A three hour drive south of AnandTech’s HQ in Raleigh, North Carolina, you will find another company that has themselves embedded in the computer hardware world - Kryotech.  We received our first formal introduction to Kryotech at the 1998 E3 Expo which, at the time, was held in Atlanta, Georgia.  Armed with AMD’s recently released K6-2 400 CPU, Kryotech managed to offer to the public a K6-2 that operated at 500MHz using their patented super cooling technology that cooled the chip down to –40 C. 

While this cooling technology was definitely a very impressive accomplishment, the fact of the matter was that cooling a K6-2 400 down to –40 C just to run it at 500MHz was not worth the added cost of the Kryotech system.  Although Intel’s Pentium II offerings at the time were more expensive than an equivalently clocked K6-2, the Kryotech Cool K6-2 500 retailed for $1695, a price which included the motherboard, CPU, case, and cooling system, which was greater than the cost of acquiring the same components for use with a Pentium II that would offer the same performance as, if not greater than, the Cool K6-2. 

This was Kryotech’s unfortunate downfall.  While they would not stray from their close relationship with AMD, it was in fact killing them because of the simple fact that AMD’s K6-2 and K6-III line were not the type of CPUs that would benefit the most from this sort of technology. 

Many asked why Kryotech simply didn’t offer an Intel version of their Cool K6-X line of systems, but what most users didn’t understand was that establishing a relationship with Intel similar to the one Kryotech had with AMD is even more difficult of a feat than developing the Kryotech cooling technology.  In spite of this, Kryotech’s cooling products have actually been used in quite a few Intel demonstrations, although Intel would never disclose it. 

It just seemed like Kryotech could never offer a system that was actually worth the added cost that the consumers were being forced to pay.  At the same time, the idea of using the Kryotech systems for professional applications was countered by the argument that the poor performing FPU of the K6-2 and K6-III was not something that could be masked by a slightly higher clock speed.  For engineers and high-end workstation users, the Kryotech Cool K6-X systems were neat things to have but not systems that were of much value. 

 

The Turning Point

In the past year AMD has gone from being the underdog to being the performance leader and then back to being a direct competitor to Intel on a clock for clock basis.  The Athlon has placed AMD in a light that they are not used to being in.  With the release of the Athlon, AMD can no longer be known as the low-cost solution; they have become the high-performing competitor that has forced Intel to greatly accelerate their processor release roadmap.  At the same time, the Athlon provided Kryotech with enough ammunition to build a killer system that would finally put the Kryotech technology to good use.  Not only was this a turning point in AMD’s history but also just the break Kryotech needed to begin releasing truly high-performing super cooled solutions that were worth the added cost.

Kryotech’s first Athlon solution was the Cool Athlon 800 which was released back when the Athlon 600 was the fastest thing out.  Because of the intense competition between Intel and AMD, an air-cooled Athlon 800 was released just a couple of months later thus making the Cool Athlon 800 worthless.  Anticipating an accelerated 800MHz Athlon release, Kryotech took the liberty of releasing their Cool Athlon 900, which is still faster than the current clock speed king, the Athlon 800. 

The Cool Athlon 900 was actually a bigger step than it seems like for Kryotech.  While the slower 500 and 550MHz parts are having tremendous success overclocking to 700 – 800MHz frequencies, the faster Athlon parts aren’t having nearly as much success.  The 700MHz Athlon is having difficulty making it higher than 800MHz and the newer 0.18-micron cores are seemingly stuck at around the 800MHz mark.  We have been able to get an Athlon 800 (0.18-micron) to run at 900MHz with a core voltage of 1.80v using a 78W Peltier but the operation was not nearly as stable as we would like it to be.  This will change as the yields on the 0.18-micron parts improves, but, for now, don’t expect these 0.18-micron cores to go much further.  So pushing 900MHz definitely shows off exactly how sophisticated and powerful Kryotech’s cooling system is. 

