• 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.
Testing UTT and TCCD memory modules in Winchester and DFI NF4 Ultra-D setup

Hardware:
Winchester 3000+ (CBBHD 0447 UPCW)
XP 90, ~ 40 CFM fan (e.g. 90 mm Enermax)
DFI LP UT NF4 Ultra-D rev A02, bios 02/17/05
6600 GT overclocked to 1.19/0.59 GHz
Antec True 550W (using 20 pin connector + 4 pin 12 V connector)

Memory:
TCCD G. Skill PC4400 LE, 2 x 256 MB
TwinMOS UTT Speed Premium PC3200 (AA4T 44D week 0506(4)), 2 x 256 MB

No tweaking of the extended memory timing for both UTT and TCCD except tCL-tRCD-tRAS-tRP. Default values are used for the rest.

The extended timing in the bios is listed as follows (for both cases):
Ratio-Cmd-tCL-tRCD-tRAS-tRP-7-A-2-2-1-2-312-A-E-A-0-A-4-A-A-256-D-16-7-D
(A = Auto, E = Enable, D = Disable)


Table 1
CPU at 2772 MHz = 308.1 MHz x 9
UTT 2-2-2-5 1T 3.5 V, 252 MHz, 5:6 ratio, CPU / 11
TCCD 2.5-3-3-7 1T 2.9 V, 308 MHz, 1:1 ratio

................................... UTT ............ TCCD
3Dmark01 .......................... 23246/23371 .... 23349/23426
3Dmark03 .......................... 9971 ........... 9978
SuperPI 1M (sec) .................. 32 ............. 31
Everest Memory Read (MB/s) ........ 7155 ........... 8029
Everest Memory Write (MB/s) ....... 2277 ........... 2473
Everest Memory Latency (ns) ....... 37.3 ........... 35.5
Sandra CPU Integer MIPS ........... 12318 .......... 12311
Sandra Int Buff BW MB/s ........... 7323 (91%) ..... 7897 (80%)
SciMark2.0 MolecularDyn (sec) ..... 63.17 .......... 62.11
SciMark2.0 Primordia (sec) ........ 316.18 ......... 311.38

1.5-2-2-5 1T:
3Dmark01 23285
ScienceMark2.0 molecular dynamic 63.18 sec


Table 2
CPU at 2792 MHz = 310.3 MHz x 9
UTT 2-2-2-5 1T 3.5 V, 254 MHz, 5:6 ratio, CPU / 11
TCCD 2.5-3-3-7 1T 2.9 V, 310 MHz, 1:1 ratio

................................... UTT ............ TCCD
3Dmark01 .......................... 23409 .......... 23497
3Dmark03 .......................... 9951 ........... 9987
SuperPI 1M (sec) .................. 31 ............. 31
Everest Memory Read (MB/s) ........ 7149 ........... 8068
Everest Memory Write (MB/s) ....... 2282 ........... 2472
Everest Memory Latency (ns) ....... 37.3 ........... 36.1
Sandra CPU Integer MIPS ........... 12403 .......... 12402
Sandra Int Buff BW MB/s ........... 7387 (91%) ..... 7953 (80%)
SciMark2.0 MolecularDyn (sec) ..... 62.09 .......... 61.53
SciMark2.0 Primordia (sec) ........ 313.76 ......... 310.56


Table 3
CPU at 2799 MHz = 311.0 MHz x 9
UTT 1.5-2-2-5 1T 3.5 V, 255 MHz, 5:6 ratio, CPU / 11
TCCD 2.5-3-3-7 1T 2.9 V, 311 MHz, 1:1 ratio

