Dissappointing Winchester o/c

H

Harrath

Limp Gawd
Joined
Jan 18, 2003
Messages
209
My 3000+ only goes to 235x9, 2115 MHz on stock voltage with an XP-90. Voltage increments do nothing for the o/c. This is on an Epox 9NDA3+.. It's not the board or the memory holding it back, and the temps are great. Oh well, almost a 3500+ for $180. Can't beat that.
 
Darth_Fluffy said:
EPOX? Kill it with lasers. You need a real board, bro.
Click to expand...
what do you haev against epox? they make some solid boards. i recall one of their nforce2 boards being an enthusiast favorite
 
Ru sure its not the memory? Because ram that ran a certain way in say a XP board will NOT run that way in a a64 board because of the onboard ram controller which requires slightly looser timings to achieve same speeds! Mine went wacko after 220 mhz and i had to use the divider to drop it alittle so the cpu runs at 240.
 
EPOX builds great boards, i don't think that's the issue.

It sounds like the ram could be to blame. loosen timings and possibly try a memory divider,

Either way that's a 300mhz overclock, which seems about average for an A64.
 
what's the hypertransport multiplier set to?
http://www.anandtech.com/mb/showdoc.aspx?i=2253&p=13
9NDA3+ had some quirks, but it does overclock quite well according to anadtech.

You wouldn't happen to be running 4 sticks of ram would you?



The last area is memory performance and here, the results are clearly disappointing. The Epox is a fine board with 2 DIMMs, but when we tested with 4 DIMMs, the 9NDA3+ sets the memory speed at DDR333 no matter what we set in the BIOS. This is the first Socket 939 board we have tested that has been so restrictive in performance with 4 DIMMs. This limitation is so severe that we cannot really recommend the Epox if you plan to use 4 DIMMs now or in the future. Choose another Socket 939 board for four DIMMs. We suspect that Epox can fix this limitation in a BIOS update, but until we see this problem corrected, performance with 4 DIMMs will be a disappointment.
Click to expand...
 
Epox makes some of the best motherboards out there. The Epox 9NDA3+ has the highest overclock out of all the other s939 boards that have been reviewed on Anandtech. The Epox 8KDA3+ was also Anandtech's Gold Editors Choice in their socket 754 roundup.
 
Harrath said:
My 3000+ only goes to 235x9, 2115 MHz on stock voltage with an XP-90. Voltage increments do nothing for the o/c. This is on an Epox 9NDA3+.. It's not the board or the memory holding it back, and the temps are great. Oh well, almost a 3500+ for $180. Can't beat that.
Click to expand...

We need some more information to help you out here.

1.) What voltage, speed and latency is your memory set to?

2.) What is your HyperTransport multiplier set to? NForce 3 250's generally do not like to be run over 800.

3.) What voltage is your CPU set to, and what are your loaded/idle temps.

Here's how I generally get the CPU's maximum to determine my overclock:

1. Drop the HT to 2x or 3x.
2. Drop the memory speed from 400 to 200.
3. Up the voltage until the temperature hits where you are comfortable for your idle temps. Remember that temps will increase as your clock cycles do. Voltage alone does not dictate temperature. 1.6V is usually OK on air-cooling. 1.65V and above is pushing it.
4.) Boot Windows, and run Prime 95 Torture test in Maximum heat mode.
5.) Run your favorite video card torture test as well. Why? Because voltage drops affect your O/C, and your video card sucks up a lot of juice.
6.) Fire up the NForce System utility, and slowly up the FSB until Prime 95 craps out.
7.) Restart Prime 95, knock the FSB down two clicks, and write the CPU's frequency and temps down. That's your new target O/C number.
8.) Drop back into the BIOS, raise memory speed back up to 400, up the memory voltage to where you are comfortable with and set your target CAS latency.
9.) Boot Windows, Fire up Prime 95 with the memory torture test.
10.) Fire up the NForce System utility, and raise the FSB until Prime 95 craps out. It most likely will NOT hit the numbers it did before, but you did just find out how far your memory will O/C.

