Chaintech VNF4 Ultra Overclocking -- I figured out how to break the 220 barrier!!

[MajorPayne]

On almost every forum where people have the Chaintech VNF4 Ultra board, people report being unable to overclock past 220 in the BIOS (they can sometimes get to 250 in the N-Tune software within windows). This has been limiting to me until this morning -- I finally figured out how to get my board past this point!

Initially, I was able to get my board up to 220HT, but as soon as I went over this, it would not boot (it would default to a lower CPU multiplier and boot to a warning screen). It definitely was not heat-related, since I was running my processor at 21 degrees C. The one thing I was missing is that I was overloading the HTT bus itself, and I did not think to lower the HTT multiplier (which default is AUTO -- set at 5x). I think people miss this setting because it is in a totally separate part of the BIOS (advanced Chipeset Features - which is 3rd option from the top on the right).

Before lowering the HTT multiplier, I benchmarked in 3dmark03, 3dmark05, and sysmark -- I then lowered the HTT multi from 5x to 4x, and re-benchmarked, so I could see if there was any negative impact from lowering the multiplier. All of the scores were within 10 points of the originals, so the negative impact seems negligable.

After lowering to 4x, I was able to go up to 225, test some, then go to 230, test more (above 229, I had to change the memory stock clock to 133 from the 266 I had it at). I gained about 250 points in 3dmark 03 going up to 230 from 220. I then bumped to 235, and tested more, and then up to 240, then up to 245. I stopped at 245 this morning (since I had to leave, and did not want to overclock my system TOO much and leave it that way all day without me there if something started to fry). My temps were at 23 degrees C for the CPU, and 24 for the case (I have a stock cooler on my A64 3000+, and MANY casefans). I had been able to reach over 2.2 GHZ (with stock cooling, and UNCHANGED voltages), and I ran it through 3dmark 03 again without problems. I also used n-tune in windows to bump to 260, where it SEEMED stable (just to see if I could), and then moved it back before I left.

So it seems that the problems in OC'ing this board may be related to people not changing the HTT multiplier (or not changing it enough). And I know that the location of the HTT multiplier in BIOS makes it hard to associate it with overclocking, since every single other clock/voltage aspect of the board is in the Clock/Voltage section -- maybe Chaintech will change this with a later BIOS.

I was really glad to see this system OC so well (once I figured it out) since everything else about the board has been great. I had been worried that I bought a great board at stock settings, but a very poor board for OC'ing!

 

[rpmcrash]

Is it prime 95 stable and can you post cpu-z results and maybe a bench post from everest.
every picture has a story.

 

[GuitarDaddy]

3.4? You mean 2.4?

 

[MajorPayne]

Originally posted by: rpmcrash
Is it prime 95 stable and can you post cpu-z results and maybe a bench post from everest.
every picture has a story.

I had not had time to run Prime this morning (which is why I clocked it back before I left). I will try later tonight though, and grab some screens. What is a bench post from everest though? I do not think I have used that program before...

 

[MajorPayne]

Originally posted by: GuitarDaddy
3.4? You mean 2.4?

I don't think I put 3.4 in there anywhere? I looked over again, but I am still missing it. I never hit close to 3 GHZ tho, so if I did put this, it was definitely a mistake (I would LOVE to hit 3GHZ with a 90nm 3000+ and stock cooling!!!!).

 

[fishmonger12]

you didn't put 3.4 in here anywhere...

pics please

 

[rpmcrash]

Originally posted by: MajorPayne

Originally posted by: rpmcrash
Is it prime 95 stable and can you post cpu-z results and maybe a bench post from everest.
every picture has a story.

I had not had time to run Prime this morning (which is why I clocked it back before I left). I will try later tonight though, and grab some screens. What is a bench post from everest though? I do not think I have used that program before...

http://www.lavalys.com/product...p;lang=en&pageid=1
EVEREST Home Edition

Sorry about that. Do a bench in everest then show result post text.
I gave you a link i hope it works.