The next step in this progression was the release of the true monster, a 0.18-micron Athlon 750 running at 1000MHz, dubbed the SuperG.  The reason for the name change is because the SuperG actually uses a more powerful compressor and a slightly altered design from the original Cool K6-X and Cool Athlon series of systems to allow for an Athlon 750 to be pushed to 1000MHz (and no, using an Athlon 800 doesn’t allow for much higher clock speeds to be achieved). 

How it works

It has been quite a while since we have talked about the technology behind Kryotech’s cooling systems, but now is as good a time as any to provide a little refresher.

As we mentioned before, we were first introduced to Kryotech’s cooling technology at the 1998 E3 Expo in Atlanta.  The technology used in the Cool K6-2s back then and the SuperG today is identical and is known as Vapor-Phase Refrigeration. 

Vapor-Phase Refrigeration is the same technology that makes the refrigerator in your kitchen keep your food cold.  Vapor-Phase Refrigeration uses the physical properties of a refrigerant, like the one mentioned above, to achieve an effective level of cooling.  The refrigerant is stored in a liquid form and heated to the point where a phase change is initiated, converting the liquid to a gas which cycles through the cooling system and is returned to a compressor which initiates another phase change and compresses the gas to a near-liquid form. This enables the heat to be removed from the source, and the cooling system itself to remain self sufficient.

Kryotech claims that Vapor Phase Refrigeration is 50 times as effective as traditional forced air cooling (your standard heatsink/fan combo cooling device) and 5 times as effective as forced liquid cooling. Using Vapor Phase Refrigeration, Kryotech has been able to allow their products to run at levels around -40 degrees Celsius, including the SuperG. 

The Kryotech SuperG uses the same Vapor Phase Refrigeration technology used in the older Cool K6-X and Cool Athlon series.  The CPU is placed and isolated in a chamber known as the KryoCavity.  The KryoCavity features a nickel plate that makes contact with the surface of the CPU itself as well as the L2 cache.  The heat is extracted from the plate and transferred through the thermal bus, essentially a fairly thick and barely malleable black tube.  This process leaves the surface of the plate at approximately –40 C and thus allows for the “thermal acceleration” of the Athlon CPU to higher frequencies (basically, Kryotech is safely overclocking the Athlon; the ‘safely’ part comes from the fact that, by cooling the CPU down to such a low temperature, you immediately increase the range of clock speeds it can operate at). 

Removing the heat that the processor introduces into the case, through the use of the KryoCavity, provides for a more stable operating environment since you don't have to worry about the presence of an extremely hot processor contributing to the growing heat problem in today's desktop computers. The only producers of heat in the case that remain are devices such as video cards and hard drives which should operate much more comfortably with the heat of the processor removed from their surroundings. 

The heart of the SuperG is AMD’s latest 0.18-micron 750/800MHz Athlon core that is “thermally accelerated” to 1000MHz, using a 10.0x multiplier and the default 100MHz FSB.  The L2 cache of this CPU, like the 750/800MHz air-cooled Athlons is set to 2/5 the core clock and, with the core operating at 1000MHz, the L2 cache runs at 400MHz, the highest for any available Athlon.   Since, currently, the Athlon doesn’t have a full speed on-die L2 cache, the performance advantage that it holds over the new Pentium IIIs based on the Coppermine core is reduced in certain situations.

With the Super G, Kryotech is a bit more flexible with the motherboard configurations.  They allow for either a Gigabyte GA-7IX or a Microstar 6167 to be used in the system since the two layouts are virtually identical.  The sample we tested was outfitted with the Gigabyte GA-7IX. 

Click to Enlarge

To answer the next question without being asked first: no, the ASUS K7M cannot be used with the SuperG.  The K7M has a row of capacitors placed flush against the Slot-A connector which would prevent the KryoCavity from being installed properly. 

Like the previous Kryotech systems that we have taken a look at, the SuperG cools the nickel plate down to –36 C before powering up the system itself.  If at any time the temperature rises too high, the base of the system immediately signals for power to be cut from the system.  The base can control when the system turns on and shuts off is because the power switch is actually connected to a set of wires that run to the base of the system.  Another set of wires is connected to the motherboard which control powering the system on and turning it off.

