My folding system started life as a Biostar TForce 6100 and the cpu above. As with all new builds and the wonder of standardization (what voltage will memory need on this system?) I started with just 1 DIMM. When it started life it did 30 minutes per checkpoint running SMP under Unbuntu. Memory is Corsair DDR2-667 CAS 4, with the ever so popular 1 stick DOA on arrival. Another package later and I have both sticks running. This saved me about 75 seconds per checkpoint. Nowing this memory OCs well and the actual speed memory runs at with all the AMD divider calcs I set the memory as DDR2-800, and picked up another 75 seconds or so, give or take a few.
I stopped running the system monitor, maybe worth 15 seconds. Temp readings in the BIOS are absurd, so I bought one of the infra-red jobs. It reports, even at full load, the base of the CPU hs about 40C, and the northbridge HS about 60C. Southbridge was 60C as well, what the...? This was in a case with the side left open and zero air flow. So I went to work. Pulled the SB HS, which had a huge amount of white compound. I really didn't want to put a stick on HS here even though I have a much bigger one. I clean it and use Artic Silver 5. The NB HS is not that big, I replace it with a Zalman NBF-47, the aluminum peacock. I think it weighs twice as much as stock, though still feather light, and must have triple the surface area. I stick little HSs on other hot chips and close it. I run two Yate Loons as exhausts at full speed. A side cone on the case sends cool air straight to the CPU, which is still using the stock cooling. SB temps drop at least 8C, northbridge temps drop over 20C! CPU just sits there at 40-42C, with the stock AMD 70mm ultra thin fan.
So now I OC, 225MHz FSB and HT divider at 4. Memory is not happy so I up the voltage to the max BIOS of 2 volts and nothing. So then I use the jumper to set it to 2.2V and that does the job. After some "adjustments", memory setting back to 667 it's running, and I pick up another minute or so, maybe 1 minute 20 seconds per checkpoint. CPU hs is only a couple of degrees warmer, 2, maybe 3.
After the weekend of running like that, and the server burp that occurred yesterday I tried to OC some more (230MHz FSB) but no luck. I up the CPU voltage and nothing. I tried looser memory timings, from 4-4-4-12 to 5-5-5-15 and nothing. To pin things down I go back to 225MHz FSB and set memory to DDR2-800. That's fine and again I pick up another minute or so. Currently at 25 minutes 15 seconds per checkpoint.
While the system only uses 105 watts even at this setting I wonder if the PS is an issue, a 300 watt Seasonic of modest amps. And I really don't know what the cpu temp is, only that the cpu HS temp is and at the outer edge. Is there good thermal conduction? Are they flat? Are either one flat? I have a big Thermalright HR-01 I will try next.
Shaving time between checkpoints - My AMD X2 4000+
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
Since I really can not tell for sure from your description, I will not say that for certain a different power supply will not help, I strongly believe that to be the case. The Seasonic is a great supply, and I doubt that you are taxing it.
I have had two systems that I worked on need a bigger power supply. Both of these systems were difficult to start, as a typical desktop system needs a lot more power to spin up the drives than it does once things are running.
Both of these systems would start the drives spinning, then nothing. If you pushed the power supply button a few times, the drives would get faster and faster, and then finally, once the power draw from the drives was not there anymore, the rest of the system would get enough power, and boot and everything would be fine till the system was turned off.
Anyway, if your system starts up fine, then I think the capacity of the Seasonic is fine, and usually thier power regulation is very solid and if anything they tend to under-rate their power supplies compared to average brands.
If you have a bigger thermalright handy, easy enough to give it a try, but bet you a beer (if you are old enough
) that it will make no difference.
I have had two systems that I worked on need a bigger power supply. Both of these systems were difficult to start, as a typical desktop system needs a lot more power to spin up the drives than it does once things are running.
Both of these systems would start the drives spinning, then nothing. If you pushed the power supply button a few times, the drives would get faster and faster, and then finally, once the power draw from the drives was not there anymore, the rest of the system would get enough power, and boot and everything would be fine till the system was turned off.
Anyway, if your system starts up fine, then I think the capacity of the Seasonic is fine, and usually thier power regulation is very solid and if anything they tend to under-rate their power supplies compared to average brands.