 

[MajorPayne]

Originally posted by: rpmcrash

Originally posted by: MajorPayne

Originally posted by: rpmcrash
Is it prime 95 stable and can you post cpu-z results and maybe a bench post from everest.
every picture has a story.

I had not had time to run Prime this morning (which is why I clocked it back before I left). I will try later tonight though, and grab some screens. What is a bench post from everest though? I do not think I have used that program before...

http://www.lavalys.com/product...p;lang=en&pageid=1
EVEREST Home Edition

Sorry about that. Do a bench in everest then show result post text.
I gave you a link i hope it works.

When I get home later tonight, I will benchmark and post some screens (I will get everest too, as I am curious now).

 

[MajorPayne]

I finally got a chance to do some REAL testing and benchmarking. I can only take the system to 240 stable currently (I can get it to 255, but I get errors in prime, and other places). At 240, I run 21-26 degrees C, at 2.6GHZ (up from a stock of 1.8GHZ), and my ram is running at 406MHZ (after being clocked down to 266 to get this far). I had to raise the voltage on my CPU and ram once to get here, and I had to lower the HTT multiplier as well.

I have hosted screenshots on my server to show where I got with my OC'ing. I may try for a better OC later down the road (I think I have to play with the voltages more). The first screenshot is from CPU-Z, and is at http://www.oregoncomputersolut...m/screenshots/CPUZ.jpg, and the Everest screenshot is at http://www.oregoncomputersolut...creenshots/Everest.jpg.

The system runs VERY Stable at this OC, and the temps are great (check out the temp numbers in the bottom right of my taskbar -- red is CPU, and yellow is Case).

By the way, if anyone needs information on getting Clockgen or Motherboard Monitor 5 to work for this board, I am happy to help, since I have both working correctly.

 

[rpmcrash]

That look about right i am running at 240 right now. ive been running at 240 for a week.
any thing after that is not stable games crashes windows locks up and htt is no help.
I have taken it to htt 1x. its going to need a BIOs fix.I want my o.cing in 1T so i can wait.
been through this before with the VNF-3.its a waiting game.

 

[MajorPayne]

Originally posted by: rpmcrash
That look about right i am running at 240 right now. ive been running at 240 for a week.
any thing after that is not stable games crashes windows locks up and htt is no help.
I have taken it to htt 1x. its going to need a BIOs fix.I want my o.cing in 1T so i can wait.
been through this before with the VNF-3.its a waiting game.

Interesting, I never had to go below Htt 3X to hit 240 (I am running a A64 3000+). I DID have to play with my voltages for CPU and RAM tho, so if you try that maybe you can get there at 3x if you up your voltages.

 

[rpmcrash]

what i mean is any thing over 240 is unstable and yes im running at Htt 3 and i have up the voltage.
PRIME 95 FAILED IN 2 HOURS.I SET IT TO 240 RAN PRIME 95 STABLE FOR 16 HOURS AND STILL RUNNING-
STRESS TEST. I would like to run in 2.6 and i will be happy. you may have a lucky board.

 

[MajorPayne]

Interestingly enough, when I got home tonight, I was able to get it to boot and run stable at 240 HTT with a HTT multi of 4X -- which made a big difference in 3d-mark 03 (100+ points). There seems to be little rhyme or reason as to when the mobo will let you boot with one setting, but not another. I will re-benchmark with the htt multi at 4x, and do a 12 hour run with prime to see if I get any errors. Interestingly enough, when I first got the board, I could not get it past 220 at all. Later, I got it past this to 240, but had to go down to 3x HTT multi to do it. Now I can run stable at 240 with a 4x multi! It almost seems like the board is getting more tolerant of OC'ing as I use it. Has anyone else seen this type of behavior from a motherboard before? Where it will not let you reach an OC initially, but after a few weeks use, it runs fine at that OC.

 

[Tab]

Holy crap! Thats a awesome speed increase.

 

[MajorPayne]

Originally posted by: Tabb
Holy crap! Thats a awesome speed increase.