................................... UTT ............ TCCD ........... TCCD (tRCD=4)
3Dmark01 .......................... 23396(?) ....... 23576 .......... 23217/23361 (to check these numbers)
3Dmark03 .......................... xxxx ........... 9980 ........... 9957
SuperPI 1M (sec) .................. 31 ............. 31 ............. 31
Everest Memory Read (MB/s) ........ 7220 ........... 8104 ........... 8012
Everest Memory Write (MB/s) ....... 2307 ........... 2493 ........... 2409
Everest Memory Latency (ns) ....... 37.0 ........... 35.1 ........... 36.7
Sandra CPU Integer MIPS ........... 12541 .......... 12434 .......... 12430
Sandra Int Buff BW MB/s ........... 7540 (93%) ..... 7987 (80%) ..... 7936 (80%)
SciMark2.0 MolecularDyn (sec) ..... 61.98 .......... 61.43 .......... 61.45
SciMark2.0 Primordia (sec) ........ 313.24 ......... 309.45 ......... 311.35


Other benchmarks, gaming scores may be added over time.


Summary of results:

Based on the runs of UTT at 252 MHz 2-2-2-5 1T and TCCD at 308 MHz 2.5-3-3-7 1T (22% higher frequency) in 3Dmark01, SuperPI, 3Dmark03, TCCD was about 100 point ahead in 3Dmark01 and 1 sec ahead in SuperPI 1M.

Also based on the runs of UTT at 255 MHz 1.5-2-2-5 1T and TCCD at 311 MHz 2.5-3-3-7 1T (22% higher frequency) in 3Dmark01, SuperPI, TCCD was about 100 point ahead in 3Dmark01 and tie in SuperPI 1M.

So within margin of error, by keeping the CPU, HTT, HT and video card frequencies the same, it is fair to say UTT at 2-2-2-5 1T ties with TCCD at 2.5-3-3-7 1T running at about 22% higher frequency, e.g. 255 MHz vs 311 MHz, based on the above program tests.

In another post (see link below) which estimated an upper bound for the tradeoff between memory frequency and latency:
In conjunction with the 30-42% for memory read of 1 to 8 burst, and the 33% typical based on analytical estimation by counting read access cycles (see link below), it is fair to establish that memory with 2.5-3-3-7 1T would need 25-30% higher bus frequency to break even with memory with 2-2-2-5 1T timing for memory performance in memory intensive applications.

From table 1, 2 and 3,
- 1 MHz of memory frequency at 2-2-2-5 is about 60-80 points in 3Dmark01.
- 1 cycle of tRCD is about 200-300 points in 3Dmark01 at 311 MHz

If motherboard and memory can run at 300+ MHz, TCCD would be more flexible to allow a wider range of frequency and timing for tweaking, with 2.8-2.9 V, from 2-2-2-5 1T at 200-220 MHz, to 2-3-2-6 1T at 240-250 MHz, to 2.5-3-3-7 1T at 280-310 MHz, to 2.5-4-3-8 1T at ~320 MHz, to 3-5-5-10 1T at ~350 MHz (not all TCCD can do this). And at such high frequency of about 20+% higher would tie the tight timing 2-2-2-5 1T of BH-5/UTT at 250-260 MHz.

If motherboard and memory can run only up to 250 - 300 MHz, and 3.3 - 3.5+ Vdimm is an option and is chosen to be used, then BH-5/UTT would be the choice (running at a ratio to HTT) as its tight timing 2-2-2-5 1T can catch up with the 20+% high frequency of TCCD running 2.5-3-3-7 1T.


But to run BH-5/UTT at 2-2-2-5 1T above 240 MHz, say 250 - 260 MHz, it would require 3.3+ V. I have some concerns about the long term reliability of using the 4V Vdimm option provided by the DFI NF4 boards due to the power/heat stress on the Vdimm regulator MOSFET (see few post down). For 24/7 usage for powering the BH-5/UTT, I would not use the 4V Vdimm option, but rather an addon memory booster to power the BH-5/UTT modules. A Vdimm booster may have the same problem depending on its exact circuitry, but at least it is not built into the board risking damaging the board.

But it is not clear whether the DFI NF4 boards would be compatible with Vdimm booster.