 

Click to Enlarge

Our test system was outfitted with a Sparkle FSP300-60GT which is, naturally, on AMD’s Recommended Power Supply list.  However, the combination of the FSP300-60GT and the 7IX resulted in a very serious problem with delivering enough current to the AGP slot. 

This combination caused the system to lock as we ran 3D applications/games under Windows 98 when certain cards that draw quite a bit of current were used.  The system seemed to work fine under NT, where not all of the AGP features are enabled as a result of NT’s limited support for AGP as anything other than an interface slot (AGP texturing/SBA are not enabled and thus the current requirements are lowered). 

The problem resides either with the motherboard or the power supply; we’re currently investigating the problem even further to see the exact cause of the issue.  This problem will affect those users purchasing the SuperG for gaming purposes, but it won’t trouble those using it for professional use under Windows NT since NT doesn’t support any AGP features other than the faster clock speed of the AGP bus. 

Hitting 1GHz

In order to push the 750 to 1GHz (pronounced Gigahertz, not Gigahert), Kryotech had to make quite a few modifications to their system.  The first was to increase the core voltage of the Athlon from 1.60v to 1.85v.  This trick is quite common to overclockers: increasing the core voltage often helps achieve stability at overclocked speeds, and this was a necessary move for Kryotech in order to get the 750 to 1GHz.  Don’t worry though - 1.85v is within the range of supported voltages for the Athlon, although it is a bit on the high side. 

The SuperG features a larger compressor capable of dissipating 130% more heat than the original design, thus paving the way for a completely new line of Kryotech products that can scale to higher clock speeds and are flexible enough to accommodate other options.  While the last part of that statement is a tad vague, we have been lead to believe that the SuperG will be much more upgrade friendly than the previous models, possibly capable of accepting future Athlon CPUs and motherboard configurations with a simple switch of the CPU or a few modifications to the existing design.

The new compressor is actually quieter than the older ones and, from our experience with it, it’s even quieter than Swiftech’s MC1000 Peltier cooler. 

Other Modifications

As we mentioned above, the SuperG is more suited for future expansion and upgrades.  While a part of the reason is because of the more powerful compressor, Kryotech also made use of a roomier case with the SuperG. 

SuperG behind the Cool K6-X

Click to Enlarge

The case itself is manufactured by Superpower, and it is a part of their Polaris II series, more specifically, the KS-298.  We have reviewed a similar case made by Superpower, the KS-299.  Although it’s a part of the Landmark series from Superpower, we noticed a very similar construction in the KS-299.  The main benefit of both cases is the ease of getting into and out of the case.  This not only allows for ample room for the KryoCavity, but it also makes installing larger motherboards an option for the future. 

Click to Enlarge

The KS-298 is capable of accepting most dual processor motherboards, and, with a dual CPU Athlon chipset due out sometime in the second half of next year, we would hope that Kryotech would allow for an upgrade to the setup to accommodate a dual CPU setup.  Whether or not this wish will come true is difficult to predict, but if you’re spending around $2500 on a case, motherboard, and CPU, you would hope that Kryotech would support the upgrades that you want in the future. 

An unfortunate problem with the KS-298 case is cooling the rest of the components.  In an effort to move air through the case itself, Kryotech has included a fairly large 92 mm intake fan at the rear of the case.  While it is a good idea to include a large fan in a case where there is no movement of air (remember, the CPU doesn’t have a fan, it’s closed off from the rest of the case), the decision to make this fan an intake fan wasn’t too great of an idea. 

Flipping the fan around so that it acts as an exhaust fan thus pulling out the hot air from the case (generated by the chipset, hard drive, video card, etc…) would make for a much more effective cooling decision as it creates an area of low pressure inside the case.  It’s not a big deal, but we would have expected much more from Kryotech. 

