• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

[Ret Sticky]Overclocking sndbx for A64 939 systems with Winchester, Opteron dual core

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

hitechjb1

Senior Member
Joined
Feb 2, 2003
[Ret Sticky]Overclocking sndbx for A64 939 systems with Winchester, Opteron dual core

Here is discussed overclocking A64 939 system(s) for Winchester and beyond (Venice, San Diego, Telodo (dual core), ...). Setup, benchmarking, experiments and results, overclocking and optimization techniques, issues and resolution, ... are discussed over time.

Results of a Winchester 3000+ and a DFI LP UT NF4 Ultra-D are posted. Testings, tuning is being done over time, and the evolved system(s) will serve as a 939 testbed through Winchester and follow on.

The techniques used for the Winchester 3000+ and DFI NF4 Ultra-D can be applied readily to Venice/San Diego, ....

Suggestions and comments are welcome.


May 2006, Opteron 939 dual core results have been added.

Link to posts: Dual core 939 Opteron testing (at 3+ GHz)

Opteron 939 165 CCBBE 0610 DPMW on air (3.3 GHz CPU-Z, unstable)

opty165_3.30GHz.JPG



Winchester 3000+ CBBHD 0447 UPCW on air

lp_ultra-d_winnie3000_cbbhd_328x9_mem_328_2.95GHz.JPG


Sandra CPU, Everest bandwidth with CPU at 2.94 GHz (327 MHz x 9), memory 327 MHz 3-5-5-10 1T
Integer Dhyrstone score 13564
lp_ultra-d_winnie3000_cbbhd_327x9_mem_327_2.94GHz_sandra_everest.JPG


SuperPI 1M with CPU at 2.90 GHz (322 MHz x 9), memory 322 MHz 3-5-5-10 1T
Best SuperPI 1M is 30 sec
- CPU at 2.86 GHz (318 MHz x 9), memory 318 MHz 2.5-4-4-10 1T or
- CPU at 2.85 GHz (317 MHz x 9), memory 285 MHz 2.5-3-3-7 1T
nf4_939_lpud_cbbhd_322x9_2.90G_superpi_sandra.JPG


SuperPI 32M 27 min 42 sec, CPU 2.83 GHz (314 MHz x 9), memory 314 MHz 2.5-4-3-7 1T
lp_ultra-d_winnie3000_cbbhd_314x9_mem_314_2.5-4-3-7_2.8V_superpi32M_27m42s_cut.JPG


Prime95 24+ hr, CPU 2.72 GHz (302 MHz x 9), memory 302 MHz 2.5-3-3-7 1T
lp_ultra-d_winnie3000_cbbhd_302x9_mem_302_2.5-3-3-7_2.8V_sp2004_28hr_cut.JPG



Contents

Hardware, testing and results
DFI LanParty UT Nforce4 Ultra-D with Winchester
Hardwares and testing
Summary of results
Screen shots (Improved results)
Screen shots (First setup test results)
ScienceMark 2.0 test results at 2.85 and 2.73 GHz
DRAM Bios Setting (for TCCD) bios 02/17/05
DRAM Bios Setting (for TCCD) bios 03/10/05 @ 315 MHz
DRAM Setting for BIOS 414-3 (TCCD) with or without Cool and Quiet

Overclocking techniques
Overclocking setting for various bus frequencies (with a memory divider table)
Some overclocking scenarios for 939 Winchester/Venice/San Diego
PSU rating estimate for some 939 CPU and system
How to compare processors of different architectures and frequencies
How to determine an upper limit for CPU and memory
Bare minimum voltage at maximal overclocking
Auto-adjusting of the HyperThread system bus frequency using AUTO LDT multiplier
Comparing various CPU cores based on SuperPI 32M run time
Do "best" CPU voltage and frequency exist for overclocking
Computing/estimating power for 90 nm CPU
http://www.ocforums.com/showthread.php?p=3734677#post3734677
Very preliminary dual core performance analysis
Dual core overclocking estimation from single core statistic

Cool and Quiet
Cool and Quiet (CnQ) Testing
Running the computer Cool and Quiet
Cool and Quiet with new higher overclocking (continued)