I have not used the memory booster, and I assume by using it, the current to the 3.3+ V BH-5/UTT would bypass the Vdimm regulator and hence would relieve the heat/power stress on the MOSFET and related components.



Memory frequency and latency tradeoff

How much frequency increase is needed to break-even with low latency
 
Last edited:
CPU at 2772 MHz = 308.1 MHz x 9
UTT 2-2-2-5 1T 3.5 V, 252 MHz, 5:6 ratio, CPU / 11

lp_ultra-d_winnie3000_cbbhd_308x9_mem_252_2-2-2-5_3dmark01_6600gt_1.19G_0.59G.JPG


lp_ultra-d_winnie3000_cbbhd_308x9_mem_252_2-2-2-5_3dmark03_6600gt_1.19G_0.59G.JPG


lp_ultra-d_winnie3000_cbbhd_308x9_mem_252_2-2-2-5_sisoft_superpi.JPG
 
Last edited:
CPU at 2772 MHz = 308.1 MHz x 9
TCCD 2.5-3-3-7 1T 2.9 V, 308 MHz, 1:1 ratio

lp_ultra-d_winnie3000_cbbhd_308x9_mem_308_2.5-3-3-7_3dmark01_6600gt_1.19G_0.59G.JPG


lp_ultra-d_winnie3000_cbbhd_308x9_mem_308_2.5-3-3-7_3dmark03_6600gt_1.19G_0.59G.JPG


lp_ultra-d_winnie3000_cbbhd_308x9_mem_308_2.5-3-3-7_everest_read.JPG


lp_ultra-d_winnie3000_cbbhd_308x9_mem_308_2.5-3-3-7_everest_write.JPG


lp_ultra-d_winnie3000_cbbhd_308x9_mem_308_2.5-3-3-7_everest_latency.JPG


lp_ultra-d_winnie3000_cbbhd_308x9_mem_308_2.5-3-3-7_superpi_sandra.JPG
 
Last edited:
CPU at 2792 MHz = 310.3 MHz x 9
UTT 2-2-2-5 1T 3.5 V, 254 MHz, 5:6 ratio, CPU / 11

lp_ultra-d_winnie3000_cbbhd_310x9_mem_254_2-2-2-5_3dmark01_6600gt_1.19G_0.59G.JPG


lp_ultra-d_winnie3000_cbbhd_310x9_mem_254_2-2-2-5_sisoft_superpi.JPG


nf4_939


nf4_939


nf4_939
 
Last edited:
CPU at 2792 MHz = 310.3 MHz x 9
TCCD 2.5-3-3-7 1T 2.9 V, 310 MHz, 1:1 ratio

lp_ultra-d_winnie3000_cbbhd_310x9_mem_310_2.5-3-3-7_3dmark01_6600gt_1.19G_0.59G.JPG


lp_ultra-d_winnie3000_cbbhd_310x9_mem_310_2.5-3-3-7_3dmark03_6600gt_1.19G_0.59G.JPG


lp_ultra-d_winnie3000_cbbhd_310x9_mem_310_2.5-3-3-7_superpi_sandra.JPG


nf4_939


nf4_939
 
Last edited:
CPU at 2799 MHz = 311.0 MHz x 9
TCCD 2.5-4-3-7 1T 2.8 V, 311 MHz, 1:1 ratio

nf4_939


nf4_939


nf4_939


nf4_939


nf4_939
 
Last edited:
CPU at 2799 MHz = 311.0 MHz x 9
UTT 1.5-2-2-5 1T 3.5 V, 255 MHz, 5:6 ratio, CPU / 11

lp_ultra-d_winnie3000_cbbhd_311x9_mem_255_1.5-2-2-5_3dmark01_6600gt_1.19G_0.59G.JPG