Now comes the tricky part. You have to essentially 'balance' the three variables (CPU speed, memory speed, and HT speed) to achieve your maximum overclock. Usually, this means dropping the HT speed a peg, memory speed down to 333, and cranking everything up so that each component is performing it's best relative to the other two. Your mileage will probably vary. You will have to play with the system a lot to find the 'sweet spot', but try not to get too comfortable with it. When all is said and done, I like to torture test my systems for at least 24 hours under full load. If it can do this, I consider it stable. A system should be able to go for a week at 100% load without crapping out.

FYI, I've had a 754 hit 2.63Ghz on Air-Cooling using this method. I had an old Opteron 146 engineering sample up to 2.42Ghz using this method as well. I've built quite a few of these over the last few months, and have yet to see one that wouldn't do at LEAST 2.3Ghz completely stable. This includes some very early samples and engineering samples. It's most likely NOT your motherboard, or your chipset, as I have personally had a SK8N NForce 3 150 (very early sample), up to 265 on the frontside, with a reduced HT bus.

Matt.
 
*applauds*

just do what enraged says. it may seem a bit complicated, but trust me, if you follow that, you should find the max oc of your stuff pretty well.
 
Harrath said:
My 3000+ only goes to 235x9, 2115 MHz on stock voltage with an XP-90. Voltage increments do nothing for the o/c.
Click to expand...

so have you tried more voltage or are you just assumming it dosnt help? you do have a pretty good Heatsink, try setting it to 1.65-1.7 and overclocking...

Also,
-what is you memory running at? what divider are you useing?
-Chipset voltage? did you up it?
-memory voltage? did you up it?
-LTD? did you drop it to 3x?



Overclcking is not feature...its a bonus.......
 
I'd be willing to bet it's the memory or mobo holding it back; my 3000+ runs stable at 2350mhz and would go higher but my ram isn't willing :(
(air cooling, xp-90
 

Excellent post enraged...
just had to throw in my kudos for that one... Write up a amd64 oc tutorial and see if you cant get it stickied/pinned
 
One other thing you can try just to eliminate the CPU's headroom from the equation, is drop the CPU multiplier down to 8x or 7x.

This will keep your cpu out of the mix for a little bit to see just how high you can get your HT bus and the memory bus.
 
Check the NB week. If it is before week 18, (taiwan) you might have problem of not being able to reach high FSB... if it is past that (korean) you might have other issues.
Before v-modding my Neo2, I couldn't get dual channel to work past 270FSB. Now it hums fine past 320FSB. Better chipset has shown result past 300FSB all over the net. When it didn't run dual channel past 270FSB, it still ran sigle channel well above 300FSB so try 1 stick to see if it is northbridge issue.
 
jinu117 said:
Check the NB week. If it is before week 18, (taiwan) you might have problem of not being able to reach high FSB... if it is past that (korean) you might have other issues.
Before v-modding my Neo2, I couldn't get dual channel to work past 270FSB. Now it hums fine past 320FSB. Better chipset has shown result past 300FSB all over the net. When it didn't run dual channel past 270FSB, it still ran sigle channel well above 300FSB so try 1 stick to see if it is northbridge issue.
Click to expand...
I have neo2 with K040908....is this ok?
MVC-587F.JPG

What other issues? Thanks.
 
Tried everything in enraged's post, 3x HT multiplier, lowest memory timings, lowest memory divider, 1.7v, won't even boot into windows at 240x9. Temps at 235x9 (highest I can get stable) at 1.6v never go above 45c. I think I just got a dud with this one.

Edit: Lowered CPU multiplier to 8, board goes to 270 no problem. Oh well.

Edit2: It seems (according to CPU-Z) that my core voltage fluctutates erratically. Anywhere from 1.510 - 1.534v when running stock. Could this be a source of troubles? Running an Antec Neopower 480W.
 
What exactly does HTT do? Also, are the Athlon64s multipliers only able to go up to it's default multipler or below? I have a 3000+ and don't have the option to set a multiplier past 9.
 
That voltage variation is pretty normal. My K8N will increase vcore while under load. So at idle if it's like 1.61, under load it will increase to 1.63-1.64.

However I have noticed that if do a cold boot, and then a warm boot, the voltages will drop lower after the warm boot by about .03-.04.
 
Memo said:
What exactly does HTT do? Also, are the Athlon64s multipliers only able to go up to it's default multipler or below? I have a 3000+ and don't have the option to set a multiplier past 9.
Click to expand...

Any Non-FX Athlon 64 is multiplier half locked. This means you can decrease the multiplier from the factory default, however you cannot go above that rated multiplier either. FX series processors are fully unlocked.
 
lithium726 said:
that is your southbridge. the memory is controlled by your processor, it has nothing to do with that chip.
Click to expand...
Yes i know that. I want to ask jinu117 b/cause he had problem b/fore with neo2/amd3500/mach1 combo. I want to run this neo2 with fx55 and vapo ls. My epox 9nda died...and Nforce3 ultra due for a new revision 2. b/cause rev.1 ridled with so many bugs...or is it?
 
centvalny said:
Yes i know that. I want to ask jinu117 b/cause he had problem b/fore with neo2/amd3500/mach1 combo. I want to run this neo2 with fx55 and vapo ls. My epox 9nda died...and Nforce3 ultra due for a new revision 2. b/cause rev.1 ridled with so many bugs...or is it?
Click to expand...
thats the one, i think you the newer ones, not sure check on http://xtremesystems.org/forums/index.php

should have sold your memory on ebay.... :cool:
 
jinu117 said:
Check the NB week. If it is before week 18, (taiwan) you might have problem of not being able to reach high FSB... if it is past that (korean) you might have other issues.
Before v-modding my Neo2, I couldn't get dual channel to work past 270FSB. Now it hums fine past 320FSB. Better chipset has shown result past 300FSB all over the net. When it didn't run dual channel past 270FSB, it still ran sigle channel well above 300FSB so try 1 stick to see if it is northbridge issue.
Click to expand...

Can we get a link for the V-Mod. Mine does 316 with dual channel, but I want 333 FSB (No good reason, I just wants it).
 
etruscan said:
Can we get a link for the V-Mod. Mine does 316 with dual channel, but I want 333 FSB (No good reason, I just wants it).
Click to expand...
With phase changer cooling it is possible.
 
Harrath said:
Tried everything in enraged's post, 3x HT multiplier, lowest memory timings, lowest memory divider, 1.7v, won't even boot into windows at 240x9. Temps at 235x9 (highest I can get stable) at 1.6v never go above 45c. I think I just got a dud with this one.

Edit: Lowered CPU multiplier to 8, board goes to 270 no problem. Oh well.

Edit2: It seems (according to CPU-Z) that my core voltage fluctutates erratically. Anywhere from 1.510 - 1.534v when running stock. Could this be a source of troubles? Running an Antec Neopower 480W.
Click to expand...

Do NOT consider voltage fluctuations to be a normal thing. I had a very similar problem with my Antec TruPower 400 supply in my FX-53. I was worried that my voltage fluctuations were limiting my overclock on the chip, so I ran a little test. I got a second power supply, shunted it open, and connected it to all of the devices excluding the main connector on the motherboard. Sure enough, my voltage fluctuations stabilized, and my overclock increased to 2.575Ghz with 100% stability. My max O/C increased to 2.75Ghz, but was not Prime stable. Check your 12 volt rail, and make sure that that's not fluctuating under load, either. You might want to try this dual power supply test for yourself, and find a bigger supply. Anytime you have more than a 10% deviation, you have a power problem.

Also, it is possible that you just might have a chip that doesn't want to do over 2.1Ghz, but I find that to be very unlikely. Your .02 volt fluctuations point to something else. Remeber, it's not the just the voltage that the chip needs, it's the amperage. Voltage fluctuations always indicate a drop in amperage as well, which will cause the chip to produce errors during Prime testing.