Memory optimization
Memory modules (for 754 and 939 platforms)
Memory bus, cache and memory bandwidth (for 940, 754, 939)
Memory frequency and latency tradeoff
How much frequency increase is needed to break-even with low latency
Testing UTT and TCCD memory modules in Winchester and DFI NF4 Ultra-D setup
Some results about comparing memory frequency, memory timing, memory divider
The 4V Vdimm option for running BH-5/UTT in DFI NF4 SLI/Ultra motherboards
Corsair - AN501: Latency Settings and their Impact on Memory Performance

Stability testing
SuperPI 32M is a relative "quick" way to test CPU and system speed and stability
Stability testing using memtest, SuperPI (32M) and Prime95
Stability testing using Stress Prime 2004 (SP2004)

A64 CPU related
A64 940, 754, 939 CPU Models, OPN code, rating
Revisions and steppings
A64 Nforce4 939 Motherboards
Dual-core
Typical Overclocking Systems for 939, 754
Nforce4 chipsets with PCI-e, SLI features
Low PR 90 nm 939 Winchester (Sept 2004)
Venice and San Diego (April 2005)
About Rev E and SSE3 instructions
Major difference between Venice (E3)/San Diego (E4) and Winchester (D0)/NewCastle (CG)/ClawHammer (CG)
Overclocking frequecy and voltage of various A64
Very preliminary dual core performance analysis
Dual core overclocking estimation from single core statistic

Related A64 threads
A64 CPUs, chipsets, motherboards
A64 Overclocking Result Collection
 
Last edited:
DFI LanParty UT Nforce4 Ultra-D with Winchester

Feb 01, 2005 ordered a DFI LanParty UT Nforce4 Ultra-D right after it became available at ZZF. It arrived two days later.

Feb 06, 2005, got a Winchester 3000+ in a local store. Had the choice of 0447, 0448, 0450, 0451 of 3000+ and 3200+. Instead of a 3200+, decided on a 0447 CBBHD 3000+ for $50 less than a 3200+. Though 3200+ is generally more flexible, but a 3000+ was picked on the assumption that the DFI NF4 Ultra-D motherboard and the TCCD memory can attain 300+ MHz at 1T to deliver a CPU overclocking of 2.7+ GHz. Also eventually may switch to Venice, ... so don't want to spend too much on a CPU.

Video card is 6600 GT PCI-e, as the system is not mainly for games. Picking 6600 GT instead of an ATI X800XL (which is faster) as it is planned for eventual experimenting dual video card configuration, either in DXG or SLI mode. A 6800 GT/Ultra is too much to spend for me.

So the list (for testing the setup) is:

- Winchester 3000+ 0447 CBBHD UPCW (brand new retail box)
- XP-90
- medium CFM fan (e.g. 80 mm panoflo, 80/90/92 mm fan ~ 40 CFM), higher CFM fan (Tornado) if needed for testing
- Arctic ceramique
- DFI LanParty UT Nforce4 Ultra-D, rev R.A02, come with bios 01/25/05
- G. Skill TCCD 2 x 256 MB PC4400 rated 2-2-2-5 @200 MHz, 2.5-3-3-7 @275 MHz
(2 x 256 MB is used to avoid potential limitation on overclocking, 2 x 512 MB or 4 x 256 MB will be used if needed for eventual usage)
- BFG 6600 GT rated 525/1050 (core/memory)
- Fortron 350 W 120 mm or Antec True 550
(reuse existing parts, will get Antec True Control II 550W or Fortron Blue Storm AX500-A if needed)
Note that the Fortron 350W and Antec True 550 have 20-pin power connectors, and are not recommended for most Nforce4 motherboard with 24-pin power socket.
02/20/05: Antec True 550 being used (with 20 pin power connector).
- A slow HD (ATA66) for testing (all fast drives are tighted up in other systems), will switch to faster drives
02/20/05: Switched to a WD ATA100 drive 160 GB with 8 MB cache (non RAID yet)
- Windows XP SP1

Immediate goal is to see how high the CPU, memory bus can achieve in this system.


It should be pointed out that the above hardware list is what was used to do the setup testing. They should be used as a reference only and result may vary due to the intrinsic electrical variation in each component even for the same part number.

