Wesley's Tool-Box

• It's a Cyrix, Only that IT ISN'T

Quite a few things have been changed to Cyrix since last year and some of them are obvious on the picture above. You can click on the picture for bigger image. First of all, since VIA owns Cyrix now, the company that produces the chip is VIA, and Cyrix is now just the brand of the chip itself. And why is it 'III' when where weren't a 'Cyrix I' or 'Cyrix II' before? It's possibly because the last Cyrix chip was M-II, which is now marketed as 'VIA Cyrix MII'. Also noticeable is the disappearance of the 'PR' speed rating system which followed with Cyrix chips and for a short time, AMD chips. Now Cyrix is 'proudly' showing its true MHz speed. Let's have a look at the chip's features.

 

Interface:	Intel's Socket 370
FSB:	133MHz or 100MHz
L1 Cache:	128KB
Inst.Sets:	MMX, 3DNow!
Power:	10W Average
Mfg.Process:	0.18 micron
Availablility:	500, 533, 600MHz

As you can see, Cyrix III is to be plugged into Socket 370, so it's trying to compete directly with Intel's Celeron processor. This move from aging Socket7(or Super7) to widespread Socket 370 means that Cyrix III may please many users as an upgrade option. Support for MMX and 3DNow! would provide decent boost in speed where it's supported, but unfortunately, lack of SSE support means it may fall short of newer Celerons with Coppermine core ('Celemine') which support it. Supporting 100MHz or 133MHz FSB is impressive, nevertheless, because Celerons only support 66MHz. The chips reviewed (533MHz, 600MHz) support 133MHz FSB, while the remaining one (500MHz) supports 100MHz FSB.

But do you spot something missing here? If you've seen the features of the chips that came out since 1999, you should know. This processor does NOT have integrated L2 cache. This is a bigger problem than it was back in Socket7 days when chips with integrated L2 cache was generally non-existent (excluding hard-to-produce AMD K6-III); on a Socket7 motherboard, you can find around 256KB to 2MB of L2 cache installed, but on a Socket 370 motherboard, there is no L2 cache installed because it assumes that the installed CPU already has one. With today's CPUs far outrunning memory in terms of speed, having L2 cache to buffer the memory is crucial in boosting performance. The lack of it in Cyrix III is a rather big mistake.

• Something's Missing Here...

Actually, the original chip that was to become Cyrix III, codenamed Joshua and in development at Cyrix at the time of VIA's acquisition, DID have integrated L2 cache. This chip still used the PR speed rating system, but if you drop the PR and put it head-to-head with Celeron, it gave Celeron run for its money, MHz for MHz. What happened now? VIA decided to throw away the original Cyrix Joshua core with Centaur(as I've noted last page, also acquired by VIA) WinChip-based core. WinChip's design lacks L2 cache, is known to have unspectacular performance, but does consume less power, as it is evident on the feature list.

The reason behind this 'stupid' decision is that VIA succumbed to the industry-wide belief that 'the MHz sells' - no matter what the actual performance of the chip is, customers regard MHz as the sole rule of speed and buys whatever that has faster MHz. This is sadly true, even though many power users know this is basically wrong. Centaur design yields higher MHz than Cyrix design (you see a 600MHz version for this review, but the original Cyrix design hardly went beyond 400MHz) even though the actual performance suffers. Because MHz yield is better, VIA was able to take off the PR system from the chip.

To wrap up, you can see that what I have here for the review is not really a 'Cyrix' - manufactured by VIA, based on Centaur design, and plugs into an Intel platform. The only thing 'Cyrix' about this chip is the name. Okay. Now that I've bored you enough, I'll move straight to the benchmark!

• Where has Cyrix been?

The CPU industry for IBM compatible (or non-Mac) personal computers was pioneered by, and still dominated by Intel, the creator of modern microprocessors. Many companies have tried to take a bite out of this ever-growing market with varying success. The most prominent competitors were Advanced Micro Devices (AMD) and Cyrix. These two companies gained notable market shares by attacking value-oriented low-end segment of the industry in the last few years.