What you should be most concerned about is the amount that the voltage DROPS when overclocked at 1.625 or 1.65 volts. The CPU will be putting MUCH more stress on the power supply when overclocked at this voltage, as it will be using much more amperage as well. At stock speeds and voltages, a power problem will not be nearly as apparent, if at all.

You will also notice a previous poster had mentioned that his voltages actually INCREASE under load. This is the mark of a very good power supply. Some new supplies automatically adjust the rails as load increases. I suspect that this poster may in fact have one of these supplies.

I like to use a little analogy when teaching people about the principles of electicity.
Electricity is a lot like water in a pipe.
Consider voltage to be the pressure of the water in the pipe.
Consider amperage to be the AMOUNT of water you are moving.
Consider resistance to be the size of the pipe.
So, higher resistance and higher voltage will not flow more water, but less.
Less presure(voltage) at equal resistance willl give you less flow (amperage).

The theory doesn't translate exactly 100%, but does give you a better understanding of how voltage, amperage, and resistance are linked.

Matt.
 
lithium726 said:
um... a64's dont have northbridges.. and the memory controller is on the chip...
Click to expand...

Just because the memory controller is integrated into the processor itself does not mean you no longer have a northbridge.

The nForce chipsets only have a Northbridge while the VIA chipsets have both a Northbridge and a Southbridge.
 
burningrave101 said:
Just because the memory controller is integrated into the processor itself does not mean you no longer have a northbridge.

The nForce chipsets only have a Northbridge while the VIA chipsets have both a Northbridge and a Southbridge.
Click to expand...

? how does that work? i thought the chips got rid of the northbridge alltogether?
 
lithium726 said:
? how does that work? i thought the chips got rid of the northbridge alltogether?
Click to expand...

Nope, it just got rid of the memory controller on the Northbridge. The idea of a Northbridge and Southbridge is just something that was designed to cut costs. Instead of packaging all of your IC's on a high speed chip, you can package all of your high speed devices on one chip (AGP controller, Memory Controller (gone now...). Then you can package all of your low speed devices on another chip which doesn't have to be the same specs as the high speed. Thus, it's cheaper. I'm sure the manufacturer found that they could save five cents by doing this. As to whether one chip is better than two, I'd rather have the single chip solution. Less interconnects, less noise, less trace paths. But then I have to fork over five cents for that. Damn. It's not all good, however, as NVidia is finding with a big, fast, single chip that they are having signal issues with HyperTransport. Specifically, getting the NForce 3 over 800 (1600DDR) is proving to be an issue.

Matt.
 
Harrath said:
Tried everything in enraged's post, 3x HT multiplier, lowest memory timings, lowest memory divider, 1.7v, won't even boot into windows at 240x9. Temps at 235x9 (highest I can get stable) at 1.6v never go above 45c. I think I just got a dud with this one.

Edit: Lowered CPU multiplier to 8, board goes to 270 no problem. Oh well.

Edit2: It seems (according to CPU-Z) that my core voltage fluctutates erratically. Anywhere from 1.510 - 1.534v when running stock. Could this be a source of troubles? Running an Antec Neopower 480W.
Click to expand...


well there have been a few cases where the 9x multi is messed up and seems to be a defective board issue though so far .
 
Hey Matt,

Great post. It made me consider a few things and i'd like to pick your brain if you'd allow it. I am using a zippy/emacs 500w that's rated at >35a on all 3 rails... I am using it on my A8V .90 3000+ setup at home and have been noticing a few strange voltage issues. Firstly, my readings in windows show the same fluctuation between 1.61 and 1.63 volts, or if set to 1.65, 1.65 and 1.67v. My 12v rail in my BIOS screen reads a SOLID 12.05v and my 3.3v and 5v are dead on also (5v is slightly high.). In windows, however, my 12v rail is only reading 11.5v (!) to 11.6v, regardless of if i am at idle or full load. I've got 2 hardrives in the machine and a geforce 6800 NU (only 1 molex). I'm wondering if (a) i really need to go higher than a 120$ 500w psupply for this machine and (b) if these voltage fluctuations are really happening. Could it be that windows 11.5v reading is wrong? Should i read the same voltage in bios as in windows?

