| View previous topic :: View next topic |
| Author |
Message |
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
 Posted: Tue Sep 11, 2018 7:24 am Post subject: What's the deal with K5 -and PR/Real Clock? |
 |
|
Hey there,
According to http://www.cpu-world.com/CPUs/K5/index.html
The K5 under PR120 had the actual frequency used as the PR rating also, from that point on, the PR rating ran through the roof 
I can't find anything, that suggests the core has been altered, except for the 5k86 or SSA/5 or whatever?
What are actually the core's names, and are there architectural differences, or did the K5-PR120 just mark the spot, where AMD figured that it would be smart from a marketing point of view?
If so - how did they handle the fact that a PR100 (@ 100MHz/66FSB) would actually outperform a PR120 (@ 90MHz/60FSB)
Please help me clarify  |
|
| Back to top |
|
 |
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
| Posted: Wed Sep 12, 2018 1:00 pm Post subject: |
|
|
https://en.wikipedia.org/wiki/AMD_K5 says:
"The "SSA/5" had its branch-prediction unit disabled and additional internal waitstates added"
So that's why there's a difference but no different number of transistors?
But here on CPU-World, it seems that some of the 75-100 MHz with PR rating as MHz, both came in SSA/5 and 5k86 |
|
| Back to top |
|
|
CPUShack
Joined: 16 Jun 2003
Posts: 34259
Location: State of Jefferson, USA
|
| Posted: Wed Sep 12, 2018 3:03 pm Post subject: |
|
|
Has nothing to do with an architectural change, it was purely marketing. They (and others, such as Cyrix) started out with using actual clock speed, they then found that in the right testing their CPUs would perform faster in benchmarks then an Intel of the same speed (integer usually)
This was used by the marketing dept for the PR speed
This is how you have a Cyrix 6x86-100 at 100MHz and then a 6x86-120 also at 100MHz
_________________
New for 2025! The CPU Shack has a co-processor!
Visit The CPU Shack of microprocessor history and information. |
|
| Back to top |
|
|
cuttingedgecs
Joined: 08 Oct 2017
Posts: 1764
Location: Australia
|
| Posted: Wed Sep 12, 2018 5:08 pm Post subject: |
|
|
| To answer your question about how it was handled that a PR120 @90 MHz was slower than a PR100 @100MHz, I would say it was partly through not having both models/ranges available at the same time (so they wouldn't appear on the one brochure or set of manufacturer benchmarks) , and partly handled through retailer behaviour. Back then, a 100MHz PR100 part would always have been marketed at retail as 100MHz. It wouldn't have been 'til there was a disparity between MHz and PR that the PR would be advertised. So if the two were advertised at the same time by different retailers, it would have been as a 100MHz vs a PR120 - putting both in the best light possible. |
|
| Back to top |
|
|
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
| Posted: Thu Sep 13, 2018 12:31 am Post subject: |
|
|
| CPUShack wrote: | Has nothing to do with an architectural change, it was purely marketing. They (and others, such as Cyrix) started out with using actual clock speed, they then found that in the right testing their CPUs would perform faster in benchmarks then an Intel of the same speed (integer usually)
This was used by the marketing dept for the PR speed
This is how you have a Cyrix 6x86-100 at 100MHz and then a 6x86-120 also at 100MHz |
What about the SSA/5 or 5k86 difference, then?
Unfortunately, not both are tested in the Ultimate 6x86 comparison, feipoa has made at vogons.org |
|
| Back to top |
|
|
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
| Posted: Thu Sep 13, 2018 12:32 am Post subject: |
|
|
| cuttingedgecs wrote: | | To answer your question about how it was handled that a PR120 @90 MHz was slower than a PR100 @100MHz, I would say it was partly through not having both models/ranges available at the same time (so they wouldn't appear on the one brochure or set of manufacturer benchmarks) , and partly handled through retailer behaviour. Back then, a 100MHz PR100 part would always have been marketed at retail as 100MHz. It wouldn't have been 'til there was a disparity between MHz and PR that the PR would be advertised. So if the two were advertised at the same time by different retailers, it would have been as a 100MHz vs a PR120 - putting both in the best light possible. |
Sounds plausible, and as the average buyer actually doesn't know the difference, it's gonna work .. |
|
| Back to top |
|
|
alvaro84
Joined: 17 Apr 2015
Posts: 80
Location: Fehérvárcsurgó, Hungary
|
| Posted: Thu Sep 13, 2018 1:22 pm Post subject: |
|
|
| H3nrik V! wrote: | What about the SSA/5 or 5k86 difference, then?