If you have a bigger thermalright handy, easy enough to give it a try, but bet you a beer (if you are old enough
) that it will make no difference.-
aristide1
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
I suspect the same thing, which is why I will do the extra cooling first. I may go so far as to lap the cpu and hs. Too bad its common for the temps reported by the motherboard to read 10-20C too high. However when I say it won't run I should say that the system is cool and it will not boot, the board immediately decides to go into safe mode before the OS even has a chance to bootup. And I even set memory down to DDR2-400 and no change.
The northbridge was interesting because when I removed the original hot cooler the NB chip itself is tiny, the top surface area maybe the size of your typical pencil eraser. Needless to say a lack of thermal goop was going to be a bad situation here, but I think it suggests that the temp on the chip is much high than the readings I get on the hs. Same situation with the cpu.
The Seasonic 300 APFC I have is not even made anymore, and I won it here at an SPRC contest. Seems fitting I give something back. Only 1 12V rail.
The northbridge was interesting because when I removed the original hot cooler the NB chip itself is tiny, the top surface area maybe the size of your typical pencil eraser. Needless to say a lack of thermal goop was going to be a bad situation here, but I think it suggests that the temp on the chip is much high than the readings I get on the hs. Same situation with the cpu.
The Seasonic 300 APFC I have is not even made anymore, and I won it here at an SPRC contest. Seems fitting I give something back.
Only 1 12V rail.-
aristide1
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
My E6400 at stock speed 20 minutes 50 second (must be a small one again). Typical the 1760 ones are 23 minutes under 32 bit Vista.
I'd love to finish a WU on this machine every 30 hours (18 minute checkpoints). The X2 would be great if it did one every 36 hours (21.6 minutes) but I think 38 hours(22.8 minutes) will be it's best when I am done OCing the thing. Currently 25 minutes at 225MHz FSB (2.36GHz).
I wanted a Q6600 until I read this:
http://www.dailytech.com/article.aspx?newsid=8928
12MB L2 Cache!
I wonder how this little put-put would run:
http://www.dailytech.com/article.aspx?newsid=8910
There's also VIA's 1 watt processor.
http://www.viapc-1.com/
Solar powered folding. Yummy.
I'd love to finish a WU on this machine every 30 hours (18 minute checkpoints). The X2 would be great if it did one every 36 hours (21.6 minutes) but I think 38 hours(22.8 minutes) will be it's best when I am done OCing the thing. Currently 25 minutes at 225MHz FSB (2.36GHz).
I wanted a Q6600 until I read this:
http://www.dailytech.com/article.aspx?newsid=8928
12MB L2 Cache!
I wonder how this little put-put would run:
http://www.dailytech.com/article.aspx?newsid=8910
There's also VIA's 1 watt processor.
http://www.viapc-1.com/
Solar powered folding. Yummy.
-
aristide1
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Damn, I was stuck at 225MHz FSB with stock voltages. I then maxed out the NB and HT voltages. I got to 230MHz. 235 was no go, so I throttled back the memory and up the cpu voltage. No go, so I repeat that, I throttle the memory down to DDR533 and up the cpu voltage again. Now the memory is significantly slower for a 5MHz cpu gain, and it doesn't work. I go back to 230MHz, DDR2800 memory setting, drop the cpu voltage back to 1.375, and drop the NB and HT voltages down a notch. That works fine. All this still on the stock hs, but the thing fails to get much passed BIOS screen, it goes into some kind of safe mode right away. No iffy stuff or lockups.
Im curious what temps are now. A full report at some point is due.
Im curious what temps are now. A full report at some point is due.
-
aristide1
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
I have discovered through a site that I need to upgrade my BIOS for more OCing.aristide1 wrote:Damn, I was stuck at 225MHz FSB with stock voltages. I then maxed out the NB and HT voltages. I got to 230MHz. 235 was no go, so I throttled back the memory and up the cpu voltage. No go, so I repeat that, I throttle the memory down to DDR533 and up the cpu voltage again. Now the memory is significantly slower for a 5MHz cpu gain, and it doesn't work. I go back to 230MHz, DDR2800 memory setting, drop the cpu voltage back to 1.375, and drop the NB and HT voltages down a notch. That works fine. All this still on the stock hs, but the thing fails to get much passed BIOS screen, it goes into some kind of safe mode right away. No iffy stuff or lockups.
Down to 24 minutes 10 second per checkpoint.
Im curious what temps are now. A full report at some point is due.
