen't allow it.

That's very interesting. I had always wondered why 3DMark2000/2001 reported that my V3-3500 could do hardware T&L.

On SSE: it can be very helpful if you're watching video on your retro rig for some bizarre reason. 😉
PIII-S @ 1629 MHz can decode 720p H.264 in software with ease. Even profile 4.1 encodes at highish bitrates (8-10 mbps) are completely smooth. Tbird at 1525 MHz? Not a chance, and I doubt that an extra 100MHz (to achieve clock parity with the PIII) would help.

Standard Def Steve wrote:

That's very interesting. I had always wondered why 3DMark2000/2001 reported that my V3-3500 could do hardware T&L.

On SSE: it can be very helpful if you're watching video on your retro rig for some bizarre reason. 😉
PIII-S @ 1629 MHz can decode 720p H.264 in software with ease. Even profile 4.1 encodes at highish bitrates (8-10 mbps) are completely smooth. Tbird at 1525 MHz? Not a chance, and I doubt that an extra 100MHz (to achieve clock parity with the PIII) would help.

My other Athlon 1400 Socket A Thunderbird should do 1600MHz, but better would be a comparison between 1,4GHz Tualatin and 1,4GHz Thunderbird. Also interesting: Athlon XP Palomino 1600+ 1,4Ghz with KT133A and 1,4GHz Tualatin with i815EP, both with SSE support to see, which CPU with SSE and same frequency is better.

SPBHM wrote:

rasz_pl wrote:

SPBHM wrote:

well

https://images.anandtech.com/old/cpu/intel-pe … um3/Image94.gif

so I found the source https://www.anandtech.com/show/254/10, but I still have no clue what exactly "Rage's Dispatch demo" is.
I was right there in the thick of it in 1999, both gaming and working in the industry, and I cant recall any Rage's Dispatch game. Is this it https://www.mobygames.com/game/rage ?

well, that would be some game by Rage Software at the time (the ones that made Incoming), so I guess it was renamed or cancelled.

that is quite conveniently unverifiable test right there ":) huge gains in a single unobtanium never released piece of software. Few vendors, Intel among them, loved to play games like this with reviewers. Anandtech seems to be the only entity to ever own this software, and they used it it exclusively to demonstrate P3 superiority.

another test, this time it changed name to Dispatched https://www.anandtech.com/show/260/9

and bingo https://www.tgdaily.com/games-and-entertainme … d-for-ten-years

At the time, I was working for Intel and was involved in the launch of the Pentium 3, aka Katmai.

We engaged a number of games manufacturers to provide demos showcasing not only Screaming Sindy's Extensions, but the arcane and mysterious Katmai New Instructions.

One such outfit was Rage Software, now sadly deceased. Rage provided demos of Incoming and an early prototype of a game called Dispatched, which as far as I know never actually saw the light of day. Dispatched featured a strangely-arousing cat riding a jet powered motorcycle. The first version I saw was running on a 400MHz Katmai and was still in wireframe. It was bloody impressive.

and https://www.hardware.fr/articles/95-5/jeux-optimises.html

Let's start with Dispatched first. This is actually a Rage Software game that should come out late 99, which Intel showed the demo at Comdex Fall to highlight the benefits of the SSE. Big interest, it is possible to enable or disable the use of SSE instructions at any time.

Nothing to say in terms of speed, it goes squarely faster once the SSE activated, + 50% to + 100% depending on the scenes! But looking closely at the demo, we notice - as you can see on the screenshots - that the SSE version is less detailed than the non-SSE version (see the ground). Intel would you try to roll the journalists in the flour?

so it was pretty much an Intel commissioned demo piece to showcase P3 during Comdex Fall, and was cheating with details
Two other SSE patched games mentioned on hardware.fr actually ran slower with SSE 😀

I also try Unreal Tournament, a beta of which was introduced to highlight the Pentium III. Two solution, my beta was not optimized Pentium III, the optimization is almost zero.
In short, for now it is the 0 pointed in the games for the SSE

Looks like this may have been its one-time Internet home prior to October 1999: http://web.archive.org/web/19991009201102/htt … ed/dispatch.htm

SPBHM wrote:

but, I remember when I used some patch on early Skyrim from x87 to SSE2 (not the same as SSE obviously) it bumping performance from 24 to over 40fps in one bad spot with my E5200

I remember that too. Core 2 has really substantial SIMD throughput though so going from x87 to SSE2 should be exciting indeed.

It's not much difference:

melbar wrote:

It's not much difference:

The Morgan is 300MHz faster and has a SSE unit, could also make differences. I would compare the 1300 Duron with a 1300 Athlon, but better would a downclocked Athlon XP Palomino (because it also has SSE like the Morgan). Clocked down to 1300MHz for a fair comparison. The Athlon XP 2100+ would be the right CPU, because of the 13x multiplicator that you need for 100x13=1300MHz.

NostalgicAslinger wrote:

melbar wrote:

It's not much difference:

The Morgan is 300MHz faster and has a SSE unit, could also make differences. I would compare the 1300 Duron with a 1300 Athlon, but better would a downclocked Athlon XP Palomino (because it also has SSE like the Morgan). Clocked down to 1300MHz for a fair comparison. The Athlon XP 2100+ would be the right CPU, because of the 13x multiplicator that you need for 100x13=1300MHz.

Don't forget the cache - SSE is almost irrelevant, having far more L2 cache is anything but. An Athlon 1300B would beat a Duron 1300, although only by a small margin. An Athlon C clocked at 1266 would make mincemeat of it. That doens't make it a bad CPU; those things were truly excellent value, their performance per MHz unbeatable - but regardless they were a low-end part, albeit a very good one.

95, 98, and ME are all designed to be backwards compatible to MS-DOS. They allow DOS programs to do what DOS programs normally do, which is access all of the hardware themselves (since DOS didn't have drivers, there wasn't much of a choice, the software had to access the hardware itself).

It has nothing to do with windows games which work through API.