Navigate: Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home	.	.
Prices: CPU Intel P4 2.4C-800 P4 2.6C-800 P4 2.8C-800 P4 3.0-800 P4 3.2-800 AMD AthlonXP XP 1700+ XP 2000+ XP 2400+ XP 2500+ XP 2700+ XP 3000+ XP 3200+ Athlon64 Athlon64 3200+ Athlon64 FX-51 Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Memory Corsair Crucial Kingston Mushkin OCZ Search Prices:

LOSTCIRCUITS		SHORTCUTS: A Spring-loaded XPC At One Glance Features Mainboard Details BIOS, Test Configuration Memory I: Sandra Memory II: Cachemem I/O performance Winstones 3D Performance Overclocking, Conclusion Flame Us Shuttle Mainboards and SFF Online
Shuttle SB61G2 "Spring(dale)-Loaded"
(Review by MS, May 27, 2003)

The BIOS offers overclocking selections of up to 255MHz, which are ID'd by WCPUID as 255.58MHz. It would appear that there is not too much incentive for overclocking the Springdale chipset other than running at a higher clock frequency. In that case, based on a DDR400 memory interface there are certain limitations that jump to mind immediately, in that there are preciously few DIMMs that will actually run at 255 MHz and 2:2:2 latency settings.

Keep in mind, however, that it is the processor bus interface that is holding back performance on the memory front, as we showed in our DDR400 roundup, overclocking the system bus at a reduced memory speed will offer more bandwidth than running the memory in synchronous mode. This is one of the main differences to any AMD-based systems where running the memory faster will give performance brownie points but you never want to run the memory slower than the FSB. On the other hand, there is the problem of running the memory asynchronously to the processor bus which requires the use of fifos and synchronizers and, thus, adds latencies .... true, if the memory runs faster than the FSB but what about the other way around?

Turning the SB61G2 into Canterwood ..

What we did was overclock the P4 to 255 MHz x 12 (unlocked Engineering sample) and looked at the different performance metrics. The first thing we were looking at was SiSoft Sandra memory bandwidth and even though there are no surprises, the scores are nonetheless very impressive.

Comanche4

At a system bus of 255 MHz with the memory running in "320 MHz" mode for a a real 407 MHz, the bandwidth takes a quantum leap.

Latencies

We would have expected some sort of a worsening of latencies, however, quite the opposite happened.

Cachemem

SB61G2 vs. Canterwood .....
There is hardly any difference left between the two chipsets, at least with respect to memory access latencies where some of the data points actually show the SB61 ahead of the Canterwood. This is surprising to say the least but on the other hand, overclocking the chipset will, of course, reduce chipset latencies and make PAT an obsolete asset.

Comanche Fo(u)r Desert

Bandwidth is bandwidth, latencies are latencies and whatever happened to real benchmarks? Suffice it to say hat running at 3 GHz (12 x 250 MHz) with the memory running in DDR320 mode (in this case DDR400) we flat-out beat most Canterwood scores, that is, 62.9 fps in Comanche4, in 3DMark2001SE we missed the Canterwood by a hare's ass with a score of only 16286. UT2003 gave us 174.52 and 78.3, Canterwood material again. What else can we say?

Conclusions

There is not much to conclude here, Shuttle has done it once again by delivering a flawless product that is another milestone achievement in the development of the PC platform. We still would like to see CSA networking capability but that is another story altogether. We could go on and hilight the superior performance or even become melodramatic but none of this would add to our impression and the final verdict that the SB61G2 deserves nothing less than an editor's award.

next page:    => More =>