Even though Intel tried to cut off the competition by suddenly moving to Slot1 design for processors in 1997, AMD and Cyrix improved the Socket7 design thrown away by Intel to come up with 'Super7', which became a viable and cost-effective alternative to Slot1 and the two companies remained competitive. Seeing their little success in the low-end, many companies like Centaur and Rise sprang up to provide their own version of the Super7 processors.

Unfortunately, though, in 1999, nearly all of the competitors to Intel faced collapse due to many factors such as aging of the Super7 platform and problems with ramping up speed or production target. Cyrix and Centaur was eventually bought by Taiwan's VIA Technologies and Rise switched their target market to internet appliances, disappearing from the PC scene. AMD was on the verge of bankruptcy while their K6-III had trouble producing in volume. That's how Cyrix, Centaur, and Rise virtually disappeared. But...

In case of AMD, however, their unusually high devotion to R&D department in the last couple of years had paid off, and their K7 processor, renamed and now known as Athlon, was able to meet or surpass Intel's high-end offering, the Pentium III on both performance and price-competitiveness. Because of this chip and its siblings that followed such as Duron and 'Thunderbird', AMD made a strong comeback and is now viewed as a significant competitor to Intel. This trend is still picking up the pace - the latest additions to this include Intel's 2000 Q3 earnings warning in which many are thinking that AMD's gain in the market share played a significant role.

So what we are now seeing is that the PC industry is enjoying a true 'duopoly' of Intel and AMD, the two companies' combined market share being more than 99%. The two companies have almost the same line-up of ultra high-end to value low-end processors already in mass production or going into production in a few months, thus overlapping each other in every aspect and having fierce competition. All the while the other PC processor companies remained in virtual silence....

Until now.

• Finally - Cyrix Makes Comeback

 

What you see above is a VIA Cyrix III 600MHz processor in a plastic box that goes inside the retail package. The hollow area on the right of the box is where the heatsink/fan goes in. I was shocked to see this handed to me for a review because I've not heard of Cyrix much since it was absorbed by VIA more than a year ago. I did hear about this chip occasionally, but it was constantly being delayed that I thought I would almost forget about it. This product is going to be available on the market almost immediately, so I can pretty much say that Cyrix has finally made a comeback. It is quite evident that, because of all the spotlight being focused at Intel and AMD, Cyrix's re-entry was rather quiet.

• Unlocking... Again

Yup! These chips are obviously locked. It's the time for the pencil unlocking again. I've posted a how-to page on this some time ago, so the people who are unfamiliar with what I'm saying can go there. I have, however, attempted to take a better picture on how the chip looks before and after the unlocking is done, so you can click on the above images for a bigger one and pay attention to the orange arrow.

Unlocking finished, I have attempted overclocking at various speeds.

CPU:	AMD Duron 650MHz (Both)
M/B:	Asus A7V (KT-133 Chipset)
Cooling:	Alpha PAL6035MUC Fan/Heatsink
RAM:	128MB x 2 Hyundai PC-133 SDRAM
Video:	SUMA GeForce2 GTS 32MB
OS:	Windows 2000 SP1
Else:	Classified ^_^

 

Speed	Config	Voltage	Green	Blue
650MHz	100 x 6.5	1.50V	Perfect	Perfect
700MHz	100 x 7.0	1.50V	Perfect	Perfect
735MHz	113 x 6.5	1.50V	Good	Perfect
750MHz	100 x 7.5	1.50V	Boot Stop	Good
800MHz	100 x 8.0	1.50V	POST Stop	Boot Stop
800MHz	100 x 8.0	1.70V	Perfect	Perfect
850MHz	100 x 8.5	1.80V	Perfect	Perfect
892MHz	105 x 8.5	1.80V	Boot Stop	Boot Stop
892MHz	105 x 8.5	1.85V	Perfect	Perfect
900MHz	100 x 9.0	1.85V	Good	Perfect
910MHz	107 x 8.5	1.85V	Good	Perfect
927MHz	103 x 9.0	1.85V	Boot Stop	Perfect
945MHz	105 x 9.0	1.85V	POST Stop	Good
950MHz	100 x 9.5	1.85V	POST Stop	Good
963MHz	107 x 9.0	1.85V	No POST	Boot Stop
1000MHz	100 x 10.0	1.85V	No POST	POST Stop

