VOGONS


First post, by NautilusComputer

User metadata
Rank Newbie
Rank
Newbie

So I grew up with PCs of the late 80's as my youngest computer memories and really remember machines of the early 90's much better. Been catching up on PC history through that time period, and in everything I've read and watch (and searched for) I haven't found a good answer for this question.

I can follow how we got bus speeds from 8 MHz (8.33333 MHz) up - it doubled to 16.67 (16), and doubled to 33.333 (33), and then doubled to 66.67 (66). Doubling performance is very easy to describe and easy for the general public to understand.

People are very familiar with 'fourths' or 'quarters,' so even 25/50/75 Mhz makes enough sense for me to follow why they'd make those - they're numbers that are familiar to people and also easy to follow; and easy to derive from their "binary exponential" brethren - 25 is 16.67x1.5; 50 is 33.33x1.5; 75 is a little weird because it would be 66x1.125 which is a weird multiplier - but 50x1.5 is nice and easy (I guess 50 could be 25x2 OR 16x2.5 as well).

But 83 MHZ...I mean, there's no seemingly "good" way to get there. 33x2.5 does, or 66x1.25, or 41.5x2. Is that why it never caught on, because there wasn't an 'even way' to multiply something to get there and by the time we did have the capability to handle things on what I'll call 'in-between steps,' we had 100 MHz which could be any or ALL of the above, practically (100x1, 66x1.5, 50x2, 40x2.5, 33x3, 25x4, 20x5, 16.67x6)?

Was it something to do with oscillators just not working at frequencies that fit well with 83? Manufacturing technology just not having tolerances tight enough for it? Something else?

Who was the first to try 83 MHz - as CPU *or* bus speed? When in history was this?

I'd love to read/hear anyone's tidbits they can shine on this.

Reply 1 of 24, by jakethompson1

User metadata
Rank Member
Rank
Member

The Pentium OverDrive 83 is the biggest (first?) example I can think of for CPU speed. Also, I believe the reason there is a 486DX4-100 rather than 486DX3-100 is that there was supposed to be a 486DX3-83.
I don't know if there is any problem with the 83 frequency so much as technology moving fast enough for there to be no need for an intermediate stop between 66 (or 75) and 100? For example, AMD and Cyrix made 486DX2/80 chips but Intel didn't.
166 MHz Pentiums, after all are 2.5*66.

83 MHz bus speed I believe was only a Cyrix thing. I've never seen one in person.

Reply 2 of 24, by devius

User metadata
Rank Oldbie
Rank
Oldbie

I always saw 83MHz FSBs as 66MHz + 16MHz. Besides the Pentium OverDrive mentioned, the next time I saw a mention of that bus speed was when I had a 440LX system back in 1998. It supported 75 and 83MHz FSB speeds, although what for I don't know since no Pentium II officially supported those speeds and I could never get the 83MHz FSB to run stable.

Reply 3 of 24, by Unknown_K

User metadata
Rank Oldbie
Rank
Oldbie

Reminds me of the original Apple G3 Wallstreets having a 83FSB bus along with the more common 66FSB that was quickly removed on the later Wallstreets (and maybe Beige G3 towers).

I figured there must have been an issue with 100Mhz SDRAM at the time as there was never 83Mhz RAM and 66Mhz would be too slow.

Collector of old computers, hardware, and software

Reply 4 of 24, by Repo Man11

User metadata
Rank Member
Rank
Member

It was an unofficial bus speed for several Socket 7 chipsets, including the Intel HX, TX, and the Via VP3. I've run K6-2+ CPUs on board with these chipsets at 500 MHz (6x83) with no issues. From back when this was cutting edge stuff: https://www.tomshardware.com/reviews/75,27.html

Reply 5 of 24, by Doornkaat

User metadata
Rank Oldbie
Rank
Oldbie

I guess that's just the +25% overclock of a 66MHz bus.
66.66MHz is the base line of the P5 generation (double of the previous 33.33MHz), then there's +12.5% (75MHz) and +25% (83.33MHz). Next stop is +50% (100MHz) and after that +100% (133.33MHz).
50MHz is the 25% underclock.
So far all numbers make sense. The 60MHz makes sense in context of early marginal Pentium chips that would only run at that speed.