-
aristide1
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Upon researching what I am doing right and wrong I found out the following:
1. My motherboards BIOS has a "SAFE MODE" that is not going to let me boot the system with an FSB above 230MHz, unless I ramp the voltage up to 1.5V from 1.35, where it will give me 5 more MHz, which is nothing.
2. People with high FSB's set their FSB in Windows after they boot up.
3. There do not appear to be any tools to accoomplish this task in Ubuntu, though I will waste some time by looking for some.
1. My motherboards BIOS has a "SAFE MODE" that is not going to let me boot the system with an FSB above 230MHz, unless I ramp the voltage up to 1.5V from 1.35, where it will give me 5 more MHz, which is nothing.
2. People with high FSB's set their FSB in Windows after they boot up.
3. There do not appear to be any tools to accoomplish this task in Ubuntu, though I will waste some time by looking for some.
-
aristide1
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
I've been working on folder #3, also a 4000+ on a Biostar 7050-M2 board. I was running with W2K Pro, because I wanted to push it to the max, OC wise. I managed to get to 2.5Ghz when memory finally gave out. I could run it slower, but I had already stepped down one speed on it so as it was. Wow there isn't a single hot chip or hot NB to be found on this board, and it seems to use about 10 watts less than the 6100 version of this board. CPU idle temps in the 30s, FAH SMP at 40C.
Here's what I noted:
1. Right off the bat FAH seems a little bit slower on W2K than it does on Ubuntu 64 bit. So eventually I will switch over to Ubuntu.
2. Compared to my other X2 which has memory speed now around DDR2-800, running at a slower memory speed showed up as a significant increase in the time it takes to take a checkpoint (1% complete message).
3. When I went from a 200 to 240MHz FSB I had to slow my memory. When I did that the checkpoints were taking longer than they did at the original speed. So what the hell was the point?
I decided to RMA the memory and buy some Team Extreme 8 layer PCB DDR2-800 which is reported to OC like mad. It's quite a bit more expensive than the 6 layer PCB of the same thing, but the price is still OK for 2*512MB.
At some point would consider faster memory for my E6400 setup as opposed to yet another folder as a way to increase my output. It also means that a quad 45nm in the future would be getting a minimum of DDR2-1066.
I also researched OCing quite a bit and the consensus is that you need at least 1 28amp 12 volt rail if you're going to be serious about it. I decided to purchase a Corsair VX450 for $70 delivered. Picked up a 150X USB stick (2GB $15) Patriot brand, that I will consider using as a C:\ drive just to play around some more and cause myself even more grief.
More results over the weekend.
My P5B-E has been rock steady at 400MHz FSB, an E6400 running at 3.2GHz, with a Seasonic 430watter and Patriot memory running synch DDR2-800 CAS 4. I'd OC the memory if the damn thing wasn't also my main pc for schoolwork. Essentially it's running twice as fast as a 4000+ at stock speeds ( 14 minute checkpoints versus 28 ).
Updated 28) to be 28 ).
Here's what I noted:
1. Right off the bat FAH seems a little bit slower on W2K than it does on Ubuntu 64 bit. So eventually I will switch over to Ubuntu.
2. Compared to my other X2 which has memory speed now around DDR2-800, running at a slower memory speed showed up as a significant increase in the time it takes to take a checkpoint (1% complete message).
3. When I went from a 200 to 240MHz FSB I had to slow my memory. When I did that the checkpoints were taking longer than they did at the original speed. So what the hell was the point?
I decided to RMA the memory and buy some Team Extreme 8 layer PCB DDR2-800 which is reported to OC like mad. It's quite a bit more expensive than the 6 layer PCB of the same thing, but the price is still OK for 2*512MB.
At some point would consider faster memory for my E6400 setup as opposed to yet another folder as a way to increase my output. It also means that a quad 45nm in the future would be getting a minimum of DDR2-1066.
I also researched OCing quite a bit and the consensus is that you need at least 1 28amp 12 volt rail if you're going to be serious about it. I decided to purchase a Corsair VX450 for $70 delivered. Picked up a 150X USB stick (2GB $15) Patriot brand, that I will consider using as a C:\ drive just to play around some more and cause myself even more grief.
More results over the weekend.
My P5B-E has been rock steady at 400MHz FSB, an E6400 running at 3.2GHz, with a Seasonic 430watter and Patriot memory running synch DDR2-800 CAS 4. I'd OC the memory if the damn thing wasn't also my main pc for schoolwork. Essentially it's running twice as fast as a 4000+ at stock speeds ( 14 minute checkpoints versus 28 ).