Yeah, I have been VERY happy with this board. And I have been running this OC for 4 days now stable, without any errors/instability at all. My only disappointment so far has been that I had 2x512 sticks of kingston hyperx ddr3200, and I bought another 2 (since this chipset supports 4 channels of dual ddr) only to find that the chipset supports 4 channels, OR ddr at 400mhz, but not both (It will run 4 channels at 333 with a command rate of 2t, which I can OC back to 400mhz, but I cannot get it back to 1t, so the performance hit is not worth it). But this is not a limitation of the board, so much as a limitation of the chipset/specification. I have never been so happy with a mobo purchase before!

 

[xpose]

Good posts MajorPayne, you've helped me get to where I can get with overclocking on my mobo with an Athlon 64 3000+ Winchester on the same mobo. This is my first time overclocking and I have a few questions. So far I have been able to reach 2123.93 MHz , at 236 MHz.Here is a link to the info of my OC. However, I am unable to get any higher. Should I lower my memory even lower .. to 133mhz and pump the fsb to 270 to compensate for the lower memory speed? Do i need to increase the chipset voltage? Should i increase memory timings? I have ocz ddr400 memory with good timings (Cas Latency: 2-2-3-5 1T) and heat spreaders on them, so Im not so sure memory is the limiting factor. Any ideas how to get closer to 2.6gz?

 

[MajorPayne]

Originally posted by: xpose
Good posts MajorPayne, you've helped me get to where I can get with overclocking on my mobo with an Athlon 64 3000+ Winchester on the same mobo. This is my first time overclocking and I have a few questions. So far I have been able to reach 2123.93 MHz , at 236 MHz.Here is a link to the info of my OC. However, I am unable to get any higher. Should I lower my memory even lower .. to 133mhz and pump the fsb to 270 to compensate for the lower memory speed? Do i need to increase the chipset voltage? Should i increase memory timings? I have ocz ddr400 memory with good timings (Cas Latency: 2-2-3-5 1T) and heat spreaders on them, so Im not so sure memory is the limiting factor. Any ideas how to get closer to 2.6gz?

Glad I could help! And congrats on a good (and cheap) mobo too! As far as pushing this board further, I would definitely try loosening the memory timings first, just to eliminate them as the possible stop point of your OC. I would also check where your HTT multi is at (in the advanced chipset section of the BIOS). I had to lower to 3x in order to reach the 240 HTT that I topped out at (2.16GHZ with the same CPU). As far as your voltages go, I would check what the RAM is rated for, and not push more than .1V over that (RAM can fry easily if you are not conservative OC'ing it), and then run memtest (it can be found here: www.memtest86.com) to make sure your RAM does not give errors -- if it does, then back the voltage down.

As for CPU voltage, since you have an A64 3000+ winchester core like I do, the stock voltage is 1.4V. I actually pushed my core up to 1.7V for a brief time, just to see if this would let me OC further (It did not). You should be able to push safely up to 1.5 or 1.6 with just stock cooling (which is what I have) without too much of a problem.

The issue that I have is that I have tried ALL of these tricks, and have not been able to get past 240HTT, when I am confidant that the board and CPU are capable of this. The only variable I have left to try (as far as I know) is to up the Chipset voltage a little. I have been told that increasing this as little as .1V will give a huge boost in stability for overclocking, but I have not tried this yet (because I was worried about the passive cooling of the Nforce chip on this board). I will be trying this tonight (hopefully I do not damage anything), and I will let you know if this helps me pass 240.

 

[smackoww]

please show me how you got mbm to work. also what is clockgen? for OC? sure, explain that too

i need mbm to see my temps . the included digidoc program for the chaintech doesnt work.
can you also tell me what ram timings i should have stock, and what i should change when i OC?
thanks!

 

[The Sauce]

What is HTT?

Secondly regarding:

My temps were at 23 degrees C for the CPU, and 24 for the case

Your CPU can not be cooler than the air that is cooling it. Your temps are wrong.

