Needless to say, that omission is a deal-breaker for many of us, especially those interested in games. Shuttle has addressed this concern with its Pentium 4 line, with both the SS51G and the SB51G. Athlon fans, however, were forced to wait, though Shuttle has showed off the nForce2-based SN41G2 at a trade show here and there.
The SN41G2 isn’t yet available for purchase, but those wanting an Athlon-based AGP Shuttle cube have another option: The SK41G, based on VIA’s KM266 chipset. The KM266 is missing features present in other VIA chipsets, like AGP 8X and support for 333MHz Athlons, but many consumers might eschew those features for a small form factor. Does the SK41G bring AGP magic to Athlon cubes? Read on to find out.
Let’s start off by taking a look at what the SK41G and its KM266 chipset have to offer.
|CPU support||Socket A-based AMD Athlon processors|
|North bridge||VIA VT8375|
|South bridge||VIA VT8235|
|PCI slots||1 32-bit/33MHz|
|AGP slots||2X/4X AGP|
|Memory||2 184-pin DIMM sockets
Maximum of 2GB of DDR266/200 SDRAM
|Storage I/O||Floppy disk
2 channels ATA/100
VIA AC97 6-channel audio
|Ports||1 PS/2 keyboard, 1 PS/2 mouse,
2 serial, 4 USB 2.0 (2 front, 2 rear), 3 IEEE 1394 (1 front, 2 rear), 1 RJ45 Ethernet via Realtek RTL8100B, 1 DB15 VGA out,
1 S-Video out via Chrontel 7005C
2 line out/front out (1 front, 1 rear), 1 rear out, 1 bass/center out, 1 mic in (front), 1 optical SPDIF out (front)
|Bus speeds||100-165MHz in 1MHz increments
(200-330MHz double pumped)
|Monitoring||Voltage, fan status, and temperature monitoring|
As mentioned previously, the KM266 chipset does give up some features relative to its newer cousins. Among those left off the list are ATA-133, AGP 8X and support for DDR333 memory. The Athlon XP 2600+ is the fastest processor supported, and those who can read between the lines will realize that the KM266 lacks support for an official (non-overclocked) 333MHz bus.
That sounds like a lot to be missing, but let’s break it down. It’s been repeatedly demonstrated that ATA-133 doesn’t really offer a significant performance advantage over ATA-100, which is underscored by the fact that most hard drive manufacturers haven’t implemented ATA-133. As for AGP 8X, our look at NVIDIA’s new 8X AGP cards revealed that “[t]he faster AGP mode just isn’t stressed by current games.” It’s extremely likely that AGP 8X will prove more useful in the future, but it will probably be quite some time before that is the case.
The lack of 333MHz bus and DDR333 support is more serious, of course, because there are already processors available that take advantage of these features. Playing devil’s advocate for a moment, though, there are currently only two Athlon speed grades that the SK41G won’t support. The Athlon XP 2600+ may not be the fastest chip on the planet, but it could hardly be considered slow.
We’ve taken a look inside Shuttle cubes before, but it seems that each model differs at least slightly from the others, and the whole line is so well-designed in general that it’s worth revisiting.
The later cubes have an appearance that’s almost Mac-like, and I mean that in a good way. The outside is all brushed aluminum and rounded edges, with a stylish faceplate that varies depending on the model but always looks good. As with the other cubes, there are bays for one 5.25″ and two 3.5″ (one internal) drives, with blanks that match the front face. The front face has all that ports you’d typically like up front, including USB for mice or keyboards, Firewire for easy camcorder hookup, and microphone and headphone jacks. Power and reset buttons round things out.
The back of the cube looks pretty similar to the SS40G, but in case you’re not familiar, here’s a picture:
You can see that three thumbscrews hold the cover in place. There are exhaust holes for the small power supply fan and the internal 80mm fan. On the right are the backplates for the expansion slots, and at the bottom is the port cluster. The SK41G has just about everything you’d expect in terms of ports, though the parallel port has been omitted, probably because it takes up too much real estate. The yellow connector is S-Video, in case you were wondering.