I'm interested in this because my box just won't stay prime stable at over 2380mhz even boosting voltage to 1.65v and i'm curious as to weather it might be based on a voltage issue. I can get the machine to boot at 2560mhz (!) but it is not prime stable at all at that speed. Adjusting HTT has no effect, memory is only running at 208 mhz or so at that point at 2-2-2-5-1t. Memtest passes with flying colors at 2380mhz and even at 2450+ but the cpu just isn't keeping it up...

enraged78 said:
Do NOT consider voltage fluctuations to be a normal thing. I had a very similar problem with my Antec TruPower 400 supply in my FX-53. I was worried that my voltage fluctuations were limiting my overclock on the chip, so I ran a little test. I got a second power supply, shunted it open, and connected it to all of the devices excluding the main connector on the motherboard. Sure enough, my voltage fluctuations stabilized, and my overclock increased to 2.575Ghz with 100% stability. My max O/C increased to 2.75Ghz, but was not Prime stable. Check your 12 volt rail, and make sure that that's not fluctuating under load, either. You might want to try this dual power supply test for yourself, and find a bigger supply. Anytime you have more than a 10% deviation, you have a power problem.
What you should be most concerned about is the amount that the voltage DROPS when overclocked at 1.625 or 1.65 volts. The CPU will be putting MUCH more stress on the power supply when overclocked at this voltage, as it will be using much more amperage as well. At stock speeds and voltages, a power problem will not be nearly as apparent, if at all.
Click to expand...
 
headless said:
Hey Matt,

Great post. It made me consider a few things and i'd like to pick your brain if you'd allow it. I am using a zippy/emacs 500w that's rated at >35a on all 3 rails... I am using it on my A8V .90 3000+ setup at home and have been noticing a few strange voltage issues. Firstly, my readings in windows show the same fluctuation between 1.61 and 1.63 volts, or if set to 1.65, 1.65 and 1.67v. My 12v rail in my BIOS screen reads a SOLID 12.05v and my 3.3v and 5v are dead on also (5v is slightly high.). In windows, however, my 12v rail is only reading 11.5v (!) to 11.6v, regardless of if i am at idle or full load. I've got 2 hardrives in the machine and a geforce 6800 NU (only 1 molex). I'm wondering if (a) i really need to go higher than a 120$ 500w psupply for this machine and (b) if these voltage fluctuations are really happening. Could it be that windows 11.5v reading is wrong? Should i read the same voltage in bios as in windows?

I'm interested in this because my box just won't stay prime stable at over 2380mhz even boosting voltage to 1.65v and i'm curious as to weather it might be based on a voltage issue. I can get the machine to boot at 2560mhz (!) but it is not prime stable at all at that speed. Adjusting HTT has no effect, memory is only running at 208 mhz or so at that point at 2-2-2-5-1t. Memtest passes with flying colors at 2380mhz and even at 2450+ but the cpu just isn't keeping it up...
Click to expand...

The difference might be whether or not you are under load, and whether or not Windows is correctly reading the same voltages that the BIOS is supplying. I've found that you really can't trust the voltage readings you are getting in MBM or some of the later programs. There is just too much difference between all of them. Personally, I like to take my VCore off of the motherboard with a good DVM, but that proves to be more of a pain in the ass than anything else. If you really want to find out if your power supply is up to snuff, then try running the test that I ran with a second power supply. Have the supply in the case run the motherboard, and the second shunted supply run the drives, fans, and video card. Record both your voltages, then swap the supplies and record voltages again. Try overclocking again with both configurations, and see what comes out of it.

Use the voltages and temperatures you are getting in the BIOS and Windows as basic indicators. Do not take them as gospel. Neither is 100% accurate.

Honestly, you have a pretty good supply, and your voltages look very good. Having a Vcore at 1.65 and having it creep up to 1.67 isn't bad at all. Having a VCore at 1.65 and having it droop to 1.61 is another story.

Matt.
 
You must log in or register to reply here.
Back
Top