Reply 20 of 50, by Mau1wurf1977
- Rank
- l33t++
I believe Quake was one of these games that used Intel instructions to peform better. Other cpus like the K6 and Cyrix would struggle with these type of applications.
Nice cooler!
I believe Quake was one of these games that used Intel instructions to peform better. Other cpus like the K6 and Cyrix would struggle with these type of applications.
Nice cooler!
Just FYI, I'd just get the fastest K6-2 2.2v for as cheap as you can buy. They are abundantly available and should be CHEAP!
The 350 and 300 parts are just less overclockable versions of say the 450 and 500 parts 😉 (except the newer ones will be CTX core, newer versions slightly better overclockable and supporting 6x multi)
The K6-2+ and K6-3+ are basically die schrinks of the standard K6-3 (with the K6-2+ having half the cache disabled, I presume)
I believe Quake was one of these games that used Intel instructions to peform better. Other cpus like the K6 and Cyrix would struggle with these type of applications.
Not just Quake but other 3d games as well. Its due to the lower floating point performance of the AMD/Cyrix CPUs- they were focused on fast integer performance which is important for desktop/office/multimedia type tasks. This is what gave them higher clock-for-clock performance for these tasks and in benchmarks- but when it comes to gaming, the Intel processors were king right up until the Athlon came out. Floating point performance is critical for 3d gaming.
I had a Cyrix 6x86 PR166 and although it was quicker on the desktop, in 3d games like Quake and others, it was slower than a Pentium 120. The had an even weaker FPU than the K6.
I knew the Pentium would be faster in Quake clock for clock but I'm surprised that its lead is so large, even with a 100Mhz advantage the K6 can't beat the Pentium.
RG100 if you do some synthetic benchmarks like Sisoft Sandra, see if you can disable 3DNOW somehow- I predict that you will see the floating point performance is significantly lower than the Pentium. Very few games of this era took advantage of the 3DNOW instructions.
What a shame, as during my FSB scaling tests without L1 cache I found that the 6x86 chips are the fastest Socket 7 cpus you can get...
wrote:There is even a config.sys driver where you can set the CPU multi when you boot the machine 😀
I recall having had a utility to change the cpu settings [multiplier, maybe others] on the fly with k6/2+ and k6/3+ chips.
Sorry I forgot to mention that the config.sys tool is for DOS. For windows there are "on the fly" tools but I am unaware of such a tool for DOS...
What a shame, as during my FSB scaling tests without L1 cache I found that the 6x86 chips are the fastest Socket 7 cpus you can get...
Yes, the K6/6x86 always did well in synthetic benchmarks... I suspect that 3Dbench, being rather primitive low resolution 3d graphics isn't FPU intensive, as evidenced by the 486SX25 giving almost equal performance to a 486DX25 in your 3dbench database.
I've got a 386DX board here and a 387 copro to go with it... maybe when I get a chance I'll try 3dbench with and without the copro installed and see if there is any difference. Another thing to add to the 'to do' list 😜
In my own testing with 3dbench, it seems to be more reliant on video chipset and bus performance than anything- on the same motherboard, using the same cpu/BIOS settings etc, I doubled my framerate in 3dbench simply by going from an ISA ET4000 to a VLB S3 805.
What would be interesting to see, as part of your cpu scaling tests, is to throw in a gaming benchmark, like Quake timedemos or some other FPU intensive game.
However for what you are trying to achieve its probably not relevant- most 2.5D games like Doom, build engine etc were not FPU intensive at all. I remember Duke Nukem 3d ran faster on the 6x86 while Quake was slower.
If you are trying to scale your machine down to 486/386 level I don't think there are many games from that era that will require fast FPU performance. So the 6x86 might be perfect for that.
If you are squeamish, don't prod the beach rubble.
Yea as soon as you need grunt > Slot 1 Pentium 3 all the way 🤣
Socket 7 is great for "slowdown" projects. Gives you a 386DX with L1 and L2 off and a 486DX2 with L1 off and L2 on.
