Upgrade to Athlon 64 "Venice" - DFI NF4 Ultra D & 6200PCX
 

With AMD getting so much favorable attention in recent months, the only way to find out rather than reading all the reviews is to have one for yourself. Not that I no longer like the Northwood; with 64-bit OS on the horizon it may be time to acquire a 64-bit system which has and continues to receive fantastic comments from enthusiasts all over the world

http://www.madshrimps.be/gotoartik.php?articID=322

A couple of unique features on DFI NF4 that I have come to enjoy; 1) the fan controller up to three fans and 2) Monitor program.

I don't know if all A64s are the same; the Venice temp drops instantly right after stress program or game. The combination makes it a quiet system.

Interesting memory temperatures -

The setup using SF609 with panel close, SF609 is powered by an external PSU.


Temp sensors taped on memory (1) top and (2) bottom.


Temp readings with and without active cooling on (1) top memory





Temp reading on (2)





The memory close to Heatsink runs cooler because it benefits from the exhaust air movement, while the memory on top runs hotter due to lack of air movement.

I always thought the hot air coming from the heatsink would make the memory close by hotter, in fact it lowers the temp. The room temp this early in the morning is only 20° C; I am sure the temp will go higher with higher ambient temp. The mem volt is 2.9 and running at 300 HTT.

I will do more when I could get my hands on more memories. Surprisingly, I was unable to run Primestable without active cooling unless I lower HTT to 270.

nice temp drops :)

Another nice story to write about ....

When I tried to hook up my new LCD display using DVI ... nothing happens .... VGA works fine, but DVI has no input signal to display or no output signal from graphic card.

Hook up the new display to another system; DVI works fine. Therefore, it is not the display fault. switching graphic cards also works with both DVI and VGA .... so, I was thinking the graphic card was bad (no output signal from DVI).

Since I hate returning thing ..... I switched the PCIe slot from the top one to the second one on the DFI board ....... everything works. I must have missed the instruction as to which the slot a single graphic card should be used. I found nothing.

My BH5@ 3.2V is about 25-30°C with active 80mm cooling on top + good airflow in case.

What i am trying to do is make an airflow behind the motherboard to chill down the proc on the backplate since its gets very hot!(70°C at stock)
But that something i am going to do after my schooltime.

owh and about the review.
The psu rails on the first pic that seems to be low are not low if you measure it with a multimeter.
The smartgardian isn't the best reference for voltlines.

take these settings they are much better http://www.dfi-street.com/forum/show...highlight=MBM5

Assuming room temp 23°C with only 5°C above inside the case (panel closed) the base line already = 28°C to start with.

25-30°C temp is darn EXCELLENT !! :)

Regarding MBM5 and Smartguadian ...... as stated in the review if the same measurement is being used on two PSUs, it really doesn't matter if the measurement is running higher or low.:)

Apple to apple measurement .... that's why multi meter is used again for comparison. Regardless, the voltage rails were lowered in one of the PSUs, disregarding low/high versus bad/good.

well if you want almost exact temperature measurement you better be off with the mbm settings then smartgardian.

anyway my memory temp is if my fans are fully turned on (roomtemp 20-25°C)

The DVI connection versus the two PCIe slots was the point I wanted to make, so that users encounter similar issue be aware that it may not be the graphic card or Display fault. :D

MBM5 would not run with NF4 chipset, hence Smartguardian which correspond to BIOS and Speedfan :)

Frankly, none of them is accurate .... should the sensors are using computations.

MBM has the same values as my measured results with my mulitmeter.

and check a few post above since mbm does work with dfi nf4 ;)

NF4 & NF3 Data File
http://premium1.uploadit.org/soundx98//MBMDataFinal.rar

Thank you for your straight foward approach - Now, I know and will try it out;)


I cannot guarantee that to my readers unfortunately; it may well work for your particular setup.:)

Owh well i thought you saw my post with my link with mbm setup files :D

Well you got it now :).


About mbm and multimeter: its not only measured on my setup and confirmed but also confirmed by a few xs members

Also found a funny layout for mbm http://www.dfi-street.com/forum/show...&postcount=226 :D

Just did everything using the NF4 Data file...