lp_ultra-d_winnie3000_cbbhd_311x9_mem_255_1.5-2-2-5_everest_read.JPG


lp_ultra-d_winnie3000_cbbhd_311x9_mem_255_1.5-2-2-5_everest_write.JPG


lp_ultra-d_winnie3000_cbbhd_311x9_mem_255_1.5-2-2-5_everest_latency.JPG


lp_ultra-d_winnie3000_cbbhd_311x9_mem_255_1.5-2-2-5_superpi_sandra.JPG
 
Last edited:
CPU at 2799 MHz = 311.0 MHz x 9
TCCD 2.5-3-3-7 1T 2.9 V, 311 MHz, 1:1 ratio

lp_ultra-d_winnie3000_cbbhd_311x9_mem_311_2.5-3-3-7_3dmark01_6600gt_1.19G_0.59G.JPG


lp_ultra-d_winnie3000_cbbhd_311x9_mem_311_2.5-3-3-7_3dmark03_6600gt_1.19G_0.59G.JPG


lp_ultra-d_winnie3000_cbbhd_311x9_mem_311_2.5-3-3-7_everest_read.JPG


lp_ultra-d_winnie3000_cbbhd_311x9_mem_311_2.5-3-3-7_everest_write.JPG


lp_ultra-d_winnie3000_cbbhd_311x9_mem_311_2.5-3-3-7_everest_latency.JPG


lp_ultra-d_winnie3000_cbbhd_311x9_mem_311_2.5-3-3-7_superpi_sandra.JPG
 
Last edited:
The 4V Vdimm option for running BH-5/UTT in DFI NF4 SLI/Ultra motherboards

If a DFI NF4 setup is capable of running very high HTT and memory frequencies, say above 300 MHz on HTT and memory, I think the TCCD such as G. Skill PC4400 LE is a better choice than the BH-5/UTT. Even though the tight timing of 2-2-2-5 1T of BH-5/UTT equates to about 20-25% of higher TCCD frequency (50-60 MHz at 250 MHz level), one thing that has to be considered is that it would generally require at least 3.3 - 3.5 V to run BH-5/UTT at 250 - 260 MHz 2-2-2-5 1T to tie TCCD 2.5-3-3-7 1T at 300 - 320 MHz using 2.8 - 2.9 V.
Testing UTT and TCCD memory modules in Winchester and DFI NF4 Ultra-D setup

One would think the DFI board would supply up to 4.0 V for Vdimm and would fit perfectly for running BH-5/UTT. BUT ....

To get 3.3 - 3.5 V on the DFI LP NF4 Ultra-D/SLI, it would require using the 5V jumper option, but this would put too much power and heat stress on the Vdimm voltage regulator. The 5V jumper mode was tested and the Vdimm voltage regulator MOSFET (with passive heatsink) was so hot to touch when the 5V jumper option was used to get the 3.3 - 4.0 V (equally hot also in this mode even for the lower 2.x - 3.2 V), whereas the default Vdimm mode (up to 3.2 V) is way lower in temperature.

Let's look at how much power/heat is dissipated on the Vdimm MOSFET:

To get 3.3 - 4.0 V using the 5V jumper, the difference between Vdimm and the 5V is dropped across or “absorbed” by the MOSFET, about 1.0 – 1.7 V. It is resulted in a large amount of power dissipated in the MOSFET and the number can be estimated as follows.

voltage drop on MOSFET = 1.0 - 1.7 V
nominal power per memory module = 10 - 15 W (single or double sided)
nominal voltage = 2.5 V
nominal current = 4 - 6 A
memory overclocked to 250 MHz from 200 MHz rated (25% increase)
overclocked current = 5 - 7.5 A
worst case power dissipated on MOSFET = 7.5 * 1.7 = 12.75 W

Without using the 5V Vdimm jumper option,
voltage drop on MOSFET = 0.1 - 0.5 V (Vdimm 2.8 - 3.2 V)
memory overclocked to 250 - 300 MHz from 200 MHz rated (25 - 50% increase)
overclocked current = 5 - 9 A (as before)
worst case power dissipated on MOSFET = 9 * 0.5 = 4.5 W (about 1/3 of the 4V Vdimm option)

When running the DFI NF4 board 24/7 using the 4V option for Vdimm, my concern (have not looked at the exact the 4V Vdimm circuit) is that it would put too much power/heat stress on that Vdimm voltage regulator MOSFET/components. IMO, the 4V Vdimm option could pose a long term reliability issue on the board for running BH-5/UTT memory, regardless the voltage level, whether it is the lower 2.x - 3.2 V or the higher 3.3 - 4 V.

