A Tragic Flaw

The quality of peripherals that go into the Kryotech system begins to matter much more than in a normal system, especially when dealing with memory.   This could be because of increased electrical noise in the system while running at such an increased frequency, or because of the cooling methods applied to the system, nevertheless the fact does remain that not all peripherals will work flawlessly in the system.  AnandTech's tests with the Cool K6-2 revealed that that Matrox G200 came closest to being one of those incompatible peripherals, normal usage of the card wasn't hindered at all, however there was the rare occasion that the card would be the cause of spontaneous crashes with the system which would not occur in an identical test bed setup without the Kryotech unit.  Most other peripherals experienced no problems at all, which is a definite plus, however the Kryotech system is extremely sensitive to the quality of memory used.  From the tests AnandTech conducted, Samsung and NEC SDRAM chips and modules seem to work fine but according to Kryotech, Advantage Memory is the recommended option in a stability and compatibility sense.

Under Windows NT there exists a tragic flaw with the Kryotech Cool K6-2's current technology, the thermal jacket which prevents condensation from forming between the pins of the processor and the motherboard's CPU socket also seems to introduce a problem when running under Windows NT, primarily when using SCSI based disk drives. 


The CPU (#2) plugs into the thermal jacket (#3) which then plugs into #1 which plugs into the motherboard.

The problem is not limited to SCSI drives only, as it seems to happen under Windows NT with IDE drives as well (just not as severe), and causes random crashes and lockups with the system.  Kryotech is aware of this problem, and they are currently working on a mainstream solution for it.  Kryotech has told AnandTech that they have a system running under NT with SCSI devices in-house, however a true solution won't be available for at least another 6 weeks.  They are working on a technology that would lower the cost of the system by up to 20%, and would effectively remove that thermal jacket from the system itself by allowing the CPU to plug into the motherboard itself.   The earliest AnandTech was told that such a technology would be available was February, so until then, it appears as if the Cool K6-2 isn't the hottest option for high end Windows NT users.

Too little too late?

At a price of $1695 for the motherboard, CPU, case, and cooling system, the Cool K6-2 is a bit steep considering the performance it offers over a readily available and comparatively cheap K6-2 400 mb/cpu/case combo.  With many users already able to overclock their K6-2 400's to 450MHz without using Kryotech's cooling system, the extra 5% performance boost isn't really worth the money; and in that respect, the Kryotech Cool K6-2 500 is in fact, too little, too late.  Had the system been released at 500MHz back in May of 1998 when it was announced at E3, things might have been different, however with the price of a Celeron 300A at under $100, and with the price of a K6-2 400 quickly approaching the $200 mark, the Cool K6-2 just isn't worth it now.

Is the system a failure?  Absolutely not.   Kryotech's Cool K6-2 is an example of what can be done, and as a glimpse at the future of Kryotech, AnandTech decided to tear out the K6-2 400 CPU that was installed in the system and replace it with a K6-3 to see how well AMD's forthcoming chip would operate at -40 degrees Celsius.  

Modifying the System

The sample K6-3 AnandTech tested exhibited overclocking results similar to that of the K6-2, in that the physical peak of the processor appeared to be 500MHz, with 550MHz not booting at all.  A final judgement on the processor's overclocking abilities with the Kryotech system can be made closer to the release date of the chip since the sample tested was an engineering sample, however the sheer thought of having a K6-3 running at 550MHz and beyond brought quite a smile to this reviewer's face.   The performance difference between a K6-3 running at 450MHz and one running at 500MHz was approximately 6% in the Winstone tests that would run, however the entire suite would not complete as a result of one or more factors, most likely due to the nature of processor with it being an engineering sample. 

Performance Conclusion
Comments Locked

0 Comments

View All Comments

Log in

Don't have an account? Sign up now