 

[xpose]

Originally posted by: MajorPayne

Originally posted by: xpose
Good posts MajorPayne, you've helped me get to where I can get with overclocking on my mobo with an Athlon 64 3000+ Winchester on the same mobo. This is my first time overclocking and I have a few questions. So far I have been able to reach 2123.93 MHz , at 236 MHz.Here is a link to the info of my OC. However, I am unable to get any higher. Should I lower my memory even lower .. to 133mhz and pump the fsb to 270 to compensate for the lower memory speed? Do i need to increase the chipset voltage? Should i increase memory timings? I have ocz ddr400 memory with good timings (Cas Latency: 2-2-3-5 1T) and heat spreaders on them, so Im not so sure memory is the limiting factor. Any ideas how to get closer to 2.6gz?

Glad I could help! And congrats on a good (and cheap) mobo too! As far as pushing this board further, I would definitely try loosening the memory timings first, just to eliminate them as the possible stop point of your OC. I would also check where your HTT multi is at (in the advanced chipset section of the BIOS). I had to lower to 3x in order to reach the 240 HTT that I topped out at (2.16GHZ with the same CPU). As far as your voltages go, I would check what the RAM is rated for, and not push more than .1V over that (RAM can fry easily if you are not conservative OC'ing it), and then run memtest (it can be found here: www.memtest86.com) to make sure your RAM does not give errors -- if it does, then back the voltage down.

As for CPU voltage, since you have an A64 3000+ winchester core like I do, the stock voltage is 1.4V. I actually pushed my core up to 1.7V for a brief time, just to see if this would let me OC further (It did not). You should be able to push safely up to 1.5 or 1.6 with just stock cooling (which is what I have) without too much of a problem.

The issue that I have is that I have tried ALL of these tricks, and have not been able to get past 240HTT, when I am confidant that the board and CPU are capable of this. The only variable I have left to try (as far as I know) is to up the Chipset voltage a little. I have been told that increasing this as little as .1V will give a huge boost in stability for overclocking, but I have not tried this yet (because I was worried about the passive cooling of the Nforce chip on this board). I will be trying this tonight (hopefully I do not damage anything), and I will let you know if this helps me pass 240.

Well, i too am stuck at 240HTT. There must be something we have to do to go further rather than changing the HTT Multi to 2x. I have my memory timings at their original values and it works fine, aside from dropping it to 166mhz. Although, I had to increase the chipset voltage slightly to reach 240HTT. I have tried a bunch of stuff and nothing has seemed to work either. If i happen to get past 240 i will let you know. Maybe this board just wont allow it?

 

[MajorPayne]

Originally posted by: Snatchface
What is HTT?

Secondly regarding:

My temps were at 23 degrees C for the CPU, and 24 for the case

Your CPU can not be cooler than the air that is cooling it. Your temps are wrong.

No, it has been confirmed with a set of temp probes, and many others with this board have reported similar (tested) temps of 1-2 degrees C lower for the CPU than the case temp. This may be because of the fact that the motherboard's chipset has a passive cooler, which radiates heat from it (this would be especially true if the temp sensor for the case temp is located near or at the chipset cooler).

Originally posted by: Snatchface
What is HTT?

HTT Stands for HyperTransport Technology (I may be wrong about the technology part), and it is roughly equivalent to the Frontside Bus of older AMD and Intel Chips. Just like FSB, HTT was found to be useful in overclocking processors. HTT is a little different than the FSB in that it is actually an I/O speed, versus an actual bus, but it has similar advantages (being able to run up the processor's speed without changing the processor's multiplier), and disadvantages (PCI speed goes up along with it necessitating a PCI-Lock during overclocking to prevent AGP and PCI cards from getting out of spec, and memory speed rises with the HTT speed as well, which also necessitates a divider).

The HTT multi which people talk about is the multiplier on the HTT's stock speed of 200MHZ. HTT has a maximum speed of 1000MHZ (older chipsets are only specd for 800, but many can still reach 1000), and the speed of the HTT multiplied by the HTT multiplier (hence the name) give you the final HTT speed. Speeds of over 1000 are possible, but not by much, and are rarely stable.