Take out the three thumbscrews on the back of the SK41G and the cover slides back and then lifts off, revealing the innards:
A tiered tray of sorts holds up to three drives, and it must be removed to access the other parts of the box. Fortunately, that’s as easy as removing two small screws and sliding the drive tray back and out. Once the tray has been removed (and the cables have been pulled out of the way) you can see more of the FX41 motherboard:
In case you’re confused, the two pictures were taken from opposite sides of the cube; look for the fan shroud in both pictures to see what I mean. Looking down into the box, you can see the heatpipe assembly installed onto the processor as well as the north bridge heatsink. Now let’s take a closer look at the heatpipe, as well as the motherboard itself.
I went to the trouble of removing the FX41 motherboard from the case, so here’s a proper picture of it:
Isn’t it cute? If the location of some of the connectors appears awkward, don’t worry, it’s not an issue. The SK41G comes with IDE and floppy cables specially designed for the enclosure. The extra-short hard drive cable includes a rubber pull ring to make it easier to remove in the tight confines of the case, while the CD-ROM cable is shaped so it will tuck into clips in the enclosure that keep it out of the way. Finally, take a look at the top of the picture and you’ll see the star of the show, the AGP slot.
I’m not certain if the heatpipe that came with our SS40G review unit was a pre-release sample or if Shuttle has just improved its design, but the SK41G’s heatpipe is considerably more professional-looking, as well as easier to use. The tubing leading from the heatsink to the radiator fins is now a nice silver color, as opposed to the bare copper of the SS40G heatpipe. The top of the heatsink may look a little dirty in this photo, but the business side is extremely clean and well-polished.
One of the things that I found frustrating about the SS40G heatsink was the retention mechanism. It was a spring clip that utilized all three mounting tabs on each side of the socket. Between the tall fins on the top of the heatsink and the tight confines of the case, securing the heatsink was certainly possible, but the experience ranged from annoying to frustrating.
The SK41G’s retention mechanism is considerably better, using four thumbscrews that go through holes in the motherboard and screw into holes in the bottom of the case. As you can see from the photo, small springs on each thumbscrew actually apply the pressure that holds the heatsink to the processor core, so it’s impossible to overtighten the screws. This new mechanism makes heatsink installation and removal a lot easier than on the SS40G.
Finally we have a look at the radiator fins. The heat from the processor is conducted up through the heatpipes, where it is in turn conducted to the fins. A shrouded fan mounted next to the fins takes air from the inside of the enclosure and blows it through the fins and out of the case. The system is so good at keeping things cool that the fan can run at a relatively low speed, keeping noise down.