Currently I would not run my system 24/7 under the 4V Vdimm mode. An alternative for getting the 3.3+ V for BH-5/UTT is to use an addon Vdimm booster. A Vdimm booster may have the same problem depending on its exact circuitry, but at least it is not built into the board risking damaging the board (especially for a board that can do 300 - 360 MHz on HTT and memory).

But it is not clear whether the DFI NF4 boards would be compatible with Vdimm booster.

I have not used the memory booster, and I assume by using it, the current to the 3.3+ V BH-5/UTT would bypass the Vdimm regulator and hence would relieve the heat/power stress on the on-board Vdimm regulator/MOSFET.
 
Last edited:
Has anyone had any luck breaking past fsb 255 on the A8Vpro mb?
The best I can do with Corsair VS 3200 2x512 is 209.
The fsb hits a wall at 255, so I'm using the HTT 600 (3x) option with memory at 5/3.
The Winchester 3000+ cpu can only be tested up to 2.29 due to the 9X multi and the 255wall.
My upper limits are based on Prime95 overnight stability.
Temps: no load/prime95= mb 26/28, cpu 28/38 using amd retail hs fan
Voltages: all default

I'm also working on an Abit754 NF8 mb, NewCastle 3000+.
The fsb wall seems to be at <240, and the cpu multi only allows 10X for 2.4GHz.
(The agp lock works, but the pci lock doesn't according to ClockGen)
Note: I also notice that even with CnQ disabled, it still starts at 50% cpu clock and 1.1Volts, which is the amount of the "thermal throttling" setting for suspend mode. Disabling "ther.throt." sets off the over-temp alarm at bootup and auto shut down.

I'm likin' the 20-25% oc, but I'm greedy and want higher cpu multi's or more fsb!
Must say you all have some enviable oc's.
 
Last edited:
Hitech - excellent post on the UTT / TCCD comparison. I am very interested in the results. I live in Canada and am desperately trying to locate a set of G.Skill LE's but am running out of luck. There are no G.Skill retailers in Canada, and there are no US retailers that will ship to the Canada. Thus, I may have to switch to a different set of ram. I was thinking of going w/ the 3200+ (for the 10x multi) and getting a set of the OCZ PC 4000VX sticks. With any luck I'll be able to run 260x10 or maybe even 270x10 @ 2-2-2-X. (DFI NF4 Ultra-D MB).

Your post on power / heat dissipation using the 5V jumper is interesting. Question - why would DFI include a 4V vdimm option if it led to a long term reliability issue? You would think the extra power dissipation would be accounted for in the design. This is interesting and definately deserves further research before I make a decision.

Can anybody confirm whether or not using the memory booster bypasses the Vdimm regulator?

-Adrayic
 
Lyser said:
Has anyone had any luck breaking past fsb 255 on the A8Vpro mb?
The best I can do with Corsair VS 3200 2x512 is 209.
The fsb hits a wall at 255, so I'm using the HTT 600 (3x) option with memory at 5/3.
The Winchester 3000+ cpu can only be tested up to 2.29 due to the 9X multi and the 255wall.
My upper limits are based on Prime95 overnight stability.
Temps: no load/prime95= mb 26/28, cpu 28/38 using amd retail hs fan
Voltages: all default

I'm also working on an Abit754 NF8 mb, NewCastle 3000+.
The fsb wall seems to be at <240, and the cpu multi only allows 10X for 2.4GHz.
(The agp lock works, but the pci lock doesn't according to ClockGen)
Note: I also notice that even with CnQ disabled, it still starts at 50% cpu clock and 1.1Volts, which is the amount of the "thermal throttling" setting for suspend mode. Disabling "ther.throt." sets off the over-temp alarm at bootup and auto shut down.