The parts to achieve good result is not limited to the above list.
E.g.
- BH-5/UTT based low latency memory modules
- Other heat sink such as XP-120 can be used, other fans can be used.
- Other cooling compound such as Arctic Silver can be used, I use Arctic Ceramique due to its non-electrical property and ease for cleaning.
- The PSU list for Antec True Control II 550W or Fortron Blue Storm 500W is not exclusive and the order does not imply preference.
- The exact model/manufacturer of the video card is not crucial for CPU, memory and motherboard overclocking. Choose one based on need and level of gaming. 6600/6800 can be configured to dual video card mode using NF4 Ultra/SLI motherboards, ATI cards may not.

CPU, memory and motherboard are the key components and the overclocking results may vary greatly, say 10-20% from what has been achieved. Based on recent results reported, they should have a good chance to achieve
- Winchester to 2.5 - 2.7+ GHz stable on air, ~ 1.55 V
- G. Skill TCCD 4400 to 280 - 310+ MHz, ~ 2.8 V
- DFI LP NF4 Ultra-D/SLI-DR to 300+ MHz on HTT
 
Last edited:
Summary of results

Winchester 3000+ 0447 UPCW (memtest86 boot at 3006 MHz on air, passed memtest86 at 2970 MHz, Windows boot at 2952 MHz)

nf4_939_lpud_3006MHz_boot.JPG


lp_ultra-d_winnie3000_cbbhd_328x9_mem_328_2.95GHz.JPG


Bios settings (bios dated 01/25/05):
LDT_multiplier = 3 or AUTO (3 is better to make sure HT bus is within specification. AUTO mode sometimes overclocks HT bus by 15% leading to crashes, especially when CnQ is ON.)
CPU_multiplier = 9 or AUTO (9 for 3000+, 10 for 3200+, 11 for 3500+, 12 for 3800+ Venice/4000+ San Diego, etc)
PCI-express bus = 100 MHz
memory_HTT_ratio (normally 1:1)
HTT is adjusted manually to set all frequencies in the system.
Memory timing other than tCL - tRCD - tRP - tRAS, all set to Auto for now.
Vdimm = 2.8 V (for TCCD), VLDT = 1.2 V, Vchipset = 1.5 V
(For BH-5, UTT, Vdimm = 3.3+ V with 4V jumper option if 3.3+ V is chosen to be used. Vdimm regulator MOSFET would be very hot.)
CPU VID Startup Value to "StartUp"
CPU VID Control to "AUTO"
CPU VID Special Control to "Above VID * 110%"
Without explicitly mentioning, all other bios settings are as default.
DRAM Bios Setting (for TCCD) bios 02/17/05
DRAM Bios Setting (for TCCD) bios 03/10/05 @ 315 MHz

Updated bios from 01/25/05
- to 02/17/05 (beta)
- to 03/10/05 (official bios)
There is no noticeable difference in overclocking and stability among these bioses.

More results
SuperPI 32M run completed in 27 min 40 sec screen shot <- this is good number
- CPU: 2.83 GHz (315 MHz x 9) 1.52 V
- memory: 315 MHz 2.5-4-3-7 1T 2.8 V
SuperPI 32M run completed in 27 min 52 sec screen shot (03/12/05)
SuperPI 32M run completed in 27 min 55 sec screen shot
- CPU: 2.86 GHz (318 MHz x 9) 1.55 V
- memory: 318 MHz 2.5-4-4-7 1T 2.8 V
3dmark runs (highest scores):
- 6600 GT PCI-e 1.19 GHz/0.59 GHz
- 3Dmark01 24407
- 3Dmark03 9995
- 3Dmark05 4019

Results 02/13/05
ScienceMark 2.0 run at 2.85 GHz 1.55 V, memory 317 MHz 2.5-4-4-8 1T 2.8 V
ScienceMark 2.0 run at 2.85 GHz 1.55 V, memory 317 MHz 2.5-3-3-7 1T 2.8 V (membench only)
ScienceMark 2.0 run at 2.73 GHz 1.52 V, memory 303 MHz 2.5-3-3-7 1T 2.8 V
Details and screenshots:
ScienceMark 2.0 test results at 2.85 and 2.73 GHz