With the same multi meter

Sensor 1 = 3.3
Sensor 2 = 5
Sensor 4 = 12

The 12 volt rail is dead on ... 3.3 and 5 Volts are lower in MBM5 than volt meter.

MBM Meter

3.20 3.31
4.91 5.02
12.01 12.01

Ran Prime95 together with 1 cycle 3Dmark01 - taken while the programs were running mid way.

I'd say MBM is better on the 12 Volt, the remaining two rails are the same as reported on Smartguardian.

Again, measurement by computation is good but there are too many variables for the fine tunings. The use of which to determine whether a power supply is good or otherwise will not be fair to the manufacturers and users alike.

Good tips and thanks .... as I like to use the familiar MBM5. =)

Sensor 4 shows 11.46 Volt ..... so I take sensor 3 which is closer.

yes you are right as my smartgardian displays the same voltrails but only the 12V is different.
And thats still the most important line.

good thing i have some multimeters in here and some pt100's :)
You can't be 100% sure.

Anyone uses DVI connection with single graphic card on the top slot?

Somehow my board does not provide DVI signal using the top slot but works fine using the second slot - D-sub works fine in both.

i'll try tomorrow lazyman as my gf6600gt arrives.
Currently stuck with my s3 virge :p

The 6200PCX turns into 6600GT after I installed Riva Tuner 15.5 :)

http://service.futuremark.com/servle...jectId=8635890

The project id is not valid - please check the URL.

Date 2005-07-03 Project ID 8635890 ?

http://service.futuremark.com/compare?2k1=8635890

Doom3 timedemo demo1 2nd run 68.2

It does not change much just the "name" :)

Not bad comparing to the real 6600GT with Doom3 1024 MQ of 68 FPS

http://anandtech.com/video/showdoc.aspx?i=2295&p=16

"not bad" does not come close, it simply VERY impressive

from budget cheapo, no good card, to top of crop in the budget line!

68FPS = :o

nice lazyman, again a new good budget item to your hardwarelist :D

congratz

Indeed a budget system with strong performance and with no fuss cooling.

With the exception of the pair of GSkill did not perform to bar and poor customer service/support in the United States (8 weeks without any reply from Tech Support while customer service was trying to get the "connection").

However, this issue has been resolved across the Atlantic Ocean (not the Pacific Ocean), EU Support (Karato) offered the RMA with upgrade within three email exchanges after I submitted screen shot of the A64 tweaker. I took a "rain check".

So, the folks in the U.S. will have to bite the bullet until such time that tech support becomes more viable.

Forgot about this entirely - set SLI to auto in BIOS. DVI signal now goes to both 1st and 2nd PCIe. If alternate slots are to be used . ocing of graphic must be performed on each slot for single card useage.

And, the graphic peforms better with age -

PCX6200 not 6600GT

http://service.futuremark.com/servle...jectId=8676789

3DMark2001 Project Details
The project id is not valid - please check the URL.
?

Just did and works ?

System Configuration
Operating System Microsoft Windows XP

DirectX Version 9.0

Mobo Manufacturer DFI Corp,LTD
Mobo Model LP NF4 Series
GP Rates (Current/Available) N/A / N/A


CPU Unknown 2710 MHz

FSB 201 MHz
Memory 512 MB
Graphics Chipset NVIDIA GeForce 6600 GT
Driver Name NVIDIA GeForce 6600 GT
Driver Version 6.14.10.7772
Video Memory 128 MB

Program Version 3DMark2001 SE
Resolution 1024x768 32bit
Texture Format Compressed

FSAA Disabled
Z-Buffer Depth 24bit
Frame Buffer Double
Rendering Pipeline D3D Pure Hardware T&L

Detailed Test Results

3DMark Score 19175 3D marks

Game 1 Car Chase - Low Detail 292.0 FPS
Game 1 Car Chase - High Detail 122.8 FPS
Game 2 Dragothic - Low Detail 294.7 FPS
Game 2 Dragothic - High Detail 168.2 FPS
Game 3 Lobby - Low Detail 297.0 FPS
Game 3 Lobby - High Detail 147.7 FPS
Game 4 Nature 78.3 FPS

