|
Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home |
. | . |
|
Prices: Mainboards ABIT ASUS Chaintech Shuttle Soyo Tyan 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 
|
LOSTCIRCUITS
|
|
| AMD Athlon64 "Venice" May Low Power be with you! | |
|
(Review by MS May 2, 2005) |
| AMD Athlon 64 4000+ |
Overclocking
We managed to overclock our 3800+ to 12 x 245 MHz or 2.94GHz at least to the point where the CPU would POST using standard air cooling. However, leaving the core voltage setting at "Auto" caused the supply voltage to automatically increase from 1.4V to 1.47-1.48V (as shown in the CPUZ screenshot below), which is one of the reasons for the increased power consumption shown in the Caligari TrueSpace results. Stable operation was possible up to 2800 MHz.
At 2806 MHz the Venice core was barely breaking a sweat. Note, however, the increased Vcore
Conclusion
What was originally meant as a quick review turned out to be a major science project in the wake of a few dead motherboards that were not capable of handling the soldering on the CPU VRM�s MOSFETs. Likewise, we went through a number of induction �current meters until we found one that could be calibrated to the specific needs of this review. And then, we spent about one week just finding the bugs and sudden runaways but in the end, it was worth it, the method is rather straightforward and appears to be sensitive and reliable enough to yield reproducible and hopefully accurate results. Keep in mind that the lower the numbers, the greater will be the systematic error, we mention this as disclaimer primarily for the Windows Idle numbers. All other numbers are well within the range where potential read errors could be cross-correlated against the built-in read-out of the BK Precision model 1692, 30V � 40 A DC power supply.
It certainly appears as if the transition from the 130 nm process to the 90 nm process in conjunction with the use of Silicon-On-Insulator and the dual stress liner technology has been implemented very successfully by AMD. The dissipation numbers deliver a very powerful message, likewise, the new core appears to have enough headroom to scale up a bit beyond the current rating and with refined silicon mixtures probably beyond 3 GHz. The sample at hand was working up to 2.8GHz, even though above 2810 MHz, some errors did occur in e.g. Prime95.
Overclocking 400 MHz above the rated speed would have been phenomenal in the days of the K6 family or the P2, the Pentium4, on the other hand appears to do the same on a regular basis without breaking any sweat as long as the infrastructure of supporting components including motherboard and memory is up to snuff. Still, for the latest generation of Athlon64 CPU, 400 MHz are nothing short of spectacular, after all, the Athlon64 was developed primarily for efficient use of clock cycles especially in gaming environments.
On the other hand, as we showed, any increase in clock speed incurs a rather hefty increase in power dissipation that needs to be factored in. At the rate we were going, we expect to hit the 70W mark for real life power consumption at about 3- 3.1 GHz, which we believe to become the commercially viable upper limit for the Venice core in its present form.
Performance per Watt
We have seen all sorts of metrics including those used by AMD to justify their performance rating. Along these lines, we feel obliged to add another equation to the pool of more or less relevant metrics, namely fps/Watt and score points/Watt.
It is clear that the early adaptation of SOI has earned AMD / IBM a sizeable headstart, especially with respect to overcoming the problems with transitioning to yet smaller process technologies. How problematic the next transition, that is, the migration to the 65 nm process is going to be is still up in the crystal ball but the overall technology roadmap does look very promising, especially in light of the successful mastering of the 90 nm process and, by extension, the Venice core which is showing all promises of becoming a champion. Along the same lines, we raise the bar of our expectations for the 4XXX-X2 (dual) core series, we�ll have something on those shortly.
|
Athlon64-3500+ (Venice Core) |
next page: => More =>
All advice and educational articles on LostCircuits are free, but if you feel you can, please make a small donation to us!
