Shuttle SN27P2 SFF: AM2 gets the Tiny Treatment
by Jarred Walton on June 30, 2006 11:00 PM EST- Posted in
- Systems
Overclocking
While overclocking and SFF systems are rarely a great match, we did test the limits of the SN27P2 in this area. Here are our results.
The initial BIOS we received with the system did not have working CPU multipliers, which severely limited our overclocking attempts when using an Athlon X2 5000+ processor. The maximum HyperTransport bus speed is 300 MHz, but if you're stuck with a 13X multiplier you'll never get there. In fact, given that the 5000+ is near the top of AMD's current dual core chips, it should come as little surprise that we were not able to overclock more than 10% without difficulties. Even at such a small overclock, with the added voltage required for stability, noise levels when the CPU was placed under a full load were substantially higher.
Our recommendation is that if you plan on getting a dual core processor, don't plan on major overclocks -- not that we really expect most people to overclock small form factor systems, but it is worth mentioning. It appears that the cooling system simply isn't capable of dealing with the heat output generated from most overclocked X2 chips, though we did test the system briefly with an FX-62 and found that it worked without difficulty at stock speeds.
We got around the BIOS multiplier limitations by installing an Athlon 64 3000+ (default clock speed of 1.8 GHz). With the lower 9X multiplier and a single CPU core, we were able to overclock to 2.7 GHz, reaching the maximum supported bus speed. Unlike many of the AM2 motherboards we've tested, however, this was only possible by dropping the HyperTransport multiplier down to 3X. In this area, overclocking was similar to what we saw on the nForce4 platform: the HyperTransport bus really didn't want to run much over 1000 MHz, but given that the HTT bus is rarely a bottleneck in terms of system performance this isn't a major concern.
If for whatever reason you are interested in buying a cheap single core Athlon 64 and putting it in the SN27P2, it appears that the system is more than capable of overclocking such a processor. On the other hand, if you want to purchase a dual core processor, either don't plan on overclocking much, or purchase an X2 3800+ and plan on topping out at around 2.6 GHz (with a noisier system than if you were to simply buy a 5000+). It could also be that the forthcoming low-power X2 chips will overclock better and/or run quieter in this type of system, but as we have not been able to get a hold of any of these we cannot say for certain.
Memory Stress Testing
The overclocking features present on the SN27P2 are a bit more limited than what we see with many high-end motherboards, and CPU voltage tops out at 1.525V. Perhaps more importantly, the DDR2 memory voltage tops out at 2.0V. Our Corsair memory is rated at PC2-8500 with 5-5-5-18 timings, but we have been able to run this memory at DDR2-800 with 3-3-3-9 timings on several other motherboards when using 2.2-2.3V. We were able to run with 4-4-4-12 timings with the RAM set to 2.0V, but we could not drop the CAS/tRP/tRCD values to 3.0 (individually or together) with the limited voltage options.
The BIOS also only offers the standard memory timing adjustments (CAS/tRCD/tRP/tRAS), which further limits our ability to tweak for maximum performance. This may result in slightly lower performance relative to the top AM2 motherboards, but for the most part the differences should be minor and not something you would notice during actual use (without running benchmarks). If you are truly concerned about memory performance, your best bet is to purchase some PC2-6400 memory that is designed to run with low latencies at lower voltages (1.8-2.0V). That type of memory may or may not overclock as much as the Corsair memory we are using, but as we just demonstrated, overclocking really shouldn't be your primary concern with the SN27P2 - or any SFF computer, for that matter.
Testing with all four DIMM slots populated stresses the memory subsystem more, but as you can see we were still able to maintain 4-4-4-12 timings. Running with 4GB of memory (or more) is really only beneficial if you are running a 64-bit operating system, however, and for the time being we continue to prefer XP over XP x64. Let's not even talk about Windows Vista x64 Beta 2 performance. However, in another year we expect 4GB configurations to become far more common.
While overclocking and SFF systems are rarely a great match, we did test the limits of the SN27P2 in this area. Here are our results.
| Shuttle SN27P2 Overclocking | |
| Processor: | Athlon X2 5000+ (AM2) |
| Chipset: | nForce 570 Ultra |
| CPU Voltage: | 1.400V (1.300V default) |
| Memory Settings: | 5-5-5-15 2T DDR2-800 Setting |
| DDR2 Results: | 13 x 215HTT = DDR2-798 |
| HT Multiplier: | 4x above 200HTT |
| Memory Voltage: | 2.0V |
| Chipset Voltage: | 1.65V |
| Maximum CPU Overclock: | 13 x 215 = 2795 MHz |
| Shuttle SN27P2 Overclocking | |
| Processor: | Athlon 64 3000+ (AM2) |
| Chipset: | nForce 570 Ultra |
| CPU Voltage: | 1.525V (1.300V default) |
| Memory Settings: | 4-4-4-13 2T DDR2-400 |
| DDR2 Results: | 9 x 300HTT = DDR2-600 |
| HT Multiplier: | 4x above 210HTT, 3X above 260 HTT |
| Memory Voltage: | 2.0V |
| Chipset Voltage: | 1.70V |
| Maximum CPU Overclock: | 9 x 300 = 2700 MHz |
| Maximum HTT Overclock: | 300 MHz (50%) - BIOS maximum |
The initial BIOS we received with the system did not have working CPU multipliers, which severely limited our overclocking attempts when using an Athlon X2 5000+ processor. The maximum HyperTransport bus speed is 300 MHz, but if you're stuck with a 13X multiplier you'll never get there. In fact, given that the 5000+ is near the top of AMD's current dual core chips, it should come as little surprise that we were not able to overclock more than 10% without difficulties. Even at such a small overclock, with the added voltage required for stability, noise levels when the CPU was placed under a full load were substantially higher.