Don't know if that's the reason but the numbers seem to make sense to me.

Reply 6 of 24, by Doornkaat

User metadata
Rank Oldbie
Rank
Oldbie

Oh and I think board manufacturers introduced 75MHz/83MHz bus speeds as overclocking steps before Intel released the 83MHz Pentium Overdrive.
CPU manufacturers probably just adapted and used the commonly existing bus settings. Later chipsets sometimes even feature proper PCI dividers for them.

Reply 7 of 24, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

83MHz not being popular was just a coincidence. If it ran well and the price was right, somebody would have bought it.

I think the reason the 486DX3-83 never came out was that it was a useless step in between the DX/2-66, DX/4-75 and DX/4-100.
The POD83 came out too late, and by that time nobody wanted an expensive upgrade for an almost obsolete platform.

The 83MHz bus had stability issues. When it came out, not many motherboards could handle it properly. You normally needed special memory in order to benefit from the increase is frequency, and it usually ran the PCI bus way out of spec. Not to mention it was something being pushed by Cyrix and frowned upon by Intel. Cyrix by that time was seen as budget at best, and duplicitous at worst.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 8 of 24, by rmay635703

User metadata
Rank Oldbie
Rank
Oldbie

I would argue
50mhz
55mhz
60mhz
FSBs

Were all equally unpopular

50mhz was really only for 2 CPU speeds and 1 very rare one

55mhz FSB was Cyrix only in a Single stepping (Rarer than 83mhz)

60mhz fsb only existed through 180mhz on normal CPUS not counting that one Winchip and was always the gimped value speed option.

66mhz was by far most common even well into the late Celeron era.

As for actual 83mhz clock you already had 80mhz and 90mhz (which were odd if you think about it) competition at that time drove Intel to up clock speeds much faster than they planned or would have liked.

Remember the original Pentium was supposed to arrive in 33mhz, 40mhz, 50mhz and 66mhz
They delayed and launched at 60mhz for similar reasons to the 83mhz debacle.

Socket 6 died for similar reasons it was too outdated to worry about 100mhz+ p24t support and not worth the investment By board makers

Reply 9 of 24, by alvaro84

User metadata
Rank Member
Rank
Member

Wait, 60MHz FSB was provided for the 60MHz and later the 90, 120 and 150MHz models, far more widespread than the other obscure frequencies you mentioned.
So it was backed by Intel, the leading manufacturer of the S5-S7 platform, while 75-83MHz was never official with its chipsets. Even though a 430TX can work really well at 83MHz.

Shame on us, doomed from the start
May God have mercy on our dirty little hearts

Reply 10 of 24, by shamino

User metadata
Rank Oldbie
Rank
Oldbie

By the time of Socket-5/7, simple oscillators were being replaced with a configurable clock generator IC on the motherboard. I would guess that the manufacturers of those chips were primarily listening to Intel, Cyrix, AMD, etc for guidance on what frequencies they'd like to have available. This could include frequencies that actual existing chipsets weren't being built to support, but might be on a roadmap.

Some of the odd frequencies ended up with little use outside of overclocking, but motherboards that emphasized this were pretty niche in the big picture, especially at that point in time. I tend to think that the manufacturers of such boards were probably not the main influence. Those frequencies probably originated from conversations with the CPU makers.