I'm likin' the 20-25% oc, but I'm greedy and want higher cpu multi's or more fsb!
Must say you all have some enviable oc's.

Please start your own thread - it is best to maintain topics within their own threads. Thanks, and welcome to the forums!

We will be happy to help you out when you start your own thread. :)
 
Adrayic said:
Hitech - excellent post on the UTT / TCCD comparison. I am very interested in the results. I live in Canada and am desperately trying to locate a set of G.Skill LE's but am running out of luck. There are no G.Skill retailers in Canada, and there are no US retailers that will ship to the Canada. Thus, I may have to switch to a different set of ram. I was thinking of going w/ the 3200+ (for the 10x multi) and getting a set of the OCZ PC 4000VX sticks. With any luck I'll be able to run 260x10 or maybe even 270x10 @ 2-2-2-X. (DFI NF4 Ultra-D MB).

Your post on power / heat dissipation using the 5V jumper is interesting. Question - why would DFI include a 4V vdimm option if it led to a long term reliability issue? You would think the extra power dissipation would be accounted for in the design. This is interesting and definately deserves further research before I make a decision.

Can anybody confirm whether or not using the memory booster bypasses the Vdimm regulator?

-Adrayic


It is not clear whether the memory booster can even work in the DFI LP NF4 SLI/Ultra-D boards.

If you look at some posts in the memory section, there are quite a number of posts about the failure of the OCZ memory booster. Many pointed out the failure was related to overheating of variable resistor in the Vdimm regulator, including an OCZ rep., ....
OCZ DDR Booster, it's dead, Jim.
 
SuperPI 32M is a relative "quick" way to test CPU and system speed and stability

SuperPI 32M is a relatively "quick" way to test speed and stability of CPU, memory and system. SuperPI 1M does not stress the CPU, memory and system enough. System that passes SuperPI 1M may have a hard time passing SuperPI 4M, 8M, 16M and 32M.

For SuperPI 32M, in additional to CPU frequency, memory frequency, memory timings affect the total time as there is heavy memory access throughout the computation. I think system paging is also involved (to some extent depending on total memory size of the system), so the chipset and the system bus are involved. Unlike the small FFT in Prime95 (8K - 32K), only the CPU and the caches are stressed. The large FFT (1024-4096K and blend) in Prime95 can test both CPU, memory and system (paging) also, and certain Prime95 setup can stress the CPU, memory and system more with extensive testing in time.

SuperPI 32M runs longer (20 - 30+ min. depending on CPU processing power), requires more memory (~ 268 MB compared to ~ 8.4 MB for SuperPI 1M) and system resources, and tests the CPU, memory, system and their stability more than that offered by SuperPI 1M which completes in a much shorter period of time (about 20 - 30 sec).

Stability testing using memtest, SuperPI (32M) and Prime95

The original X86 executable is dated back to 23/09/1995, and it is still being widely used and downloadable as SUPER_PI.EXE. Do a search will find sites for downloading it.


About SuperPI

Quoted from SUPER_PI.TXT of the SUPER_PI package:

"In August 1995, the calculation of pi up to 4,294,960,000 decimal digits was succeeded by using a supercomputer at the University of Tokyo. The program was written by D.Takahashi and he collaborated with Dr. Y.Kanada at the computer center, the University of Tokyo. This record should be the current world record. ( Details is shown in the windows help. ) This record-breaking program was ported to personal computer environment such as Windows NT and Windows 95. In order to calculate 33.55 million digits, it takes within 3 days with Pentium 90MHz, 40MB main memory and 340MB available storage. The software is free and the circulation of program is also free!"