When it comes to benchmarks, we’re in an interesting position, in that the most obvious choice for comparison is the SN41G2, which isn’t available quite yet. However, previous cube reviews have shown that the performance of a cube system with a given chipset is comparable to a full-sized motherboard using the same chipset. Thus, we decided to benchmark the SK41G against the Asus A7N-8X, an nForce2 motherboard. Once the SN41G2 arrives, expect a direct SK41G to SN41G2 comparison in our review of the SN41G2. As ever, we did our best to deliver clean benchmark numbers. Tests were run at least twice, and the results were averaged. Our test systems were configured like so:
|Shuttle SK41G||Asus A7N-8X (pre-release)|
|Processor||Athlon XP 2600+ 2.13GHz||Athlon XP 2600+ 2.13GHz|
|Front-side bus||266MHz (133MHz DDR)||266MHz (133MHz DDR)|
|Chipset||VIA KM266||NVIDIA nForce2|
|North bridge||VIA VT8375||nForce2 SPP|
|South bridge||VIA VT8235||nForce2 MCP-T|
|Chipset drivers||VIA 4 in 1 4.45v||2.77|
|Memory size||512MB (1 DIMM)||512MB (2 DIMMs)|
|Memory type||Corsair XMS3200 PC2700 DDR SDRAM||Corsair XMS3200 PC2700 DDR SDRAM|
|Graphics||S3 ProSavageDDR (13.93.65 drivers)
ATI Radeon 9700 Pro 128MB (Catalyst 7.76 drivers)
|ATI Radeon 9700 Pro 128MB (Catalyst 7.76 drivers)|
|Sound||VIA VT8235 AC97 audio||Creative SoundBlaster Live!|
|Storage||Maxtor DiamondMax Plus D740X 7200RPM ATA/133 hard drive|
|OS||Microsoft Windows XP Professional|
|OS updates||Service Pack 1|
We used the following versions of our test applications:
- Cachemem 2.6
- SiSoft Sandra Standard 2002 SP1
- ZD Media Business Winstone 2002 1.0.1
- ZD Media Content Creation Winstone 2002 1.0.1
- Sphinx 3.3
- MadOnion 3DMark 2001 SE Build 330
- Unreal Tournament 2003 demo benchmark
- Comanche 4 demo benchmark
- Quake III Arena v1.31
- Serious Sam SE v1.07
The test systems’ Windows desktop was set at 1024×768 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.
All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
Our previous tests with Shuttle cubes have revealed that the onboard video comes at a performance penalty, and the SK41G is no exception. Because a chunk of main system RAM is reserved for the onboard video, the video effectively steals memory bandwidth from the rest of the system, as evidenced by the Sandra scores.
While the Radeon 9700-equipped SK41G configuration does considerably better than the onboard video alternative, the nForce2 motherboard does even better, turning in significantly higher scores in both Sandra tests.
While Sandra’s raw bandwidth numbers are impressive, they aren’t terribly representative of real-world memory bandwidth. When it comes to measuring memory bandwidth, we look to Cachemem for a more realistic picture. This benchmark reveals less of a hit for the onboard video solution, but at the same time, the nForce2 extends its lead over the KM266 considerably.
Bandwidth isn’t everything of course, and this graph explores that other important memory statistic, latency. Here the KM266 and nForce2 are quite a bit closer, but the nForce2 still grabs the win. The onboard video configuration trails by a large margin.
The onboard video exacts a heavy penalty here, with that configuration scoring over eleven percent slower than the Radeon 9700 config. The nForce2 once again manages to pull ahead.
Content Creation Winstone
Content Creation Winstone’s applications are more resource intensive than Business Winstone, and the results reflect that fact. This time, the onboard video scores over seventeen percent lower than the Radeon 9700 configuration. The nForce2 comes in six percent faster than the KM266.
To paraphrase the Simpsons, all work and no games make Homer something something, so let’s look at some gaming scores. If you’re uncomfortable with the sight of carnage, I suggest you avert your eyes from the following graphs, because the results aren’t pretty.
Once again the nForce2 comes out ahead of the KM266, video cards being equal, but the real story is the sickly triple digit score from the onboard video. If the rest of the gaming scores look like this, anyone considering the SK41G’s onboard video for gaming would do well to consider something else.
The nForce2 comes out ahead once again, clocking in around seven percent faster than the KM266. If you thought the onboard video couldn’t do worse than the last test, you were wrongit gets a DNF here, as the benchmark refused even to run.
Quake III Arena
For those of you who were wondering if the KM266 onboard video was capable of performing better with older games, the answer is perhaps, if you’re willing to crank the resolution and detail way down. At our standard test settings, however, the KM266 turns in a disappointing twenty frames per second, not exactly a playable frame rate. The nForce2 configuration scores another victory here, but it’s largely a symbolic one; if you’re unsatisfied with a 255 fps rate in Q3A, you should probably seek professional help.
The story continues in UT2003, as the onboard video continues to set all the wrong records. The nForce2 chalks up a couple more wins here, turning in significantly higher scores than the KM266.