Results 02/10/05
Memtest86 boot at 3006 MHz 1.65 V, memory 334 MHz 3-5-5-10 1T 2.8 V
Memtest86 pass at 2970 MHz 1.62 V, memory 330 MHz 3-5-5-10 1T 2.8 V
Windows XP boot at 2.95 GHz 1.60 V (screen shots)
Sandra CPU run at 2.94 GHz 1.60 V (Integer Dhrystone 13564 MIPS) (screen shots)
Highest HTT = 360 MHz with CPU = 360 x 8 = 2880 MHz (screen shots)
Highest memory = 350 MHz 3-5-5-10 1T with HTT = 350 MHz, CPU = 350 x 8 = 2800 MHz (screen shots)
SuperPI run at 2.90 GHz 1.60 V, memory 322 MHz 3-5-5-10 1T 2.8 V
Sandra memory bandwidth (int buffered) score at 8217 MB/s, memory 318 MHz 2.5-4-4-8 1T 2.8 V (screen shots)
SuperPI 1M 30 sec, 3DMark01, ScienceMark 2.0 run at 2.85 GHz 1.55 V, memory 317 MHz 2.5-4-4-8 1T 2.8 V
SuperPI 32M, SuperPI 1M 30 sec, 3DMark01/03/05 at 2.85 GHz 1.55 V, memory 285 MHz (9:10) 2.5-3-3-7 1T 2.8 V (screen shots)
Prime95 run stably at 2.73 GHz 1.50-1.52 V, memory 303 MHz 2.5-3-3-7 1T 2.8 V, 23 C idle, 38 C load (11+ hours, user aborted)
Details and screenshots:
1. Screen shots (more testing)
2. How to determine an upper limit for CPU and memory

Results from first setup (02/07/05)
1. Highest Windows boot is 2.92 GHz (324 x 9), 1.58 V
Memory bus 324 MHz 3-4-4-10 1T (with memory_HTT_ratio 1:1), 2.9 V
2. Winchester 3000+ to 2.91 GHz (323 x 9), 1.55 V, 29 C light load
Memory bus to 323 MHz 2.5-4-4-10 1T (with memory_HTT_ratio 1:1), 2.9 V
- Sandra CPU (13354 Dhrystone MIPS)
- Everest memory test (memory read 8057 MB/s, latency 34 ns)
3. Winchester 3000+ to 2.86 GHz (318 x 9), 1.55 V
Memory bus to 318 MHz 2.5-4-4-10 1T (with memory_HTT_ratio 1:1), 2.9 V
- Sandra CPU (13150 Dhrystone MIPS)
- Sandra memory bandwidth (integer buffered raw bandwidth 8225 MB/s)
- SuperPI 1M at 30 s
4. Winchester 3000+ to 2.70 GHz (300 x 9), 1.52 V
Memory bus to 300 MHz 2.5-3-3-7 1T (with memory_HTT_ratio 1:1), 2.8 V
- Prime95 tested stably for 4+ hour and further testing is underway. Temperature 24 C idle, 41 load.
- UPDATE: Prime95 tested stably for 10+ hours. Temperature 24 C idle, 41 load (user aborted)
- SuperPI 1M at 32 s
- 3DMark 01 22572 (video card 6600 GT at stock 525/1050 MHz)
Screen shots (initial testing)


Low PR 90 nm 939 Winchester (Sept 2004)

Some overclocking scenarios for 939 Winchester

Nforce4 chipsets with PCI-e, SLI features
 
Last edited:
As shown in the picture, the board has a 24-pin power socket so requiring PSU with 24-pin connector. The board has three heat sinks for the Vcore regulator (in the middle) and one heat sink for the memory voltage regulator (on the right). There is also a 4 pin molex socket (left) to supplement current flow to high power video card(s).

nf4_939_lpud_mb3_b.JPG
 
Last edited:
The DFI LanParty UT Nforce4 Ultra-D is rev. R.A02 with bios dated 1/25/05. R.A02 includes a list of fixes after the pre-released version. It is the revision that one should get.