Our recommendation is that if you plan on getting a dual core processor, don't plan on major overclocks -- not that we really expect most people to overclock small form factor systems, but it is worth mentioning. It appears that the cooling system simply isn't capable of dealing with the heat output generated from most overclocked X2 chips, though we did test the system briefly with an FX-62 and found that it worked without difficulty at stock speeds.
We got around the BIOS multiplier limitations by installing an Athlon 64 3000+ (default clock speed of 1.8 GHz). With the lower 9X multiplier and a single CPU core, we were able to overclock to 2.7 GHz, reaching the maximum supported bus speed. Unlike many of the AM2 motherboards we've tested, however, this was only possible by dropping the HyperTransport multiplier down to 3X. In this area, overclocking was similar to what we saw on the nForce4 platform: the HyperTransport bus really didn't want to run much over 1000 MHz, but given that the HTT bus is rarely a bottleneck in terms of system performance this isn't a major concern.
If for whatever reason you are interested in buying a cheap single core Athlon 64 and putting it in the SN27P2, it appears that the system is more than capable of overclocking such a processor. On the other hand, if you want to purchase a dual core processor, either don't plan on overclocking much, or purchase an X2 3800+ and plan on topping out at around 2.6 GHz (with a noisier system than if you were to simply buy a 5000+). It could also be that the forthcoming low-power X2 chips will overclock better and/or run quieter in this type of system, but as we have not been able to get a hold of any of these we cannot say for certain.
Memory Stress Testing
| Shuttle XPC SN27P2 Stable DDR2-800 Timings - 2 or 4 DIMMs 2 or 4 slots populated - 1 or 2 Dual-Channel Bank(s) |
|
| HTT Clock Speed: | 200MHz |
| Timing Mode: | 800MHz |
| CAS Latency: | 4 |
| RAS to CAS Delay: | 4 |
| RAS Precharge: | 4 |
| RAS Cycle Time: | 13 |
| Command Rate: | 2T |
| Voltage: | 2.0V |
The overclocking features present on the SN27P2 are a bit more limited than what we see with many high-end motherboards, and CPU voltage tops out at 1.525V. Perhaps more importantly, the DDR2 memory voltage tops out at 2.0V. Our Corsair memory is rated at PC2-8500 with 5-5-5-18 timings, but we have been able to run this memory at DDR2-800 with 3-3-3-9 timings on several other motherboards when using 2.2-2.3V. We were able to run with 4-4-4-12 timings with the RAM set to 2.0V, but we could not drop the CAS/tRP/tRCD values to 3.0 (individually or together) with the limited voltage options.
The BIOS also only offers the standard memory timing adjustments (CAS/tRCD/tRP/tRAS), which further limits our ability to tweak for maximum performance. This may result in slightly lower performance relative to the top AM2 motherboards, but for the most part the differences should be minor and not something you would notice during actual use (without running benchmarks). If you are truly concerned about memory performance, your best bet is to purchase some PC2-6400 memory that is designed to run with low latencies at lower voltages (1.8-2.0V). That type of memory may or may not overclock as much as the Corsair memory we are using, but as we just demonstrated, overclocking really shouldn't be your primary concern with the SN27P2 - or any SFF computer, for that matter.
Testing with all four DIMM slots populated stresses the memory subsystem more, but as you can see we were still able to maintain 4-4-4-12 timings. Running with 4GB of memory (or more) is really only beneficial if you are running a 64-bit operating system, however, and for the time being we continue to prefer XP over XP x64. Let's not even talk about Windows Vista x64 Beta 2 performance. However, in another year we expect 4GB configurations to become far more common.
Facebook
Twitter
RSS
17 Comments
View All Comments
SoSpartan - Tuesday, August 1, 2006 - link
Why do you state that the SN27P2 has a 400W power supply when your pictures clearly show a 350W power supply? Shuttle's website says 400W prominently, but their PDF spec sheet says 350W! What's the deal? The extra 50W could make the difference between running a nVidia 6800 vs 7900GTX or 7950GX2 stably!akp - Sunday, July 2, 2006 - link
Nice review in general, but I feel like there's one thing really missing. As you say on page 6 (Benchmark setup):The key feature for the SN27P2 of course is its size, but noise control is also likely to be a major consideration.
If you are simply planning on using the system as the core of an HTPC, our recommendation would be to grab one of the many fan-less GPUs that are coming to market, probably a 7600 (GT as an upgrade) or an X1600 Pro/XT would strike a decent balance between performance and power/heat requirements.