However might as well get a 486 as this gives you even more options with the Turbo button. But I went with Socket 7 as the board is ATX and the boards happen to be on ebay when I was looking...
Duke Nukem 3D ran quite well on my 486DX4-120, even in video modes between 320x200 and 640x480 (got a bit slow/choppy at the 640x480 end though). Quake of course clocked under 10 FPS. Daggerfall used a floating-point engine but used sprites for characters and monsters, so it was more playable but still noticeably sluggish.
wrote:Sorry I forgot to mention that the config.sys tool is for DOS. For windows there are "on the fly" tools but I am unaware of such a tool for DOS...
Yeah, it was ctu I was thinking of, a windows application.
wrote:Duke Nukem 3D ran quite well on my 486DX4-120, even in video modes between 320x200 and 640x480 (got a bit slow/choppy at the 640x480 end though). Quake of course clocked under 10 FPS. Daggerfall used a floating-point engine but used sprites for characters and monsters, so it was more playable but still noticeably sluggish.
Hehe the other day I tried Duke 3D on my Phenom II desktop. I made this DOS Usb Boot disk and it works great.
VESA 2 not an issue with todays video cards. I set it to 1024 x 768 and yea it looked amazing and it was super smooth. I couldn't measure the frame rate but easily 60 frames (or more) for sure...
Had no sound though 🙁
My AMD K6-2 300 and 350 chips have arrived. Also, I found a K6-2 500 in my attic. The only chip I don't have is a 450. Nevermind, it doesn't matter too much. I ran Quake (1.06 shareware in DOS 6.22), with exactly the same "fast settings" BIOS values, and in full screen.
K6-2: the revenge. 😉 After testing the K6-2 300 chip, I decided to see how far it could go. I ended up with it running @ ~440 MHz! (See speedsys below.) This was a multi of 3.5x (from 3), and an FSB of 125 (from 100). I needed to bump the voltage to 2.3v (from 2.2). The BIOS settings were maxed out. The external clock reading is wrong on the Speedsys image. It seems to have difficulty identifying OC'd FSB values.
However, I tried running Quake and it failed. I bumped the voltage up to 2.5v, but it still failed. However, Quake worked fine with the FSB reduced to 120. I got 67.7 FPS, which is half a point less than the OC'd P233 chip, in the table from my previous post.
The K6-2 300 chip is stable at 420 MHz. Multi = 3.5, FSB = 120, Voltage = 2.3 in DOS and 2.4 in Windows. Here's the Windows 98 screenshot. I ran all of the SuperPi tests up to and including 1M. There may be a bug in this app. When you resize the screen, the values get altered. The real score for the 1M test was 7 minutes and something. The Sandra images show the 420 clock speed (to the right), and also the CPU and mulitmedia tests.
Just for a bit of fun, I managed to get "the chomper" (a K6-2 300 MHz) up to 460 MHz! I needed to loosen up the BIOS settings, so that they were not "aggressive", and consequently the system's potency was reduced. Also, the CPU's voltage was set to 2.5. For the general timings, the multi was set to 4 (from 3), and the FSB was set to 115 (from 100).
Edit: I found some info about overclocking the K6-2 CPU here.
Sorry for this slight necro.... but anyone know how to clear the %&¤# cmos on this board ??? (ver 4.1)
Me stumped.... 😵
(bios password has accidentally? turned itself on it seems)
No mention of any jumper or even a blasted 'short circuit' in the manual...
(never understood the genius in using that last thing though.. I notice Asus has it on their version of this board)
Unplug from power, Remove the battery and let it sit for a minute or two? that should do it
If you are squeamish, don't prod the beach rubble.
Nah, sounds to 'easy'.... 😜
BUT, waddayouknow, that actually seemed to work!!
Beyond me though that the manual doesn't even mention how to do it...
Why does other/newer boards even have a jumper then ?
Ah, nevermind my tech ignorance...
Thanks a lot for quick tip !!! 😁
These days, removing the battery is all I do. I fried a board by accidentally leaving the jumper to CLEAR_CMOS and applying power 🙁
Btw, RG100 where the heck are you hiding mate!