Perfect = No Crashes Whatsoever
Good = Benchmark Crashes and Freezes Occasionally on Normal Use
Boot Stop = Computer Crashes During Boot or Immediately After Boot
POST Stop = Computer Crashes Before POST finishes
No POST = Cannot Boot At All

• Blue IS Better

As the table shows, while the GREEN one hit the ceiling at 892MHz, the BLUE one had no problems up to 927MHz. The latter even showed me 1000MHz on the screen... I think I may try a voltage boost some time to get that 1GHz mark. For now, though, I think I'll stay at this point. BLUE chip overclocking 5.4% higher than the GREEN chip at the same condition is not too shabby.

You should note that the GREEN chip showed almost an identical overclockability as the original green Duron shown at the pencil unlocking article, if you click here. In fact, I almost thought the new green Duron was my old dead one.

But this isn't the end of the story - the two green Durons that went through my hands were all made in WEEK 25, while reports of the blue Durons reveal that they are either of WEEK 27, like mine, or WEEK 22. From this, I think the logical explanations for all this could be either...

1. The blue Durons are somehow made in Dresden, defying the current knowledge. The copper interconnects contributed to the better overclockability of the blue type compared to the green type.

2. The blue Durons are from Austin, but on week 27, they had to use a different polish. The better overclockability may be from the better bin-splits that week had.

Some other speculations are possible, of course. What's your take on the issue? How did your 'blue' Durons fare compared to the 'green' Durons? Let me hear it at my new message board! Click below!

P.S. No, I did not post a benchmark comparing the two.... I thought it's pointless to do it... but if many of you want it, maybe I'll post it.

• I had a Crush(?) on Duron

Durons are fragile.. Durons are fragile.. uh..? To confess, I was not being 'nice' to my Duron. Guess what? It died a painful death... yes... I knew those weird sounds were not good..... The Duron you saw in the last article is no more, sadly.

Why am I saying that there? Because that 'incident' had apparently left me with two Durons for a while.

Confused? You see, because that Duron 'failed', I had sent it back for a replacement. Wonderful replacement policy at that shop I bought the chip, I tell you. While I was waiting for the replacement to arrive, I had to run the machine anyways, so I got another Duron. I was back on track. And after a few days, the replacement finally came. I pulled out the Duron from the system...

• Duron on Blue

The surface of the two Durons actually differed! The pictures above were not taken at the same time of the day (sorry about that) but you can clearly see the blue tint on the edges of the core on the right chip while the left chip is generally greenish all around. The 'blue' one was the replacement chip, by the way.

The bluish-greenish tint difference is well-known in the Thunderbird chips. The bluish one is the one from Dresden, and thus using the newer 'copper interconnect' process, while the greenish one comes from Austin, using the normal 'aluminium interconnect' process. The process itself does not account for the colour; rather, the two fabrications use the different polishes and the resulting tint gets different.

Trouble is, Durons are known to be manufactured only at Austin, so normally, Durons are 'green'. But with this 'blue' Duron now at my hand, one would have to wonder what's going on. There's a lot of debate on whether it's really coming from Dresden and using copper interconnects or not, but I just decided to put the two to the test.

What test? Overclocking, of course! Since copper interconnects are known to have better yields of high speed grades, it is possible that the 'blue' one would overclock better than the 'green' one.