Hey BK

[Joe O]

<< The Cyrix part is designed to be more thermally efficient than Celeron >>


Is this the end of an era?
From http://www.theregister.co.uk/content/1/12055.html

 

[BurntKooshie]

True. It will have a smaller die size than the Celeron, and use less transistors (the celeron 2 is a P3E with 128K disabled). However, you will note that this makes Cyrix's even worse Whereas Cyrix previously used the PR rating, and the Joshua version of the Cyrix III came at least close to it in many integer apps (multimedia and FPU bit big time). The new Cyrix III (sameul) is worse Even when they dissabled a celerons L2 cache, so that the two chips would be competing with L1 caches and memory only, the Celeron STILL EASILY beat out the Cyrix III (samuel). So the addition of a L2 cache shouldn't change things much, as, with a large L1 cache, the L2 cache at that size is less important to the CPU (especially one that isn't so bandwidth starved as the celeron is), so even with the L2, is SHOULDN'T be able to compete performance wise with a Celeron (though only time will trully be able to tell)

For example, a 700mhz celeron dissipates, at maximum, on a .18 micron process, 18.3 watts.

The Joshua (cyrix III), on POS .18 micron process (national semiconductors .18 process is said to be as good as intel/amd's .25 micron process), at 450mhz, dissipates 24.6 watts.

BUT, then you have to take into consideration the fact that the NEW Cyrix III (samuel) is a 135 degree twist from (yes, 135 degree) the Cyrix teams previous philosophies, as it uses the Centaur teams philosophies. Cyrix thought PR rating would help them sell. Centaur thinks that a high mhz number will sell, because they can be honest about the mhz, as at least in that one area, they will be slightly competative. The not-quite-180 degree role reversal is because one went for a lot of high IPC (instructions per clock), while the other went for high CPS (clocks per second). The reason its not 180 degrees is because they both have FPU's that SUCK....and imagine this....the Samuel FPU sucks worse than the Joshual FPU....*shudders*.


Sure the thing gets up to some mhz that it never got up to before, but then we are back to the 586 days, where AMD and Cyrix had 133 and 120mhz (the AMD chip was a 486 with a larger chache running at high frequencies, the Cyrix chip was a hybrid 4/5 generation chip, hence the disparity between the mhz and the PR rating) 586 chips respectively, that had PR ratings of a pentium 75. They never advertised the clock speed so much as the PR rating.

The new Cyrix III, even though it is using the Mhz instead of PR rating to advertise, is WORSE OFF than the Joshua core (in terms of performance). It was shown that a >600mhz Samuel wasn't competative with anything current - even a K6-2 450mhz chip beat it out. So if this thing had a PR rating, it would be LOWER than its mhz rating, just as the 586 days. The only way that the Cyrix III (samuel) will sell well is for it to be VERY cheap, in which case, it'll be used in the places its best suited....embeded and wordprocessing/browsing environments. The advantage that it has though will be its small die size.

The Samuel (Cyrix III) has a very small die size (i believe 77mm compared to a celerons current 90mm - and the next version of the Samuel/Cyrix III with 64k L2 exlusive cache should be smaller, as it will be on a smaller process, though the addition of L2 might keep it around the same), AND, it has been shown that its FPU marks are SO low, that it has been deduced that it has a OLDER winchip (the samuel is really a slight variation on the Winchip4) running at HALF the clock speed. The saving grace will be a LOW price, and LOW heat dissipation. Lacking one, or both, the chip, as it stands now, cannot survive.

By running at half the clock speed, that should also reduce (somewhat) the amount of heat dissipated, though I don't think it helps THAT much, as they OBVIOUSLY didn't spend that much time/transistors into the FPU, and hence, it should dissipate much heat anyway.

So, on Nat semi's .15 micron process, it should be competative (process technology wise) with intels .18 micron process, and, as it has a smaller die size, and an FPU that is rather horrid, i would HOPE that it would dissipate less heat

However, if you want to get into an efficiency rating, as in terms of performance per watt dissipated, I have a VERY strong feeling the Celeron would win It would probably still win on a Bang/Buck perspective.

[EDIT]Holy crap. I didn't mean to write that much....
[EDIT]Ack! It keeps getting longer! Maybe I should turn this into an article

[EDIT] Now to answer your question:Q I think this is the end of ANOTHER era in Cyrix history. Once, they actually made FAST FPU's (during the 386's, they were actually good math co processors). Then, they made some fairly poor 486's (they were like 386.5's). Their 686 generation, at least at first, was good. That was the golden age for Cyrix. Then National Semiconductor bought them out. And in the x86 CPU business, they didn't do ANYTHING (not even higher speed grades except at the VERY end just before Via bought them), though they did design the GEODE (nat semi still has the group that designed that part - which is a PCOAC [PC on a chip] - but from a 586 core - most notably, an extention of the mediaGX). This new era of Cyrix, has little to do with Cyrix. All the good engineers have left. Many were fired. High performance projects that were in the works were canned (Jalepeno'). For the most part, the Cyrix era is GONE, as there are few remnants of Cyrix left, and what makes up the CPU design team is mostly Centaur now.

So no, its not the end of an era. Its the end of a company.

 

[Joe O]

B.L., Maybe you should turn it into an article. It was very well written.






But you didn't tell me if I could use it as a space heater this winter!

 

[ViRGE]

Nooooooo!!!!!!! Why oh why do the good joke ideas die so young?!?!

 

[BurntKooshie]

Space heater eh.....well, if you're in dire straights that bad....I have a PR 200 i could let you have for the winter