Mancomb Seepgood, Internet Pirate
@16bitmanfred.bsky.social
220 followers 58 following 3.4K posts
Enjoying hardware and games from ISA to AGP graphics, with a special love for old sound cards.
Posts Media Videos Starter Packs
Pinned
16bitmanfred.bsky.social
I have been posting less and the reason is I had another epileptic seizure, had to raise medication, my hand tremors got even worse, and I got depressed over it.
Am officially looking for someone to help me fix some hardware.
I can't pay much, but might have vintage hardware for you instead.🧵
16bitmanfred.bsky.social
Oh, my bad, I kinda didn't see the part where the system POSTs, so it's really just the fans.
Yeah then it's either the PSU or the distribution from the board to the fan connectors - most other stuff is powered by the 5V rail, I don't know about HDD and CD drives, but CPU, GPU, floppy and so on are.
16bitmanfred.bsky.social
Uh... with that old a board it is very much possible that a card with the wrong voltage was inserted and fried the AGP slot.
In the beginning, the slots didn't have notches to prevent you from inserting wrong cards, but the board could actually be done for.
16bitmanfred.bsky.social
Finally feels like a peaceful autumn evening.
16bitmanfred.bsky.social
Yeah, Chris' 3D Benchmark is definitely the wrong choice for benchmarking different CPU configs.
I got the same 6.9 FPS for three different configs:
-533 MHz instruction cache disabled
-533 MHz instruction cache & branch prediction disabled
-533 MHz instruction cache & L2 cache disabled
16bitmanfred.bsky.social
If you had to run the same benchmark on 8086 to Pentium III Coppermine CPUs, is there one that would actually give out useable data on all of them?
Like the Quake timedemo often only gives you .1 FPS differences between 486 class configs but the DOOM timedemo may glitch out on fast machines.
16bitmanfred.bsky.social
The AWE64 CT3520 is currently inside my Neoware CA2 thin client, upgraded with an original Creative 8 MB memory expansion (CT1930), because I was unable to solve the problem with the Guillemot Maxi Sound 64 Dynamic 3D not outputting any sound despite the driver/resource side looking perfectly fine.
16bitmanfred.bsky.social
At the very least, it supports 3.3V CPUs as well as 5V, so you have the complete range of 486 CPUs available, which is nice.
I saw a guy on youtube push his AMD 5x86 to 200 MHz (4x50) on the same chipset, might have been Bits und Bolts.
16bitmanfred.bsky.social
With PCI slots and the UMC chipset, that's gotta be a very fast one.
Might even support up to 66 MHz FSB.
You found a real gem!
16bitmanfred.bsky.social
Like all AWE64 cards, it is unaffected by the hanging note bug, so you don't have to either fix it in hardware or software.
And yes, Creative CQM is at best the fourth best OPL3 alternative BUT that's only until you use the AWE's special sauce to add reverb and chorus effects to it!
16bitmanfred.bsky.social
One card that grew on me a lot lately is the Creative Sound Blaster AWE 64 CT4520.
Unlike the CT4500, this is about 20mm shorter and therefore fits in my thin client.
I managed to get my hands on an 8 MB memory expansion for some cool General MIDI sound fonts, since the default one sounds ass.
16bitmanfred.bsky.social
4:3 all the way!
Eizo F931 CRT: up to 2048x1536 at 75 Hz, up to 120 Hz at 800x600.
Dell Ultrasharp 2007 FPB IPS: up to 1600x1200 at 75 Hz.
16bitmanfred.bsky.social
2 MB is fine in most situations, unless you want to output high resolutions in Win98.
For DOS it's all you need, and the quality and compatibility of the card is excellent.
16bitmanfred.bsky.social
So I can document which specific settings make the VIA C3 Nehemiah equivalent to a 386-25, a 486 DX2-66 and such, for future reference.
16bitmanfred.bsky.social
This is a textbook example of Dunning Kruger.
Trish apparently still does not know how an LLM actually works.
Trish just happens to have found the solution to a problem that actual AI developers have been looking for for more than a decade & this doesn't make her suspect it might have been too easy.
desfitzgerald.bsky.social
Seeing people nonchalantly post stuff like makes me feel like I'm going crazy honestly.
A post from an academic on blue sky that says: "I'm learning about "prompt engineering" when asking LLMs to extract data. I now add this to all instructions:

Style
Be concise, analytic, and specific (cite page/figure if provided). If essential info is missing, keep going but flag Unclear and exact data needed. Never invent data.
 
#academicsky"
16bitmanfred.bsky.social
Well, in the end I automated the generation of an Excel sheet lol, now to find the single benchmark that best represents a CPU's DOS speed since it's bad enough to run it 336 times and I can't be arsed to run the numbers 772 times or more instead.
16bitmanfred.bsky.social
I gotta try all speed settings my VIA C3 Nehemiah can run at (4-14 multiplier in .5 increments=) 21 multiplier settings, plus branch prediction, instruction cache, l1 cache and l2 cache can each be set to enabled or disabled, that's 336 different speeds I believe, there must be a way to automate.
a woman with a bandana on her head is smiling and says aint nobody got time for that
ALT: a woman with a bandana on her head is smiling and says aint nobody got time for that
media.tenor.com
16bitmanfred.bsky.social
At its slowest, it's slower than a 386.
The important part really is to find a way to batch script each possible SETMUL setting and a benchmark that outputs its result somewhere so it can afterwards be sorted what score gave what result.
16bitmanfred.bsky.social
The funniest part is that Philscomputerlab had this "136 in 1" Pentium MMX thing, and this is exactly 200 more different speed settings.
The FSB is 100 MHz (can be set to 133 but then we get less working multipliers to work with) and at its peak it equals a 700 MHz Pentium III Coppermine.
16bitmanfred.bsky.social
I gotta try all speed settings my VIA C3 Nehemiah can run at (4-14 multiplier in .5 increments=) 21 multiplier settings, plus branch prediction, instruction cache, l1 cache and l2 cache can each be set to enabled or disabled, that's 336 different speeds I believe, there must be a way to automate.
a woman with a bandana on her head is smiling and says aint nobody got time for that
ALT: a woman with a bandana on her head is smiling and says aint nobody got time for that
media.tenor.com
16bitmanfred.bsky.social
I did not suggest that at all.
But in 2006, before everyone had smartphones, every household had a PC, so most people only had to add a graphics card to a device they already had anyway.
16bitmanfred.bsky.social
The PS3 sold for 499$ USD (20 GB model) or 599$ USD (60 GB model) at launch 9 days later.
I remember they weren't easy to get, and then you'd wait unusually long for the good exclusive titles, so for the first year it was more of a bluray player that could also play games.
16bitmanfred.bsky.social
One of three GTS versions came out the same day as the GTX though, November 8 2006, for 440$ MSRP, although sold for 409$ upwards.
16bitmanfred.bsky.social
While the S3 Vision 968 has no 3D functionality at all, and unlike the Trio64 V+ or V2/DX doesn't have its RAMDAC integrated, this ELSA Winner 2000 Pro/X specimen has 4 MB of VRAM and a 220 MHz RAMDAC (vs 135/170), allowing for much higher, ergonomical refresh rates.
The quality is astounding!
16bitmanfred.bsky.social
Looking forward to it!
The 8800 series was probably the biggest graphics cards gamechanger since the Voodoo 1!