Reply 11 of 24, by auron

User metadata
Rank Member
Rank
Member
Doornkaat wrote on 2020-08-20, 22:29:

66.66MHz is the base line of the P5 generation

think you forgot about the pentium 75 here, which was apparently quite an OEM favorite for a while.

regarding the 83mhz bus speed, selling CPUs that effectively require you to run PCI off-spec is an odd move in hindsight... though i suppose the DX50 on VLB was somewhat of a precursor to this. were users supposed to raise PCI latency to help those 83mhz bus systems to more stability? what about possible IDE corruption, such as what can be observed on BX@133mhz, if unlucky?

Reply 12 of 24, by H3nrik V!

User metadata
Rank Member
Rank
Member
NautilusComputer wrote on 2020-08-20, 20:02:

I can follow how we got bus speeds from 8 MHz (8.33333 MHz) up - it doubled to 16.67 (16), and doubled to 33.333 (33), and then doubled to 66.67 (66). Doubling performance is very easy to describe and easy for the general public to understand.

Besides all the real good arguments in this thread, 83(.33333) is just a multiple of the 8.3333 😀

But yes, I agree that the 83 MHz bus doesn't seem "logical".

Please use the "quote" option if asking questions to what I write - it will really up the chances of me noticing 😀

Reply 13 of 24, by H3nrik V!

User metadata
Rank Member
Rank
Member
rmay635703 wrote on 2020-08-21, 02:26:

50mhz was really only for 2 CPU speeds and 1 very rare one

Would these be:

The Pentium 75 (and AMD K5-75)
The 486DX-50
And Pentium-50, which only came as an engineering sample?

Or am I missing out?

Please use the "quote" option if asking questions to what I write - it will really up the chances of me noticing 😀

Reply 15 of 24, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

Pentium 100 could also run in 2x50 mod, but probably almost never used unless you had a *really* crappy board.
50MHz FSB wasn't exactly unpopular. Even though only a few models used it, the P75 sold like hotcakes. Cyrix had the 6x86-100GP/P120+
The 486DX-50 was also a thing in high-end EISA workstations for a while.
Don't forget both AMD and Cyrix also made 486DX-50s. Some intel DX4 and Cyrix 5x86-100s also officially supported 50MHz FSB (also seldom used).
The 486DX-50 came out in 1991. It was supposed to be a real powerhouse, but making fast boards that could run stable at that speed proved too challenging, not to mention difficulties in manufacturing fast DRAM and SRAM to keep up. By the time the Socket5 chips came around 50MHz was already slow.

55MHz was something only used on the 6x86-133, and hardly any boards support it.
40MHz FSB was also an unpopular option for the 6x86-80GP/P90+. (this is only rare in the pentium world)
30MHz FSB was used by AMD 486DX4-90, which was probably OEM only. It may have also been intended for the Cyrix 5x86-120/4X, but that chip was never officially released. I think I might have heard that Intel possibly used a 30MHz FSB in a mobile version of a 486, but not too sure about that one.

So yes, 83MHz was far from the only unpopular FSB speed. The issue was mostly related to the PCI bus. If the bus ran too fast it was unstable, and if too slow then performance suffered. Therefore 60 or 66MHz on a PCI system was preferred. A few manufacturers tried to find clever work arounds like async mode, but usually the performance was still lousy.

I also agree that 60MHz was not unpopular. For a number of years the a 66MHz FSB pentium was basically unobtainable because they couldn't produce them in large enough quantities for an affordable price. P133 was the first mainstream 66MHz FSB Pentium. The 60MHz chips were far more popular until that time.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 16 of 24, by Doornkaat

User metadata
Rank Oldbie
Rank
Oldbie
auron wrote on 2020-08-21, 04:12:
Doornkaat wrote on 2020-08-20, 22:29:

66.66MHz is the base line of the P5 generation

think you forgot about the pentium 75 here, which was apparently quite an OEM favorite for a while.

I mentioned the 50MHz bus. I guess this is a misunderstanding, I was talking primarily about bus speeds, not core speeds.
What I mean is the Pentium was likely originally designed around the 66MHz bus speed. Board manufacturers came up with +12,5% (75MHz) and +25% (83MHz) overclocks while the 60MHz bus speed option was the result of poor yields on the original P5 chips. It was then simply simply continued to be used on socket 5.
The budget Pentium (P75) used a -25% underclocked (50MHz) bus so it wouldn't outperform last year's flagship CPU (the 66MHz Pentium) while still allowing Intel to sell their new platform within the mainstream market and probably getting rid of marginal/underperforming P54C chips.

Again, I don't claim this was absolute truth, it just makes sense to me looking at the numbers.

Reply 17 of 24, by mkarcher

User metadata
Rank Member
Rank
Member
Doornkaat wrote on 2020-08-21, 08:43:

The budget Pentium (P75) used a -25% underclocked (50MHz) bus so it wouldn't outperform last year's flagship CPU (the 66MHz Pentium) while still allowing Intel to sell their new platform within the mainstream market and probably getting rid of marginal/underperforming P54C chips.

Again, I don't claim this was absolute truth, it just makes sense to me looking at the numbers.

As I heard it 20 years ago, the P75 was meant to target the laptop market only, where you wanted lower FSB speed to decrease idle current. The 5V Pentium 60/66 never were a good fit for laptops due to their high power consumption (compared to contemporary CPUs), and IIRC the Pentium 66 actually preferred to be run at slightly above 5 Volts. The desktop market was targeted with the Pentium 90 and Pentium 100. Later on, the Pentium 75 was sold into the desktop market because OEMs demanded an entry-level socket 5 option to get rid of their socket 3 and socket 4 systems in the line-up. The "real" P75 also likely was a great competitor to the Am5x86-P75 (133 MHz), and helped getting those sales to Intel instead of AMD.

Reply 18 of 24, by Intel486dx33

User metadata
Rank l33t
Rank
l33t

The 83mhz Pentium Overdrive CPU was probably to expensive and did not justify the upgrade on a socket 3 motherboard.
The socket 3 motherboard had other limitations like bus speed and i/o and controller issues, bios limitations, video and memory performance issues.

Back in 1995 it was best to just upgrade to a socket 5 motherboard and Pentium CPU.
Most people have already had there 486 computer for a few years and have already max it out.

NEW EDO ram and and better motherboards for cheap.
better video cards , Onboards controllers, better and faster memory management

From the posts I have see. There is about a 15% performance improvement when going from a VLB motherboard to a PCI motherboard using the same CPU.
faster ram in EDO and better memory and video performance.

There is a noticeable performance difference when going from a 486 to a Pentium CPU.

If you benchmark a socket 3 Pentium Overdrive 83mhz against a Pentium 75mhz computer I bet the Pentium 75mhz would win. Just because of the better performing motherboard.

Reply 19 of 24, by Doornkaat

User metadata
Rank Oldbie
Rank
Oldbie
mkarcher wrote on 2020-08-21, 09:08:

As I heard it 20 years ago, the P75 was meant to target the laptop market only, where you wanted lower FSB speed to decrease idle current. The 5V Pentium 60/66 never were a good fit for laptops due to their high power consumption (compared to contemporary CPUs), and IIRC the Pentium 66 actually preferred to be run at slightly above 5 Volts. The desktop market was targeted with the Pentium 90 and Pentium 100. Later on, the Pentium 75 was sold into the desktop market because OEMs demanded an entry-level socket 5 option to get rid of their socket 3 and socket 4 systems in the line-up. The "real" P75 also likely was a great competitor to the Am5x86-P75 (133 MHz), and helped getting those sales to Intel instead of AMD.

Good info! Makes a lot of sense to me having the P75 be a mobile chip at first and then feeding it into the mainstream desktop market. It would also make a lot of sense to position the existing mobile P75 as a desktop chip against AMD's and Cyrix's 486 offerings in mid/late 1995, especially considering the socket 5 platform offered a viable upgrade path while the socket 3 likely didn't.