Links to the very original source:
http://www.super-computing.org/
ftp://pi.super-computing.org/windows <- e.g. Windows user, download "super_pi.zip", unzip the package and run super_pi.exe
ftp://pi.super-computing.org/Mac_OSX
ftp://pi.super-computing.org/Linux
ftp://pi.super-computing.org/Itanium_Linux
ftp://pi.super-computing.org/AIX_Power3

- Use the above original (version 1.1) for general testing for speed and stability, to avoid potential problems of the "mod" version in certain setups and processors.
- Use the following "mod" version only for tweaking as it can show milli-second resolution or for validation if needed.


d]g[ts said:
http://superpi.radeonx.com/ <- Link taken from XS froum. original thread there linked HERE
this is a version that supports millisecond times, and can be validated via checksums. Also has anti cheat feature to where it cannot be patched.


This thread lists results about running SuperPI 32M from members.
SuperPI 32M for testing CPU and system speed and stability



CPU: Winchester 3000+ CBBHD 0447
memory: G. Skill 4400 LE 2 x 256 MB (Samsung TCCD)
motherboard: DFI LanParty UT NForce4 Ultra-D (rev. A02, bios 03/10/05)
cooling: XP-90, 80 mm Tornado with fan control
OS: Windows XP Professional SP1

- CPU: 2.83 GHz = 314 MHz x 9, 1.52 V
- memory: 314 MHz, 2.5-4-3-7 1T, 2.8 V
- SuperPI 32M completed in 27 min 42 sec

lp_ultra-d_winnie3000_cbbhd_314x9_mem_314_2.5-4-3-7_2.8V_superpi32M_27m42s.JPG



Winchester 3000+ CBBHD 0447
G. Skill 4400 LE 2 x 256 MB (Samsung TCCD)
DFI LP UT NF4 Ultra-D (rev A02, bios 02/17/05)

- CPU: 2.85 GHz = 317 MHz x 9, 1.55 V
- memory: 317 MHz, 2.5-4-4-8 1T, 2.8 V
- SuperPI 32M completed in 27 min 52 sec

lp_ultra-d_winnie3000_cbbhd_317x9_mem_317_2.5-4-4-8_2.8V_superpi32M.JPG
 
Last edited:
Got CPU to 2.86 GHz with SuperPI 32M completed.

Winchester 3000+ CBBHD 0447
G. Skill 4400 LE 2 x 256 MB (Samsung TCCD)
DFI LP UT NF4 Ultra-D (rev A02, bios 02/17/05)

- CPU: 2.86 GHz = 318 MHz x 9, 1.55 V
- memory: 318 MHz, 2.5-4-4-7 1T, 2.8 V
- SuperPI 32M completed in 27 min 55 sec

lp_ultra-d_winnie3000_cbbhd_318x9_mem_318_2.5-4-4-7_2.8V_superpi32M.JPG
 
Shuzzy said:
nice results there. if you dont mind me asking, where do you get that nvmonitor program?

That NVmonitor ultility program is part of the nVidia nTune program which came with the motherboard CD.

I do NOT use nTune for automatic overclocking and tuning. Everything is done manually using the bios.

The NVmonitor shows the CPU, board and GPU temperature, also the memory, HT bus, PCI-e bus, GPU/memory clock speed. The voltage reading of it is not correct.

The latest version of nTune can be download from the nVidia web site (nTune_2.00.23 as of 03/12/05).

I also updated the latest Nforce4 chipset drivers, nView and audio mixer. Currently the drivers, nView and the audio mixer are package as a single .exe file (nForce_6.39_WinXP2K_WHQL_english.exe as of 03/12/05).

I found there is no noticeable difference in overclocking and stability between the original driver that came with the CD and this latest version of the driver.
 