HTT speed (before multi is applied) times the processor multiplier give you the processor's final speed (for example, my A64 3000+ winchester has a stock multiplier of 9, which multiplied by the stock HTT speed of 200 results in a speed of 1800 MHZ. I have the HTT clocked up to 240, so the actual speed is 9*240, which gets me a processor speed of 2160 -- lowering the HTT multiplier does NOT lower the final processor speed, just the speed that the HTT itself runs at, and this does not noticeably reduce system performance unless it drops lower than about 600).

Hope this explains enough to answer your questions.

 

[MajorPayne]

Originally posted by: xpose
Well, i too am stuck at 240HTT. There must be something we have to do to go further rather than changing the HTT Multi to 2x. I have my memory timings at their original values and it works fine, aside from dropping it to 166mhz. Although, I had to increase the chipset voltage slightly to reach 240HTT. I have tried a bunch of stuff and nothing has seemed to work either. If i happen to get past 240 i will let you know. Maybe this board just wont allow it?

I have NOT yet tried to up my chipset voltage (I meant to last night, but did not get off work until almost midnight, and since I was out the door at 5:30am, I never had time), so I do not know if this will help. At this point, it seems like something is artificially holding me back, since everything is cool and stable at 240 HTT.

One thing I heard mentioned earlier in Anandtech's review of the SIS 755FX Chipset for the A64 is that the that board (not the VNF4, which as far as reviewers etc. are concerned does not exist!) has a poor stock clock generator built in, which will not let you past an effective speed of 233 (even though it appears to pass this, the CPU speed does not really change past this point). I am starting to wonder if this board has a similar situation -- I do not know much about motherboard clock generators, so I cannot say for sure, but it definitely seems that some kind of artificial limitation is preventing passing the 240HTT point. Hopefully this is not the case tho, since the clock generator is a physical part (that cannot be changed by a BIOS update) and this would relegate us permanently to the 240 point.

 

[The Sauce]

Originally posted by: MajorPayne

Originally posted by: Snatchface
What is HTT?

Secondly regarding:

My temps were at 23 degrees C for the CPU, and 24 for the case

Your CPU can not be cooler than the air that is cooling it. Your temps are wrong.

No, it has been confirmed with a set of temp probes, and many others with this board have reported similar (tested) temps of 1-2 degrees C lower for the CPU than the case temp. This may be because of the fact that the motherboard's chipset has a passive cooler, which radiates heat from it (this would be especially true if the temp sensor for the case temp is located near or at the chipset cooler).

Well then you would be measuring the chipset temps and not the case temp, no? I just always considered the case temp to be the ambient air inside of hte case, not any one particular component such as the video card or chipset.

Also, thanks mucho for the explanation. My last build was with an NF2 chipset so I am a HTT noob but it makes sense. I will be helping a friend build with the Chaintech mobo soon and that information will be very useful. I appreciate it.

 

[MajorPayne]

Originally posted by: Snatchface
Well then you would be measuring the chipset temps and not the case temp, no? I just always considered the case temp to be the ambient air inside of hte case, not any one particular component such as the video card or chipset.

Also, thanks mucho for the explanation. My last build was with an NF2 chipset so I am a HTT noob but it makes sense. I will be helping a friend build with the Chaintech mobo soon and that information will be very useful. I appreciate it.

Yes, you would be measuring at wherever the sensor is located, and if the thermal sensor is located at (or near) the chipset, then effectively you would be measuring that temp (although the BIOS calls it a case sensor). A true case sensor would hang in the middle of the case itself, and not be attached to the motherboard, but then it could not be built into the motherboard in order to measure this temp either. Any built in temp sensor can only read the temp on the board itself, and is subject to fluctuations of temps caused by board components around it.

And I am glad I can help understand the HTT bus.... It can be quite different from working with the old FSB, and the differences have tripped many people up (myself included!).

 

[urfage]

There is a BIOS update on the Chaintech website. No changelog or any documentation with it. The timestamp is dated 12/29 (the one that came w/ my board was dated 12/20). I've installed it, but there seems to be no improvment in overclocking.

-Bryce

EDIT: using clockgen I was able to get much higher clockspeeds than the BIOS would let me choose. MajorPayne discovered this, check out the thread here.