The Test

 

BAPCo's SYSMark 2000 features up-to-date versions of CorelDRAW 9, Microsoft Excel 2000, Dragon Systems NaturallySpeaking Preferred 4.0, Netscape Communicator 4.61, Corel Paradox 9, Microsoft PowerPoint 2000, Microsoft Word 2000, MetaCreations Bryce 4, Avid Elastic Reality 3.1, Adobe Photoshop 5.5, Adobe Premiere 5.1, and Microsoft Windows Media Encoder 4.0. The SSE SIMD FP optimizations in these applications coupled with the Pentium III's full speed on-die L2 cache help keep the Pentium III at 800MHz virtually on the heels of the SuperG Athlon at 1GHz.

The SuperG is the fastest, although not by a huge margin, in this test. It is noticeably faster than the previous Athlon king, the 800, but it should be noted that the Athlon in its current form (with the K75 core) will most likely never hit an air-cooled 1GHz clock speed. Instead, the first 1GHz Athlon will probably based on a core with a full-speed on-die L2 such as the Thunderbird or Spitfire.

The SuperG enjoys a similar lead over the competition in the Content Creation Winstone 2000 test, although it is followed up by the Athlon 800 rather than the Pentium III 800.

Once again, although the SuperG is the current fastest processor in this test this performance is not indicative of the air-cooled 1GHz Athlon that will ship next year. The air-cooled 1GHz Athlon should be noticeably faster due to its faster L2 cache speed and other possible optimizations.

We were able to get Quake III Arena to run on the SuperG test system although the system would crash intermittently as a result of not being able to deliver enough current to the AGP slot. Running these video cards on the SuperG is an excellent way to see how they will perform with faster CPUs as they are released.

The 640 x 480 test in particular stresses the CPU more than the higher resolution tests and thus gives us a better picture of CPU performance.

Unreal Tournament would unfortunately crash before allowing us to run any tests, thus removing the GeForce and DDR GeForce as potential candidates in this comparison. This problem, once again, seems to be the result of not being able to deliver enough current to the AGP slot. The cause of this could be because of the GA-7IX motherboard, the Sparkle FSP300-60GT, or a combination of both. While both the power supply and motherboard work just fine in normal conditions, with the Athlon running at 1000MHz and running at 1.85v, the strain on both the motherboard and the power supply is tremendous.

SPECviewperf

The Standard Performance Evaluation Corporation, commonly known as SPEC, managed to come up with a synthetic benchmark with real world implications. By running specific "viewsets" SPECviewperf can simulate performance under various applications. To be more accurate, according to SPEC, "A viewset is a group of individual runs of SPECviewperf that attempt to characterize the graphics rendering portion of an ISV's application." While this method is by no means capable of identifying the performance of a card in all situations, it does help to indicate the strengths and weaknesses of a particular setup.

SPECviewperf 6.1.1 currently features five viewsets: the Advanced Visualizer, the DesignReview, the Data Explorer, the Lightscape and the ProCDRS-02 viewset. Before each benchmark set we've provided SPEC's own description of that particular viewset so you can better understand what that particular viewset is measuring, performance-wise.

Each viewset is divided into a number of tests, ranging from 4 to 10 in quantity. These tests each stress a different performance element in the particular application that viewset is attempting to simulate. Since all applications focus on some features more than others, each one of these tests is weighted meaning that each test affects the final score differently, some more than others.

All results are reported in frames per second, so the higher the value, the better the performance is. The last result given for each of the viewsets is the WGM or Weighted Geometric Mean. This value is, as the name implies, the Weighted Geometric Mean of all of the test scores. The formula used to calculate the WGM is as follows:

With n being the number of tests in a viewset and w being the weight of each test expressed as a number between 0.0 and 1.0.

If you'd like to know more about why a Weighted Geometric Mean is used, SPEC has an excellent article detailing just why, here.

We ran the SPECviewperf 6.1.1 package under NT for a high-end workstation performance comparison. In order to place the strain on the CPU, we replaced the GeForce 256 with a regular TNT2 Ultra which does not feature any on-board geometry acceleration thus offloading all transforming & lighting requests onto the CPU.