Given those factors, wouldn't it have made a lot of sense to include a fanless GPU in your benchmark setup? I really have to wonder how much of the power draw and noise was coming just from the graphics card in those tests.
JarredWalton - Sunday, July 2, 2006 - link
Actually, the majority of the noise under full load with the 7600GT comes from the CPU. However, the GPU is contributing a bit in that configuration. A fan-less GPU wasn't included because I don't have one at present. The idle noise with the 7600 GT used is almost entirely generated by the CPU/case cooling - maybe subtract 2 dB for the GPU at best.artifex - Saturday, July 1, 2006 - link
What was the temperature like when all bays were full of either hard drives or an active DVD burner? I'm wondering if it can really keep up in a situation like that. In my tower, my drives seem to get really hot without extra cooling. Please consider adding a test like this the next time you review SFFs.Also, e-SATA sounds cool, but that's strictly a drive interface. Wouldn't another firewire port have been more useful? Is e-SATA faster than FW400? If I use the firewire in the back for one external firewire drive, that means I'm left with a minifw in front for video cameras, etc. I'd really like more flexibility, there, so I'll be looking forward to your reviews of other AM2 SFFs for sure.
JarredWalton - Saturday, July 1, 2006 - link
SATA tops out at 300 MB per second these days (though in reality hard drives can't come near that number for sustained performance). FireWire 1394b tops out at 800 Mbps (one third as fast) and 1394a only hits 400 Mbps. That said, FireWire is a far more flexible interface, as there are many peripherals that use it. So far SATA is only for hard drives.As for installing three HDDs, I'd be very nervous if they were all hot drives. However, I have an SN25P with two HDDs that has been in use for about a year and it has no problems with temperatures. Under the right conditions (or wrong conditions if you prefer), the fans in the case will simply have to spin faster and make more noise. I don't think 51 dB is as loud as the SN25P can get, and the SN27P2 is basically the same in terms of cooling.
Note that the rear of the case does have two 60mm fans just for cooling the HDD area (mostly), so the only HDD spot I'd be concerned with is the floppy/HDD area, and then only if you were planning on running three HDDs along with a big GPU like a 7900 GTX or X1900 XTX. Even then, those large GPUs would probably just end up adding more noise and helping ventilation.
artifex - Monday, July 3, 2006 - link
Well, this might replace my Antex Soho tower with 3 accessory fans, so I'm used to the noise already. Also, I only have a 5500 since I'm not a big gamer, so this is looking better. Thanks again.Howard - Saturday, July 1, 2006 - link
What's an SPL meater?JarredWalton - Saturday, July 1, 2006 - link
Sound Pressure Level. Just a small device to measure how much noise is present. For example, I'm sitting in front of a 24" fan right now, which is rather noisy. It generates 55 dB of noise at a distance of six feet. Scary that some PCs are as loud as a standing fan. :|AnonymouseUser - Saturday, July 1, 2006 - link
What is the point of the last paragraph? This was a decent review until the very end."There are two major events that need to take place before we would recommend most people go out and buy this system, however. First, it needs to be available for purchase"
OK, that makes sense, but that pretty much goes without saying that it needs to be available before you can buy it.
"More importantly, as we've mentioned repeatedly over the past few weeks, you might as well wait a month now and find out what happens with the Core 2 Duo launch. That also gives the side benefit of lower X2 CPU prices, [[ so even if you're not interested in Intel's new processor lineup for whatever reason ]], the AM2 price cuts are likely to keep you waiting another month."
How does this affect the peformance of the SFF in the review? How does this affect the price of the SFF in the review?
This was a review of an AMD based SFF, not a comparison of AMD/Intel cpus, so we don't need your opinion on which CPU brand to choose.
Frumious1 - Saturday, July 1, 2006 - link
OMG Jarred! How DARE you mention crazy things like launch dates and prices cuts. WTF do those have to do with a review about an expensive-ass Shuttle SFF!? We AMD fanboys prefer to bury our heads in the sand and pretend that Core 2 Duo is just a bunch of lies and that AM2 will remain superior until hell freezes over. (Which, incidentally, is scheduled to happen around July 25th last I heard....)Seriously Anonymouse do you work for AMD or something? You many not give a damn about Core 2 Duo, but most of us don't have our blinders on. "How does this affect the peformance of the SFF in the review? How does this affect the price of the SFF in the review?"
I'd say that it means future SFFs like the SD37P2 will offer better performance than anything you can put in the SN27P2 (at least until AMD ships something other than a three-year-old K8 derivative), and it means that for the same $400+ that the Shuttles are going to cost you could end up with a better (faster and cooler running) SFF. Yes I said it: Core 2 Duo will be BETTER than AM2 X2! So stick THAT where the sun don't shine (right next to your head).
Freaking amazing how that thing called logic works, innit?
For the record, I don't work for Intel or AMD, and I wouldn't buy a SFF unless the price was the same as that of an ATX case + mobo + moderate PSU. Just about any reasonable ATX case is going to end up quieter than these SFFs if you build it right.