DRAM Bios Setting

Motherboard: DFI LP UT NF4 Ultra-D (rev. A02)
Winchester 3000+
G. Skill 4400 LE (TCCD)

The default bios setting are shown.
Those that are altered are shown by using the “->” sign.
The actual values of the AUTO settings are also shown.

Have not tweaked the settings for SuperPI, 3Dmark01/03/05, ScienceMark02, ....
The TCCD modules used allows for a wide range of memory frequency, from 200 to 320+ MHz, a tight timing tweak would not fit all frequency setting.
(Sensitivity numbers of individual setting on performance and stability will be detailed when available, see next post).

There is no noticeable difference in overclocking, stability and performance between the 01/25/05 bios (official) and the 02/17/05 bios (beta). Both bios work fine with TCCD and UTT memory modules.

Will try the latest official 03/10/05 bios with NVMM 4.85 and NV RAID 4.81, and memory compatibility enhancement.

Current Nforce4 drivers, nView and the audio mixer are from a single .exe file nForce_6.39_WinXP2K_WHQL_english.exe downloaded on 03/12/05.


Bios 02/17/05

DRAM Frequency Set ............... AUTO -> 200 MHz (DRAM/FSB: 1/1)
Command Per Clock (CPC) .......... AUTO -> Enabled
CAS Latency (Tcl) ................ AUTO -> 2.5
RAS to CAS Delay (Trcd) .......... AUTO -> 3 or 4 (for 313+ MHz)
Min RAS Active Time (Tras) ....... AUTO -> 7 or 8 (for 313+ MHz)
Row Precharge Time (Trp) ......... AUTO -> 3 or 4 (for 313+ MHz)
Row Cycle Time (Trc) ............. 7
Row Refresh Cycle Time (Trfc) .... AUTO (= 19)
Row to Row Delay (Trrd) .......... 2
Write Recovery Time (Twr) ........ 2
Write to Read Delay (Twtr) ....... 1
Read to Write Delay (Trwt) ....... 2
Refresh Period (Tref) ............ 3120 (= 200 MHz 15.6 us)
Write CAS Latency (Twcl) ......... AUTO (= 1)
Bank Interleave .................. Enabled
DQS Skew Control ................. AUTO (= OFF)
DQS Skew Value ................... 0
DRAM Drive Strength .............. AUTO (= Weak)
DRAM Data Drive Strength ......... Level 4
Max ASYNC Latency ................ AUTO (= 8 ns)
Read Preamble Time ............... AUTO (= 5 ns)
Idle Cycle Limit ................. 256
Dynamic Counter .................. Disabled
R/W Queue Bypass ................. 16x
Bypass Max ....................... 7x
32 Byte Granularity .............. Diabled (8 Bursts)



lp_ultra-d_winnie3000_cbbhd_313x9_mem_313_A64_Tweaker.JPG
 
Last edited:
Sensitivity of DRAM settings on performance and stability

To be completed, very preliminary results, please don't quote

The TCCD modules used allows for a wide range of memory frequency, from 200 MHz 2-2-2-5 1T to 320+ 2.5-4-4-8 1T, a tight timing tweak would not fit all frequency setting.

These settings have the most impact on performance

Tcl
Trcd
Trp .... 4 -> 3 ............... -0.25/31 sec SuperPI 1M
Trfc ... 19 -> 16 ............. -0.10/31 sec SuperPI 1M
Trrd ... 2 -> 1 ............... -0.05/31 sec SuperPI 1M
Tref ... 15.6 -> 3.9 us ....... +0.06/31 sec SuperPI IM

(Approximate average number quoted, for reference only).


These settings have the most impact on stability and booting
Tcl
Trcd: 3 up to 310 MHz, 4 for up to 320 MHz, 5 for up to 350 MHz (each module maybe different, for reference only)
Max Async Latency: 8 ns for 300+ MHz HTT, 7 ns may be OK


More sensitivity numbers on individual setting will be detailed when available.
 
Last edited:
Back