nf4_939_lpud_mb4_b.JPG
 
Last edited:
Hardware Setup


The original heat sink retention bracket (yellow) was removed, and replaced the one for the ThermalRight XP-90 (black).

nf4_939_lpud_removebracket_b.JPG
 
Last edited:
The CPU with the IHS and the XP-90 were installed.

nf4_939_lpud_mb_xp90_b.JPG
 
Last edited:
Everything were put in place. The system boots up and is running :).

nf4_939_lpud_running_b.JPG
 
Last edited:
The system is running and being tested with OS installed, with colorful fans :).

nf4_939_lpud_light_b.JPG
 
Last edited:
Installation issues

There were two scary problems during installation.

1. The two on-board power and reset switches were broken as shown in the picture. Those two switches are supposed to be very handy during testing for switching the system ON/OFF and resetting the system without connecting the power and reset pins to the power and reset buttons of a case.

nf4_939_lpud_mb_broken_b.JPG


The broken pieces fell out all over the places when the motherboard was taken out of the bag. These two buttons are probably not well built parts, they are tall and unprotected in the package. This indicates there is still QC problem, do not know how wide spread this is.

I worried at that time without them the system might not boot at all. Attempting to fix and assemble them back was unsuccessful. I took a chance and went out to get a CPU and hoped that those two button switches would not be needed for system boot and reset.

After installation, the system boots fine as I had thought that they are really not necessary. I use a pen tip to short the power and reset pins for system boot and reset during testing.


2. When the CPU fan is connected to the CPU fan header, the bios has an auto-speed control that the fan would not spin if CPU temperature is below certain temperature (say < 25 C). When the system was first started, during posting the CPU fan span momentarily and stopped while the GPU fan and maybe also the chipset fan were still ON, I immediately shut down the system. In combination with the two broken switches, I thought the motherboard was DOA. After a while, I figured out that it was due to the auto-fan control feature, and actually the system posted and booted up without problem (even with the auto-fan control). This is a non-issue but rather an auto-fan feature.

After these two small and scary incidents, the system boots fine and has been stable without any problem.


About the CPU auto-fan control

When the CPU fan is connected to the CPU fan header and the motherboard controls the fan speed automatically based on CPU temperature. It is good this way for regular 24/7 usage as fan noise can be cut down significantly and I find that the system is much quieter unless the system is used for Prime95 the whole day :).

It is actually a non-issue but rather an auto-fan speed control feature of the motherboard. It may be scary at the very first time since during posting the CPU fan stops after spinning momentarily if the CPU is below certain temperature (say < 25 C) at cold start. Chipset fan has that auto speed control feature too.

To bypass such feature, the fan can be connected directly to a molex connector and the fan speed wire (single wire) to the CPU fan header for fan speed monitoring. There should/may be a bios option to disable the auto fan speed control.
 
Last edited:
3. Power connector

The Nforce4 939 board requires 24 pin power connector. I attempted to use existing 20-pin PSU to test out the system first instead spending $100+ for a Fortron Blue Storm 500W/Antec Neeo Power 480W.

During testing, a Fortron 350W which has a 20-pin power connector has been used to test the system continuously. By connecting the 20-pin to the on board 24-pin socket, as well as the 4 pin 12 V connector, the system has been running without problem. With this testing setup with 1 HD, 1 optical drive, 1 video card 6600 GT at rated speed, 320 MHz memory frequency and 2.9 GHz Winchester overclocking have been achieved. Eventually, a 24-pin PSU with higher 12V current should be used for 24/7 full system setup.


Added July 24, 2005:

From my experience with the DFI LP NF4 Ultra-D, the 4-pin power connector that is next to the main 24-pin power connector is needed to boot. The 4-pin connector supplies the 12V to power the CPU.

As of July 2005, I am still running an old Antec True 550 (24A on 12V) with only 20-pin power connector to power the DFI NF4 system to 310+ MHz for memory and HTT, and 2.8+ GHz for CPU for almost 6 months.

