Articles :: Project Swiftech Part One, Extreme Overclocking Of the P4 2.4Ghz and MSI GF4 Ti4600 :: Motherboards.org

Doc Overclock · 06-14-2002 · Category: Guides

The Milestone Report


Initial Components Installation and preliminary testing report

Swiftech Quiet Power case: Stock configuration using:

  • Motherboard: Gigabyte GA-8IRXP
  • CPU: Intel P4 2.4 Northwood O/C to 3.0GHz (125FSB x24) @ 1.8 volts - ~35 C at full load
  • Memory: Crucial DDR PC2700 o CL= 2.5 o Unbuffered o Non-parity o 6ns o 2.5V o 32Meg x 64
  • Hard Drive: Maxtor 80GB ATA133 7200RPM on RAID channel
  • OS: Windows XP

Step 1: Quiet Power Active cooling components installation.

  • Installed MCW-UT (3/8" OD tubing) water-block with 226 watts thermo-electric module
  • Replaced stock (53CFM) radiator fans by two Nidec 92 CFM fans
  • Installed S320-12 300W 25A power supply
  • Modification on top of case panel to install blowhole w/53 CFM 120mm fan


Bench system test data:

  • Ambient 23 C
  • CPU temp ~23C
  • Coolant temp ~39C
  • GPU temp ~ 41.5 C

Conclusion:

Mix-and-match of active and non-active liquid cooling solutions on the same loop is ineffective in current environment, and should not be implemented as the results show poor performance in the outcome. Action => Install MCW40-T with 78Watts thermo-electric module on Quiet Power GPU.

Quiet Power Test Data

Test data 1:

Objective of test: establish reference/comparative data between straight liquid cooled CPU and both active (Peltier) combined with a liquid cooling system.

  • Ambient 20 C
  • CPU setting @ 125x24=3Ghz @ 1.775 volts
  • GPU setting @ 320/675
  • CPU Temp data: 13C in bios
  • CPU Temp data: ~18 C under Prime 95 load
  • GPU Temp ~10 C
  • 3D Mark 2001 score: 12,222

Test data 2:

Objective of test: analyze performance improvement with active cooling system.

  • Ambient 20 C
  • CPU setting @ 126 x 24=3.045GHz @ 1.775 volts
  • CPU temp data: ~ -3 C idle - ~ 20 C under 3DM2001 benchmark
  • GPU setting @ 324/693
  • GPU temp data: up to ~ 14 C under 3DM2001 load
  • 3D Mark 2001 score: 12770
  • Remark/problem: memory setting low compared to 3Dmark database
  • Action: => Installation of GF4 memory heatsinks

Test Data 3:

Objective of test: analyze performance improvement with active cooling system.

  • CPU setting @ 126 x 24=3.045GHz @ 1.775 volts
  • GPU setting @ 324/772
  • Installation Detonator 27.70 drivers
  • 3D Mark 2001 score: 13,007
  • Remark/problem: CPU frequency is low versus calculated data based on operating temperature: 15 C gain compared to non-active should result in a ~90-120Mhz CPU frequency gain => 3120Mhz
  • Action=> investigate AGP frequency/Graphics Card as limiting factor
  • Action=> investigate Memory as limiting factor

Reference test using the ATI RADEON 8500

Objective of test: investigate AGP frequency as limiting factor in VGA overclocking.

  • CPU setting @ 130 x 24=3.121GHz @ 1.8 volts
  • GPU setting @ 300/300
  • 3D Mark 2001 score: 10791
  • Test reveals CPU clocks are higher with ATI graphics card, hinting at a higher CPU frequency potential

Test data with GF4 TI4600

Objective of test: investigate memory frequency as limiting factor, and finalize GF4 sensitivity to an overclocked AGP frequency.

  • Multiple tests at high AGP frequency (above 66Mhz) evidence failure of GF4 to sustain overclocked AGP bus frequency.
  • Multiple tests at lower memory multiplier of 2.0 instead of 2.66 reveal CPU clocks higher than in Test Data 3, confirming higher CPU frequency potential
  • Solution => set AGP clock speed to strict specification compliance
  • Solution =>Install DDR Ram prototype heatsinks to enhance system memory cooling

Test Data 4:

Objective of test: analyze performance improvement with above modifications.

  • CPU setting @ 130 x 24=3.120GHz @ 1.775 volts
  • GPU setting @ 345/775
  • Ambient 20 C
  • GPU temp data: up to ~ 14 C under 3DM2001 load
  • 3D Mark 2001 SE score: 13438
  • Remark/problems:Core GPU frequency still relatively low compared to Mad onions 3Dmarks database
  • 3Dmark score still unsatisfactory compared to goal of 14,000
  • Solution=> Installation of dedicated GPU liquid cooling including:
  • Prototype Swiftech 120mm radiator
  • 93 CFM Nidec fan
  • 1048 water pump
  • F&B kit in top panel
  • Solution: investigate PC3000 DDR memory
  • Solution: investigate WIN2K performance as OS vs. XP

Test Data 5:

Objective of test: analyze performance of dedicated GPU cooling, and compare performance of 3Dmarks under WIN2k vs. XP.

  • Ambient 20 C
  • CPU setting @ 130 x 24=3.120GHz @ 1.775 volts
  • GPU temp data: up to ~ 20 C under Prime 95 (2 hours)
  • GPU setting @ 345/775
  • GPU temp data: up to ~ 7 C under 3DM2001 load
  • 3D Mark 2001 SE test scores dropped ~200 points, hinting XP as possibly superior OS for 3Dmark benchmarking Multiple attempts to increase the GPU frequency failed despite gaining lower operating temperatures. There are uncertain issues with core GPU voltage.
  • Solution=> revert to windows XP
  • Solution=> Install OCZ PC3000 DDR memory

Test data 6:

Objective of test: Reanalyze performance gains with PC3000 memory.

  • Note: Case temp (closed case) 21 C
  • CPU setting @ 134 x 24=3.216GHz @ 1.8 volts
  • Memory settings CAS 1.5 -5-2-2 @ 7.8V - Frequency 353Mhz
  • GPU settings @345/775
  • 3D Mark 2001 SE score: 13917
  • Remark/problems: System Memory and GPU frequency/settings being optimized, CPU frequency is only parameter left to improve.
  • Solution => The Installation of half inch tubing and a high power flow pump for CPU cooling system.

Due to unavailability of some of the necessary components, running water will be used to test the scalability of the CPU performance by adding an additional reduction in operating temperature. Yes I do mean running water from a sink or outside garden hose that is slightly a pain to have to deal with but can be done rather painlessly depending on the location of your PC. The water only has to be set with very low pressure.

Test Data 7:

Objective of test: analyze performance results with running water.

  • CPU setting @ 138 x 24=3.312GHz @ 1.8 volts - Running water
  • Memory settings CAS 1.5 -5-2-2 @ 7.8V - Frequency 359Mhz
  • GPU settings @345/775 - running water
  • Ambient temp 20 C
  • CPU temp at load ~10 C
  • GPU temp at load ~0 C
  • 3D Mark 2001 score: 14107

Contents

  1. Introduction
  2. Setup
  3. What is a Peltier Device?
  4. The Milestone Report
  5. Performance
  6. Conclusion