In most applications K7 Thunderbird is just as fast as PIII Coppermine with similar clock. The catch is Athlon needs twice as much power to get this kind of performance...
HW museum.cz - my collection of PC hardware
I remember the dubious Pentium vs 486 discussions.
Often a run through Quake would lay the argument to rest. Even in Doom the Pentium, despite its low clock speed, did a great job.
Regarding optimization for the ol' Pentium - next to Abrash, Agner Fog did some outstanding work but unfortunately not much, if any, software was written that delivered the goods in an official product.
wrote:The Pentium II was basically a Pentium Pro with MMX added, with some small tweaks (mainly to fix the performance issues in 16-bit code). And the PIII was a PII with SSE added and some more small tweaks.
They are all known as the P6 microarchitecture.
Therefore all of them run at triple the speed of the Pentium, amirite? 😁
94 MHz NEC VR4300 | SGI Reality CoPro | 8MB RDRAM | Each game gets its own SSD - nooice!
wrote:Therefore all of them run at triple the speed of the Pentium, amirite? 😁
Not sure what you mean, but I believe I already said that a big part of the P6 microarchitecture was that it made the x86 scale much better in clockspeed.
Which we can't really see with the PPro itself yet, because manufacturing was not mature enough yet (it was a HUGE chip for its day), so it only ran at 200 MHz.
But PII and PIII ramped up the clockspeed very quickly, with virtually no changes to the original PPro microarchitecture as I said, so yes. Eventually it did run at triple the speed of the Pentium, and more.
wrote:just as I remember, was underlined the greatness of the original Pentium architecture
Fist couple of years there was no much need in Pentiums. That's why aggressive advertising wich had some inertia on 1995. I see normal evolution of x86 CPUs where it's doubtful to notice concrete ones.
wrote:basically P3 and Athlon perform about the same at the same clockspeed
Coppermine, not Katmai. in some apps
wrote:Coppermine, not Katmai. in some apps
I am talking about the microarchitecture of P6 vs K7, not specific implementations which tend to go back and forth a bit, depending on when they were released, and who implemented what new tech first (faster buses, more cache, higher clock speeds, SSE instructions etc).
I had an AMD 486DX4 100 MHz when a student in my class told us he got a Pentium 60. It had this calculation bug, so he had to put up with lots of jokes in that regard. He did swap the chip out, I think there was a recall.
When I upgraded, I went straight to a Pentium 133. That machine was a massive upgrade. 3dfx Voodoo also came out and I was in PC gaming heaven. That machine lasted me a very long time.
wrote:I had an AMD 486DX4 100 MHz when a student in my class told us he got a Pentium 60. It had this calculation bug, so he had to put up with lots of jokes in that regard. He did swap the chip out, I think there was a recall.
When I upgraded, I went straight to a Pentium 133. That machine was a massive upgrade. 3dfx Voodoo also came out and I was in PC gaming heaven. That machine lasted me a very long time.
My 386SX lasted longer... until the K62-350.. 😁 The only thing that was permitted at those expensive times was the ram upgrade.
So I didn't exactly lived the late 486 /Pentium I generation, but only with friends configs.
But when the 486DX4 and all the variants came out, weren't they sell like "low/middle budget" configurations? Someone that bought the last 486s already knew they were old right?
wrote:But when the 486DX4 and all the variants came out, weren't they sell like "low/middle budget" configurations? Someone that bought the last 486s already knew they were old right?
Yup, as I said, when the Pentium 60/66 were introduced, the 486DX2-66 was still the fastest x86.
The DX4 and other derivatives were released later. You could see them as 'Celeron'/budget options, where Pentium was the high-end.
This is the model that was always used in the PC market back in the day. 8088-based PCs were sold into the early 90s, as 'budget' options.
Likewise, 286/386SX didn't really become mainstream until 1990 or so, when the 486 was already out, and 386DX had been around for years as well as high-end option.
In my experience the 386DX never really became mainstream... except until the 386DX40 was released well into the 486-era, as a budget alternative to the 486 (competing with that other budget option, the 486SX).
So basically the older models were being sold as mainstream/budget, years after new CPUs were launched.
These days it's different, where you have a single line of CPUs from low-end to high-end, and older CPUs are discontinued much faster.
wrote:Yup, as I said, when the Pentium 60/66 were introduced, the 486DX2-66 was still the fastest x86. The DX4 and other derivatives w […]
wrote:But when the 486DX4 and all the variants came out, weren't they sell like "low/middle budget" configurations? Someone that bought the last 486s already knew they were old right?
Yup, as I said, when the Pentium 60/66 were introduced, the 486DX2-66 was still the fastest x86.
The DX4 and other derivatives were released later. You could see them as 'Celeron'/budget options, where Pentium was the high-end.This is the model that was always used in the PC market back in the day. 8088-based PCs were sold into the early 90s, as 'budget' options.
Likewise, 286/386SX didn't really become mainstream until 1990 or so, when the 486 was already out, and 386DX had been around for years as well as high-end option.
In my experience the 386DX never really became mainstream... except until the 386DX40 was released well into the 486-era, as a budget alternative to the 486 (competing with that other budget option, the 486SX).So basically the older models were being sold as mainstream/budget, years after new CPUs were launched.
These days it's different, where you have a single line of CPUs from low-end to high-end, and older CPUs are discontinued much faster.
At those times I didn't even know there was a 386DX-40. Basically friends had 386 from 16 to 25 or 486SX/DX-33. At school also some 286, the only ones where I could say to myself my pc was faster. 😁
wrote:Not sure what you mean, but I believe I already said that a big part of the P6 microarchitecture was that it made the x86 scale […]
wrote:Therefore all of them run at triple the speed of the Pentium, amirite? 😁
Not sure what you mean, but I believe I already said that a big part of the P6 microarchitecture was that it made the x86 scale much better in clockspeed.
Which we can't really see with the PPro itself yet, because manufacturing was not mature enough yet (it was a HUGE chip for its day), so it only ran at 200 MHz.
But PII and PIII ramped up the clockspeed very quickly, with virtually no changes to the original PPro microarchitecture as I said, so yes. Eventually it did run at triple the speed of the Pentium, and more.
94 MHz NEC VR4300 | SGI Reality CoPro | 8MB RDRAM | Each game gets its own SSD - nooice!
wrote:wrote:But when the 486DX4 and all the variants came out, weren't they sell like "low/middle budget" configurations? Someone that bought the last 486s already knew they were old right?
Yes, as a student I was always behind the curve 🤣
I moved to the DX4 from a 486DLC 40, so for me it was still quite an upgrade.
wrote:wrote:wrote:But when the 486DX4 and all the variants came out, weren't they sell like "low/middle budget" configurations? Someone that bought the last 486s already knew they were old right?
Yes, as a student I was always behind the curve 🤣
I moved to the DX4 from a 486DLC 40, so for me it was still quite an upgrade.
Well I used the 386SX so much that when I saw Windows 95 in the 1997 I was confused..i upgraded from W3.1 to Windows 98... poor me. 😁
wrote:
Indeed, the revolution in the P6 architecture was that they moved to a RISC-backend, translating x86 instructions on-the-fly. That's why the clockspeed could scale so much better than on Pentium and earlier architectures.
wrote:wrote:Indeed, the revolution in the P6 architecture was that they moved to a RISC-backend, translating x86 instructions on-the-fly. That's why the clockspeed could scale so much better than on Pentium and earlier architectures.
It was necessary, but a nasty thing to do from a programming standpoint. I think the Cyrix 6x86 had a feature where the programmer could program the RISC core, but that idea never really caught on. I guess Intel didn't want to worry about compatibility between their RISC cores. They left programmers/compilers stuck behind the artificial x86 layer, so in essence the code is being written for a phony CPU. How do you optimize code for something like that? People do it, but it's a damn ugly mess IMO.
In it's purest form, assembly should have direct 1:1 control of the CPU, and that's been lost unfortunately.
It's funny to think that in reality, the x86 architecture died with the Pentium. Everything since has been legacy support.