Serious Sam SE
SSDG: Same story, different graph. Serious Sam benefits even more from the nForce2, as that configuration scores eleven percent higher than the KM266.
Sphinx is a high-quality speech recognition routine that needs the latest computer hardware to run at speeds close to real-time processing. We use two different versions, built with two different compilers, in an attempt to ensure we’re getting the best possible performance.
There are two goals with Sphinx. The first is to run it faster than real time, so real-time speech recognition is possible. The second, more ambitious goal is to run it at about 0.8 times real time, where additional CPU overhead is available for other sorts of processing, enabling Sphinx-driven real-time applications.
None of the configurations tested here achieves the real-time goal, but the nForce2 configuration comes considerably closer than either of the KM266 configs. The shared memory of the onboard video obviously takes a toll on performance in this benchmark.
The SK41G is an impressive machine in its own right, and has the distinction of being the first Athlon-based Shuttle cube with an AGP slot. If you look at it on its own merits, it achieves very respectable performance that is on par with many full-sized desktop PC’s, at least when it’s paired with a good graphics card. It lacks some of the newer features available on other chipsets, such as support for DDR333 and a 333MHz front-side bus, but for the moment, it’s the only thing going.
However, those who follow this stuff have their eyes on the other shoe, and it’s about to drop. The SN41G2 actually arrived in our offices as this review was going to press, and you can expect a full review of it very soon. Like the Asus board tested in this article, the SN41G2 has support for DDR333 and 333MHz front-side bus Athlons, and dual-channel DDR support for an extra burst of speed. It also has AGP 8X, ATA-133 and dual Ethernet ports. It even has dual VGA outputs, enabling you to run dual monitors with the onboard graphics. And of course, those onboard graphics will likely offer considerably more 3D punch than the ProSavage DDR in the SK41G. With all that, the obvious question is, why would anyone want an SK41G?
Well, let me try to answer that question as best I can at this point. We don’t yet have pricing information on the SN41G2 from Shuttle, and Pricewatch is mum on that point as well. What follows requires a bit of supposition, but let’s look at the SB51G as an example: Retail is $349, and Pricewatch prices are around $330 shipped. The SB51G came out around the beginning of November, so it’s been on the market for about six weeks now. Given that the SB51G and the SN41G2 are both “top of the line” for their respective processor architectures, it’s likely that the SN41G2 will cost around the same as the SB51G, or perhaps a bit more given its better onboard video and dual video outputs.
The SK41G, on the other hand, has only been available for a couple of weeks, yet it’s currently available on Pricewatch for $285 shipped, about $45 cheaper than our made-up (but probably fairly close) SN41G2 price. If you’re building a cube for mom and dad, who don’t care about 3D graphics, let alone DDR333, that’s a significant savings. If the performance hit from the onboard video is too much for you, slap in a cheapo video card for $10. And hey, if you’re interested in a cube for LAN parties or light-duty gaming, but don’t need the absolute latest and greatest, that $45 could go towards a nicer AGP card.
I know what you’re thinking: “But the SN41G2’s onboard graphics are good enough that you don’t need an AGP card!” Well, it’s true that they’re a damn sight better than the SK41G’s onboard offering, but don’t forget that the SN41G2 doesn’t have onboard frame buffer memory, so it’s using main memory just like the SK41G. It’s possible that the use of dual channels of DDR333 will mean less of a benchmark-neutral performance hit than we saw with the SK41G, but it’s also possible that the SK41G with an AGP card will equal or better the SN41G2 with onboard video, even on benchmarks with no 3D graphics. These questions will be answered in our SN41G2 review shortly. Regardless, the SK41G packs a lot of value for the money. If you’re looking to build a DirectX 8 (or even DirectX 9) capable LAN party box for as little as possible, the SK41G and a nice AGP card is a tough combo to beat.