IMO, contrary to most would believe, the 24-pin connector may not really be "necessary" to high CPU and video card overclocking provided that the PSU has sufficient rating on the 12V and the video card has a separate 12V PCI-e power connection, as the CPU and the video card are each drawing from a separate 12V power connector (not from the 24-pin). The 20-pin connector already has 3 pins for 3.3V, 4 pins for 5V and 1 pin for 12V. That 4 pins add one more line for each of the GND, 3.3, 5 and 12V, such as for memory (3.3V or 5V boost).


nf4_939_lpud_powerconnect_b.JPG
 
Last edited:
Improved results (02/10/2005)

Booted into Windows at 2.95 GHz, memory 328 MHz 10-5-5-3 1T, ...

lp_ultra-d_winnie3000_cbbhd_328x9_mem_328_2.95GHz.JPG


nf4_939_lpud_cbbhd_328x9_2.95G_everest_read.JPG



Sandra CPU, Everest bandwidth, latency at 2.94 GHz, integer score 13564

lp_ultra-d_winnie3000_cbbhd_327x9_mem_327_2.94GHz_sandra_everest.JPG


lp_ultra-d_winnie3000_cbbhd_327x9_mem_327_2.94GHz_sandra_everest_2.JPG



SuperPI at 2.90 MHz

nf4_939_lpud_cbbhd_322x9_2.90G_superpi_sandra.JPG



3DMark01 and SuperPI 1M 30 sec at 2.85 GHz (317x9) memory 317 MHz 2.5-4-4-8 1T
6600 GT at stock 525/1050

nf4_939_lpud_cbbhd_317x9_2.85G_superpi_3dmark01_sandra.JPG



CPU = 2.85 GHz (317 x 9)
memory = 285 MHz (ratio 9:10, CPU/10) 2.5-3-3-7 1T 2.8 V
6600 GT overclocked at 600/1200 (core/memory)

3DMark01 and SuperPI 32M

lp_ultra-d_winnie3000_cbbhd_317x9_mem_285_superpi32M_sandra_3dmark01.JPG


3DMark01 and SuperPI 1M 30 sec

lp_ultra-d_winnie3000_cbbhd_317x9_mem_285_superpi1M_sandra_3dmark01_2.JPG


lp_ultra-d_winnie3000_cbbhd_317x9_mem_285_superpi_sandra_3dmark01_24k.JPG



Prime95 tested stably at 2.73 GHz for 11+ hours (user aborted), temperature 23 C idle, 38 load
6600 GT at rated 525/1050 MHz

nf4_939_lpud_cbbhd_303x9_prime95_3dmark01_sandra.JPG
 
Last edited:
First setup test results (02/07/2005)

Highest Windows boot is 2.92 GHz, 1.58 V
HTT 324 MHz
Memory bus 324 MHz 3-4-4-10 1T

nf4_939_lpud_cbbhd_324x9_2.92G.JPG


Winchester 3000+ to 2.91 GHz on air, 1.55 V, 29 C light load
HTT 323 MHz
Memory bus to 323 MHz 2.5-4-4-10 1T (with memory_HTT_ratio 1:1)

- Sandra CPU (13354 Dhrystone MIPS),
- Everest memory test (memory read 8057 MB/s, latency 34 ns)

nf4_939_lpud_cbbhd_323x9_2.91G_everest_read_sandra.JPG



nf4_939_lpud_cbbhd_323x9_2.91G_everest_latency_sandra.JPG



2.86 GHz on air, 1.55 V
HTT 318 MHz
Memory bus to 318 MHz 2.5-4-4-10 1T (with memory_HTT_ratio 1:1)

- Sandra CPU (13150 Dhrystone MIPS),
- Sandra memory bandwidth (integer buffered raw bandwidth 8225 MB/s)
- SuperPI 1M at 30 s

nf4_939_lpud_cbbhd_318x9_superpi_sandra_30s.JPG



Prime95 tested stably at 2.70 GHz for 4+ hour. Temperature 24 C idle, 41 load.
3DMark 01 22572 (video card 6600 GT at stock 525/1050 MHz)

nf4_939_lpud_cbbhd_300x9_3dmark01_sandra_prime95_stop.JPG


nf4_939_lpud_cbbhd_300x9_prime95.JPG
 
Last edited:
Back