Unfortunately, not both are tested in the Ultimate 6x86 comparison, feipoa has made at vogons.org |
There is a definite difference between the two versions. I've tested both the PR100 and PR133 (both 100MHz) and though the PR133 isn't 1/3rd faster the difference is often at least 20-25 percent. Then sometimes it's much less. If I don't forget I'll fish out a few results. |
|
| Back to top |
|
|
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
| Posted: Fri Sep 14, 2018 12:15 am Post subject: |
|
|
| alvaro84 wrote: | | H3nrik V! wrote: | What about the SSA/5 or 5k86 difference, then?
Unfortunately, not both are tested in the Ultimate 6x86 comparison, feipoa has made at vogons.org |
There is a definite difference between the two versions. I've tested both the PR100 and PR133 (both 100MHz) and though the PR133 isn't 1/3rd faster the difference is often at least 20-25 percent. Then sometimes it's much less. If I don't forget I'll fish out a few results. |
Very interesting - I'll hope you remember  |
|
| Back to top |
|
|
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
| Posted: Fri Sep 14, 2018 3:10 am Post subject: |
|
|
| alvaro84 wrote: |
There is a definite difference between the two versions. I've tested both the PR100 and PR133 (both 100MHz) and though the PR133 isn't 1/3rd faster the difference is often at least 20-25 percent. Then sometimes it's much less. If I don't forget I'll fish out a few results |
Probably FPU intensive tests, that don't differ so much. The FPU is probably such a big bottleneck, that the enhancements are negligible .. |
|
| Back to top |
|
|
alvaro84
Joined: 17 Apr 2015
Posts: 80
Location: Fehérvárcsurgó, Hungary
|
| Posted: Thu Sep 20, 2018 12:19 pm Post subject: |
|
|
| H3nrik V! wrote: |
Very interesting - I'll hope you remember |
I do, I just was a little bit busy...
So, here's my little comparison. It's no comprehensive test, just a few haphazard results from games and demos I often test with, all running under DOS 7.1.
AI Bench [basically a rotating textured torus in true color]: 894 vs 871 points, a measly 2.6 percent lead for the PR133.
Doom: 73.2 vs 86.0 fps - a more healthy 17% difference.
Duke Nukem 3D (320x200): 64 vs 76 fps. Almost 19%.
Duke Nukem 3D (640x480): 26 vs 29 fps, mere 11.5%, probably heavier load from frame buffer copies.
Quake 320x200, nosound: 19.8 vs 24.6 fps, 24.2% difference in probably the most FPU intensive benchmark.
The board was a FIC PT-2003 (i430 FX chipset, 512kB PB COAST cache), the VGA an Asus V3000 PCI (Riva 128). |
|
| Back to top |
|
|
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
| Posted: Thu Sep 20, 2018 1:59 pm Post subject: |
|
|
| Wow, thanks for sharing results. There's definately some things, that has been optimized. Interesting indeed! |
|
| Back to top |
|
|
CPUShack
Joined: 16 Jun 2003
Posts: 34259
Location: State of Jefferson, USA
|
| Posted: Sat Sep 22, 2018 12:29 am Post subject: |
|
|
In more PR fun, PR ratings were not always higher then the clock speed, sometimes they were actually LOWER, for example a Cyrix 5x86-120 has a PR90 rating, a 6x86 120MHz is PR150, again the baseline was a Pentium 120, which was faster then the 5x86 but slower (in the right tests) then a 6x86 PR150
_________________
New for 2025! The CPU Shack has a co-processor!
Visit The CPU Shack of microprocessor history and information. |
|
| Back to top |
|
|
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
| Posted: Sun Sep 23, 2018 3:47 am Post subject: |
|
|
| CPUShack wrote: | | In more PR fun, PR ratings were not always higher then the clock speed, sometimes they were actually LOWER, for example a Cyrix 5x86-120 has a PR90 rating, a 6x86 120MHz is PR150, again the baseline was a Pentium 120, which was faster then the 5x86 but slower (in the right tests) then a 6x86 PR150 |
And AMD DX5-133 had a P75 rating  |
|
| Back to top |
|
|
CPUShack
Joined: 16 Jun 2003
Posts: 34259
Location: State of Jefferson, USA
|
| Posted: Sun Sep 23, 2018 11:56 am Post subject: |
|
|
| H3nrik V! wrote: | | CPUShack wrote: | | In more PR fun, PR ratings were not always higher then the clock speed, sometimes they were actually LOWER, for example a Cyrix 5x86-120 has a PR90 rating, a 6x86 120MHz is PR150, again the baseline was a Pentium 120, which was faster then the 5x86 but slower (in the right tests) then a 6x86 PR150 |
And AMD DX5-133 had a P75 rating |
Exactly!
_________________
New for 2025! The CPU Shack has a co-processor!
Visit The CPU Shack of microprocessor history and information. |
|
| Back to top |
|
|
|
