Intel 570 CPU + Soltek SL-915GPRO-FGR Motherboard

CPUs|Motherboards
Viewing page 4 of 4 pages. Previous 1 2 3 4

CPU TEMPERATURE & SYSTEM POWER


Screenshot of full 2xCPUBurn load at 200FSB, 1.175V.

After running the CPU at default settings I decided to try some mild overclocking to see how this effected the system temperature and overall power draw. I left the Vcore at default and increased the FSB until I started getting errors while running 2 x Prime95. This particular CPU started maxing out around 215FSB at default Vcore. I backed down to 212FSB and took temp and AC power readings at that speed. Temps and AC power draw increased slightly over the default setting with this 6% OC.

Next I wanted to see how far I could undervolt the system while still remaining at default FSB. I managed to get down to 1.120V before I started getting errors in Prime95. I bumped the Vcore back up to 1.175V and ran my temp and load tests at that setting. The temps and AC power draw decreased by a significant amount with this drop of 0.15V or just over a 10% reduction in voltage. This is one of the most dramatic drops in temperatures and current draw that I've ever seen from a mild undervolt. It is an indication of just how much current is drawn by the CPU so that even tiny changes in the voltage impacts total power significantly. (Basic Ohm's Law: Power = Voltage x Current)

The AC power draw of this undervolted 3.8GHz Prescott processor was essentially the same as the current draw of my P4-3.0 Northwood system running at default Vcore. That's pretty impressive for a processor with the power-hungry reputation of the Prescott. Keep in mind this is one of the new "EO" Prescott cores that are supposed to use less power than the first versions. We haven't been able to find any firm documentation about the power draw of these new steppings though, so for now things will remain a little blurry.

CPU TEMPERATURE with XP-120 + 5V OEM Panaflo 120mm fan
Front Side Bus & CPU voltage
200FSB
Default Vcore (1.312V)
212FSB
Default Vcore (1.312V)
200FSB
Vcore: 1.175V
Temperature
AC Power
Temperature
AC Power
Temperature
AC Power
Idle
38°C
108W
40°C
111W
34°C
86W
2 x Prime95
59°C
176W
63°C
184W
49°C
137W
2 x CPUBurn
63°C
217W
65°C
223W
51°C
158W

NOTE on FAN SPEED: The temperature results shown above are only with the big Panaflo fan at 5V, which measures ~20 dBA/1m and ~25 CFM airflow. With the Panaflo fan at 12V, the noise jumps to ~35 dBA/1m and airflow reaches ~70 CFM. Under full load at default speed and Vcore, the CPU temp then dropped to just 51°C.

Mike's Notes on System Power Requirements

The 217W system AC power draw at default settings translates to about 167W DC power delivered to the components. This is based on the 77% AC/DC conversion efficiency measured at this power level for the Seasonic Super Silencer Rev.A1 PSU used for the test system. It's highly unlikely that a P4-3.8 would be used in such a minimalist configuration as our test setup. If we add a typical high power VGA card, another hard drive and a PCI sound card, we would see the AC power draw jump up to a little over 300W at maximum loads. DC power demanded of the PSU would be ~250W. Still, to ensure adequate headroom for a P4-3.8 system, I would recommend a quality PSU honestly rated for 400W output, preferably one that provides high current on the 12V line (as called for in the ATX12V PSU Design Guide version 1.3 and higher). If you opt for SLI (dual VGA cards), then perhaps a ~450W PSU compliant with ATX12V v2.0 is safer.

One final thing I wanted to test was to determine at what temperature CPU throttling kicked in. The 570J is equipped with Thermal Management Throttling, as are all P4s. The function is triggered by the CPU's internal temp monitor to protect the processor from failure due to overheating, and works by inserting idle processing cycles in the CPU's instruction pipeline (TM1), or by reducing the internal clock frequency of the CPU (TM2).

A utility called Throttlewatch by Panopsys was used to detect and report CPU throttling activity. The fan on the HS was swapped for one with lower airflow, and its speed turned down to slowly allow the CPU temperature to rise. Throttlling began at a low 67°C and became quite extensive at any higher temperature. In fact, the throttling was extensive enough so that the CPU temperature never rose past 70°C.

CONCLUSIONS

This was a triple first for me: First experience with...

  • a Socket T board
  • a Prescott CPU
  • a Soltek motherboard.

I like jumping in the deep end. :)

The Soltek Sl-915GPRO-FGR board was 100% stable throughout my testing. The drivers on the included CD worked fine and the manual was well written and thorough. The accessory package was remarkably complete, even down to the color-coordinated PATA IDE cables. I appreciated all the space around the CPU socket as it let me use the very effective, but very large Thermalright XP-120 cooler. The only real downside of this board was the odd placement of the PATA IDE and FDD headers, and the single native parallel ATA IDE channel. Yes, I'm aware this is a built-in limitation of the 915 chipset, but it reduces the utility of the board nonetheless. Overall this is a decent board that performed well, was 100% stable and had good adjustments for undervolting (albeit none for underclocking).

PROS

* Extremely full-featured
* Clearance around CPU for large coolers
* Fanless Northbridge chipset
* Internal USB 2.0 & Firewire headers
* Complete accessory package
* PCI-E x1 & x16 slots for early adopters

CONS

* Inconvenient location of PATA & FDD headers
* Unable to underclock
* Multicolored IDE, SATA, PCI & ATX headers

I was also pleasantly surprised with the thermal characteristics of the sample Intel 570J (P4-3.8GHz) Prescott "EO" core processor. At default settings it's warmer than I'm used to dealing with but doesn't seem quite as hot as the earlier Prescott DO steppings, and when undervolted by about 10% it runs surprisingly cool, considering its ancestry. This CPU is almost historic, as it is supposed to be the fastest P4 Intel will release now that they have shelved the goal of reaching 4 GHz and are moving away from marketing clock speed as the main index of performance. Performance benchmarks are naturally near or at the top of all processors, but in desktop apps, the difference I perceived with a P4-3.0 or even a Pentium M 2.0 Dothan system is surprisingly marginal. One has to trust the ~20% improvement in benchmarks over a P4-3.0 would translate into more perceptible gains in a gaming or processor-intensive multimedia applications.

* * *

Much thanks to Soltek for the SL-915GPRO-FGR sample, to Intel for the shiny new 570J Prescott, and to Thermalright for the LGA775-RM adapter kit. .

* * *

Discuss this this article in the SPCR Forums.



Previous 1 2 3 4

CPUs|Motherboards - Article Index
Help support this site, buy from one of our affiliate retailers!
Search: