Pro/Forums - View Single Post

Budwise · 12-13-2005, 08:43 PM

This is Gabes post ealier this evening at Xtremesystems.org:

The following tests have been conducted specifically for the purpose of verifying the allegations published by a forum member (Orkan) comparing overclocking performance between an Apogee and a Storm water-block, where it was reported that under the same circumstances (same CPU/motherboard platform etc..)the Storm yielded a 100Mhz higher overclock than the Apogee (2.8 Ghz instead of 2.7 Ghz).

The reason why we questionned this report in the first place was the scale of the alleged difference. In effect a 100 Mhz increase on a CPU operating at 2.7Mhz is a 3% increase in processor frequency. Why doubt such a difference ? An Integrated Circuit Industry rule of thumb states that an IC will gain an average of 2 to 3% in frequency for every 10°C drop in operating temperature. We have practically verified this rule many, many times. So, while the actual CPU temperatures from one testing methodology to another can be argued ad nauseam, the report published by this user suggested to us that the CPU was actually "seeing" about a 10C drop in junction temperature, in effect resulting in this 3% frequency increase.

This, we knew to be highly implausible.

Our tests were conducted in the same fashion as a typical overclocker would.

Equipment:
Motherboard: Asus A8N Sli Deluxe
CPU: AMD A64 X2 4400+ - Unmodified (with IHS)
Waterblocks: Storm and Apogee taken straight from inventory and used "as is"
Radiator: Triple 120mm CF prototype with fans operating at 5 Volts (silent mode)
Pump: MCP655
Vcore: 1.52V (motherboard max)
OS Window 2003 Server

Since CPU overclock was the main object of this test, the memory was set to a 1/2 ratio in order to make sure that no instability would result from the memory.

Processor Load was induced by using 2 tasks of CPU Burn, one per core (Set Affinity function).

The CPU temperature was read using Asus Probe. The Air temperature was measured with an Omega thermometer and its thermocouple placed 1 foot away from the fans intake.

OC procedure:
Starting windows at 2750MHz (250*11), starting CPUZ (v1.30), set 2 tasks of CPU burn (as described above).
If no crash after 20minutes, record the results, launch clockgen and increased the HTT by one increment.
Repeat until windows crashed.

Important Note: this procedure does not provide a full measure of CPU stability but we feel is perfectly adapted to indicate overclockability within a relatively short period of time.

This being said, the results are:

With Storm, we started at 2750, and subsequently passed 2755, 2764, 2771 and reached an immediate fail at a 2781Mhz setting. So for the Storm the final "stable" overclock was 2771MHz with a final air temperature of 23.1C. The CPU temperature (reported by Asus probe) occillated between 47C and 48C.

With Apogee, we passed 2750, 2755, 2764, 2771 and 2781MHz. Windows crashed finally at 2788MHz after 10minutes. So the "pass" for this test is 2781 at an air temperature of 23.2C and CPU temperature stable on the 47C mark.

To us, the above results simply indicate one thing: there does not appear to be ANY significant difference in overclocking abilities between these two water-blocks and with this processor.

12-13-2005, 08:43 PM	#519
Budwise Cooling Neophyte Join Date: Dec 2005 Location: Texas Posts: 29	Re: Apogee from Swiftech... This is Gabes post ealier this evening at Xtremesystems.org: The following tests have been conducted specifically for the purpose of verifying the allegations published by a forum member (Orkan) comparing overclocking performance between an Apogee and a Storm water-block, where it was reported that under the same circumstances (same CPU/motherboard platform etc..)the Storm yielded a 100Mhz higher overclock than the Apogee (2.8 Ghz instead of 2.7 Ghz). The reason why we questionned this report in the first place was the scale of the alleged difference. In effect a 100 Mhz increase on a CPU operating at 2.7Mhz is a 3% increase in processor frequency. Why doubt such a difference ? An Integrated Circuit Industry rule of thumb states that an IC will gain an average of 2 to 3% in frequency for every 10°C drop in operating temperature. We have practically verified this rule many, many times. So, while the actual CPU temperatures from one testing methodology to another can be argued ad nauseam, the report published by this user suggested to us that the CPU was actually "seeing" about a 10C drop in junction temperature, in effect resulting in this 3% frequency increase. This, we knew to be highly implausible. Our tests were conducted in the same fashion as a typical overclocker would. Equipment: Motherboard: Asus A8N Sli Deluxe CPU: AMD A64 X2 4400+ - Unmodified (with IHS) Waterblocks: Storm and Apogee taken straight from inventory and used "as is" Radiator: Triple 120mm CF prototype with fans operating at 5 Volts (silent mode) Pump: MCP655 Vcore: 1.52V (motherboard max) OS Window 2003 Server Since CPU overclock was the main object of this test, the memory was set to a 1/2 ratio in order to make sure that no instability would result from the memory. Processor Load was induced by using 2 tasks of CPU Burn, one per core (Set Affinity function). The CPU temperature was read using Asus Probe. The Air temperature was measured with an Omega thermometer and its thermocouple placed 1 foot away from the fans intake. OC procedure: Starting windows at 2750MHz (250*11), starting CPUZ (v1.30), set 2 tasks of CPU burn (as described above). If no crash after 20minutes, record the results, launch clockgen and increased the HTT by one increment. Repeat until windows crashed. Important Note: this procedure does not provide a full measure of CPU stability but we feel is perfectly adapted to indicate overclockability within a relatively short period of time. This being said, the results are: With Storm, we started at 2750, and subsequently passed 2755, 2764, 2771 and reached an immediate fail at a 2781Mhz setting. So for the Storm the final "stable" overclock was 2771MHz with a final air temperature of 23.1C. The CPU temperature (reported by Asus probe) occillated between 47C and 48C. With Apogee, we passed 2750, 2755, 2764, 2771 and 2781MHz. Windows crashed finally at 2788MHz after 10minutes. So the "pass" for this test is 2781 at an air temperature of 23.2C and CPU temperature stable on the 47C mark. To us, the above results simply indicate one thing: there does not appear to be ANY significant difference in overclocking abilities between these two water-blocks and with this processor.