Advanced Visualizer (AWadvs-03) Viewset

Taken from http://www.spec.org/gpc/opc.static/awadvs.htm

Advanced Visualizer from Alias/Wavefront is an integrated workstation-based 3D animation system that offers a comprehensive set of tools for 3D modeling, animation, rendering, image composition, and video output. All operations within Advanced Visualizer are performed in immediate mode with double buffered windows. There are four basic modes of operation within Advanced Visualizer:

• 55% material shading (textured, z-buffered, backface-culled, 2 local lights)
• 95% perspective, 80% trilinear mipmapped, modulated (41.8%)
• 95% perspective, 20% nearest, modulated (10.45%)
• 5% ortho, 80% trilinear mipmapped, modulated (2.2%)
• 5% ortho, 20% nearest, modulated (.55%)
• 30% wireframe (no z-buffering, no lighting)
• 95% perspective (28.5%)
• 5% ortho (1.5%)
• 10% smooth shading (z-buffered, backface-culled, 2 local lights)
• 95% perspective (9.5%)
• 5% ortho (.5%)
• 5% flat shading (z-buffered, backface-culled, 2 local lights)
• 95% perspective (4.75%)
• 5% ortho (.25%)

If your usage patterns are characterized by the Awadvs-03 viewset as described above, then the SuperG manages to distance itself from the competition quite nicely. The only system that comes close to beating it is the Pentium III 800 on an i820 motherboard, but if you factor in the cost of the 128MB of RDRAM used in the tests (approximately $1000), the SuperG isn't that much more than an equivalently configured Pentium III 800 system on an i820 motherboard with RDRAM.

DesignReview (DRV-06) Viewset

Taken from http://www.spec.org/gpc/opc.static/drv.htm

DesignReview is a 3D computer model review package specifically tailored for plant design models consisting of piping, equipment and structural elements such as I-beams, HVAC ducting, and electrical raceways. It allows flexible viewing and manipulation of the model for helping the design team visually track progress, identify interferences, locate components, and facilitate project approvals by presenting clear presentations that technical and non-technical audiences can understand. There are 6 tests specified by the viewset that represent the most common operations performed by DesignReview.

While the SuperG is definitely faster than the air-cooled Athlons (obviously), it seems to be beat by the Pentium III as a result of its faster L2 cache (2X as fast with the Pentium III 800) and SIMD FP instructions provided for by the P3's SSE enhancements.

It all depends on your usage patterns when dealing with professional applications, and if this viewset is representative of the tasks you normally use your system for then the SuperG may not be a wise investment.

Data Explorer (DX-05) Viewset

Taken from: http://www.spec.org/gpc/opc.static/dx.htm

The IBM Visualization Data Explorer (DX) is a general-purpose software package for scientific data visualization and analysis. It employs a data-flow driven client-server execution model and is currently available on Unix workstations from Silicon Graphics, IBM, Sun, Hewlett-Packard and Digital Equipment. The OpenGL port of Data Explorer was completed with the recent release of DX 2.1.

The tests visualize a set of particle traces through a vector flow field. The width of each tube represents the magnitude of the velocity vector at that location. Data such as this might result from simulations of fluid flow through a constriction. The object represented contains about 1,000 triangle meshes containing approximately 100 verticies each. This is a medium-sized data set for DX.

This is definitely not a strong point for the Athlon platform in general. Not only is the superior L2 cache speed of the Pentium III making the Athlon crave an on-die L2 cache (it'll get one in time) but the added memory bandwidth provided for by the RDRAM on the i820 test system gives the Pentium III 800 a clear lead over everything else.

Lightscape (Light-03) Viewset

Taken from: http://www.spec.org/gpc/opc.static/light.htm

The Lightscape Visualization System from Discreet Logic represents a new generation of computer graphics technology that combines proprietary radiosity algorithms with a physically based lighting interface.

There are four tests specified by the viewset that represent the most common operations performed by the Lightscape Visualization System

We have a return to the performance throne with the Lightscape viewset and the SuperG as the 1GHz Athlon is approximately 11% faster than the Pentium III 800 on the i820 board. The Athlon is performing very well in this test and manages to distance itself from the Intel counterparts on a clock for clock basis, even in spite of Intel's faster RDRAM in the case of the P3-800/i820 tests.

This just goes to show you that performance depends entirely on what applications you're going to be using your system for, especially in the professional world.

ProCDRS-02 Viewset

Taken from: http://www.spec.org/gpc/opc.static/procdrs.htm

The ProCDRS-02 viewset is a complete update of the CDRS-03 viewset. It is intended to model the graphics performance of Parametric Technology Corporation's CDRS industrial design software.

For more information on CDRS, see http://www.ptc.com/icem/products/cdrs/cdrs.htm

The viewset consists of ten tests, each of which represents a different mode of operation within CDRS. Two of the tests use a wireframe model, and the other tests use a shaded model. Each test returns a result in frames per second, and a composite score is calculated as a weighted geometric mean of the individual test results. The tests are weighted to represent the typical proportion of time a user would spend in each mode.

All tests run in display list mode. The wireframe tests use anti-aliased lines, since these are the default in CDRS. The shaded tests use one infinite light and two-sided lighting. The texture is a 512 by 512 pixel 24-bit color image.

This last viewset once again falls into Intel's court, especially with the greater memory bandwidth provided by the RDRAM on the i820 test platform. This is just another sign of how badly AMD needs the arrival of the Thunderbird and Spitfire cores with their full speed on-die L2 cache.

The SuperG is obviously the fastest x86 system on the market, and Kryotech has definitely gotten their act together with making sure that enough resellers can get their hands on these systems in order to provide these to the public.  Kryotech is still selling the bare bones systems (this includes the CPU, motherboard, case and cooler) from their on-line store for $2495, but you can also get them fully configured with peripherals from their authorized resellers.  Next week, we will be taking a look at one such system from one of their resellers, SYS Technology, Inc. 

With air-cooled Athlons currently available at 700MHz, and the 750/800MHz Athlons due out for public sale early next month, a 1000MHz Athlon at this point wouldn’t be worth the added price premium to business users or gamers looking for the latest and greatest.  In a few months time we can expect to see air-cooled Athlons approach 900MHz+, thus making the SuperG obsolete.  So why on Earth would Kryotech even bother producing the SuperG?

Unlike the Cool K6-X series of Kryotech systems, the SuperG is actually ideal for high end workstations that require the fastest performance available today.  Especially for those individuals working off of grant money that are given a certain amount to spend at one given time, the SuperG makes the perfect product if you have the money now and you won’t be given an opportunity like this again. 

Unfortunately, the low cache speed of the SuperG in comparison to the newly released Pentium III 800 makes the 800 an overall better value for some applications that are very L2 cache dependent, especially applications that contain Pentium III specific SSE optimizations, a rapidly growing population. 

Overall, the SuperG is a very powerful and quite fast system to use, but it is definitely not worth the price premium.  The SuperG does come with more of a guarantee of future CPU compatibility than previous Cool K6-X systems, whether or not this will allow you to simply stick in an Athlon with on-die L2 cache and run it at a much higher clock frequency is up to you to decide.  Chances are that you will be able to use the SuperG with future Slot-A CPUs on future Slot-A motherboards, so long as they meet the layout requirements of the KryoCavity.  Fitting even the largest motherboards into the KS-298 case shouldn’t be a problem at all, the main thing is making sure that the area around the Slot-A connector is as unpopulated as possible by high standing capacitors and other obstructing components.

Once again, it seems like Kryotech is offering the fastest of a certain class of CPUs at an added price premium, but the performance advantage they offer just never seems to be enough to justify selling the systems to the majority of users out there.  The SuperG definitely has its own niche and what an expensive one it is.