Fill Rate (Single-Texturing) 1224.3 MTexels/s
Fill Rate (Multi-Texturing) 3870.5 MTexels/s
High Polygon Count (1 light) 95.6 MTriangels/s
High Polygon Count (8 lights) 19.9 MTriangels/s
Environment Bump Mapping 384.2 FPS
DOT3 Bump Mapping 211.7 FPS
Vertex Shader 170.8 FPS
Pixel Shader 311.3 FPS
Advanced Pixel Shader 248.2 FPS

Point Sprite 29.4 MSprites/s
---------------------------------------------------

3DMARK03 = 7401 :)

I also installed the Pheonix 70 using P4/A64 retention bracket; CPU temp dropped 4°C. I was toying with it just now, the house is so damn hot because I was out the whole evening with A/C off. @29°C room, it is running below 51°C Prime95.

My Venice doesn't like anything over 55°C when OC'ed. I have to use CV and attached the memory cooling fan over the PWMIC to keep it below 50°C. With DFI auto fan setting feature, the noise level is acceptable, it runs a bit louder when stressing the CPU.

I stated originally that I had problem with SATA setup; then, I broke the SATA connection. I finally RMA'ed the "Hitachi" a.k.a. IBM drive two weeks ago.

As it turned out the drive was bad. An upgraded replacement drive arrived from Thailand in a record time (total turn around time was 2 weeks).

May be I have four "blue envelopes" (litigation from IBM drives) requesting plaintiff information that the new owner Hitachi is fast in replacing bad drives :)

nice servive :)

I like to share with you some of the updates since I acquired the system back in May.

Summary:
1) The SATA drive having problem with Windows install was due to a bad drive and no fault to the installer;) nor the board. The drive was RMA'ed and a new drive came without any problem.
2) GSkill memory was also RMA'ed recently; the replacement sticks come with heatspreader. They performs as advertised this time and better.
3) I acquired additional two pairs of GSkill 4x512 at a basement price.

Other than the 2x256; the others 4x512 are of the same SKU, the 1GB LE. While one pair performs better than other. I am sure you know not all memory modules are created equal :)



Of course, hardcores and gamers know how not so well heatspreader performs. A closer look at the contact area covering 1/4th the size of the memory can't be effective in heat dissipation



- 2x256 LC replacement reaches 300HTT 1T 2.8V 2.5-4-3-8
- 2x512 LE(1) reaches 300HTT 1T 2.8V 2.5-3-3-6
- 2x512 LE(2) reaches 300HTT 1T 2.8V 2.5-4-3-8

As for DFI NF4 board using 2x512MB double sided modules, I ran SuperPi and played a game of BF2.

In order to boot the system using 4x512, 2T setting must be done first otherwise I don't even see the boot screen.

There is hardly any noticeable performance hit using divider and 2T; less than half a second.



BF2 - I was unable to use High settings in Texture and Lighting (all others are at high) when 1GB was used, while using 2GB everything runs smoothly at high setting.

Doom3 has no effect on the amount(size) of memory used, nor does it care for 1T/2T setting.

Bottom line-

1) 2GB 2T 4x512mb do not give up much if any performance hit, when HTT/FSB runs high enough.
2) 2GB is good if BF2 is the only game you play.
3) Buy only from reputable resellers with excellent customer service support; and don't count on the manufacturer support at times. Large resellers will have the "purchasing power" to "collect" from the manufacturers.
4) As more people are rushing into the 2GB, 1GB pair would be more abundant in the used market for a good price.

I also have "pencil in" the SLI bridge on the NF4 chipset making it SLI ready. As advised, there is no gain in price/performance ratio using two cards.

If 4x512 will generate more heat, here is my solution.

Using the 4 heatpipes AMD stock cooler
Antec Sonata II with med exhaust fan setting
CPU fan maxed out ~2,800 rpm (auto mode) and CV fan at about the same.
Room temp ~21°C



Sound level 42.5 dBA from 70 cm front.

good temps; although the noise level might be too high for some

Perhaps by spending $10 on insulating material will reduce the noise ;)

sorry for the unsteady hand in taking the picture this morning:D

39 dBA from about 6 inches away, rear exhaust set to low. How can I complain when white noise is 37 dBA? This is when I was running DVD movie titled "Over There".

- DVD ran from HD and volume off -