Updated 28) to be 28 ).
Last edited by aristide1 on Thu Dec 20, 2007 4:53 pm, edited 1 time in total.
x2 3800+ windsor @2.0GHz (stock)
2x1GB DDR2
AMD 690G chipset....
Command-line client. 34 minutes between checkpoints, 100 checkpoints per work unit for 500 points. Thats like 500 points every 2.5 days?
Am I doing something wrong? Also, my 5000+ @ 3.0GHz is doing 26 min checkpoints... which seems reasonable I guess?
2x1GB DDR2
AMD 690G chipset....
Command-line client. 34 minutes between checkpoints, 100 checkpoints per work unit for 500 points. Thats like 500 points every 2.5 days?
Am I doing something wrong? Also, my 5000+ @ 3.0GHz is doing 26 min checkpoints... which seems reasonable I guess?
-
aristide1
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Why aren't you running SMP? In both cases I see you using 1 cpu, and you have pairs of cpus. Task Manager will confirm if you're using one or two cores while folding.djkest wrote:x2 3800+ windsor @2.0GHz (stock)
2x1GB DDR2
AMD 690G chipset....
Command-line client. 34 minutes between checkpoints, 100 checkpoints per work unit for 500 points. Thats like 500 points every 2.5 days?
Am I doing something wrong? Also, my 5000+ @ 3.0GHz is doing 26 min checkpoints... which seems reasonable I guess?
My other 4000+ is OC'd 15%, just under 2.4GHz. 22 minutes per checkpoint about a day and a half for 1760 points. The dogs, the 1500 point jobs, are slower, 30 minutes per checkpoint.
Switch and see what happens, you might as well finish what you started with the -oneunit command. When it ends delete everything. Then download the SMP version, it's near the bottom of the list of Windows downloads.
Do you logon with a password? You need to turn this on for SMP to work. I did that and Windows doesn't ask me for a password, it simply defaults to me, but you still need a password "assigned" to your ID. Follow the install instructions on the FAQ page. The setup will ask you for your password TWICE, the cursor will not move. When successful you get duplicates messages as indicated, then you click on FAH, complete the configuration, and you're running SMP. Your Black Edition should kick some serious SMP butt.
Have you thought about using a VM?
The slight difference Ari reports in the performance of the W2K and Ubuntu folding is likely down to the fact the linux version is compiled to take advantage of the SSE extensions in the 64-bit CPUs. As reported in the the viewtopic.php?p=385094#385094 thread here, there is no 64-bit version for Windows, but using a VM enables you to host a 64-bit OS such as Ubuntu or Mandriva and run your folding in that "appliance". This is also, according to conventional wisdom, the best way to squeeze ppd out a four-core system, if Santa drops one of those down the chimney this year.
-
aristide1
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
SMP!djkest wrote:How do you guys do 1700 points a day?
Windows 2000/XP/Vista SMP client console version
http://folding.stanford.edu/English/Download
5.91 beta
Back up to my post from Thu Dec 20, 2007 12:08 am.
I tried for 30 min to install the SMP client on my power machine and it would NOT work. I explicitly followed the instructions and it didn't work. I think it said that it wasn't getting permission. I guess I dont' fully understand the account thing but I did put a password on it.. and isn't working.
I did notice that the console client only did about 50% of the max load on both cores, so there is definately some untapped potential there. BUT.. the plus side is, I was able to fold while playing Guitar Hero III (which is resource intensive) at max detail settings and even during the 2 hours of play I was ripping off 26 minute checkpoints. lol
If there is anyone out there with a lot of experience setting up SMP clients, especially in stupid Vista, please help me out here. I want to contribute!
I did notice that the console client only did about 50% of the max load on both cores, so there is definately some untapped potential there. BUT.. the plus side is, I was able to fold while playing Guitar Hero III (which is resource intensive) at max detail settings and even during the 2 hours of play I was ripping off 26 minute checkpoints. lol
If there is anyone out there with a lot of experience setting up SMP clients, especially in stupid Vista, please help me out here. I want to contribute!
-
spookmineer
- Patron of SPCR
- Posts: 749
- Joined: Sat Nov 11, 2006 6:02 pm
I was skimming through the results page over at extremeoverclocking and was shocked to see so many people getting so many ppd while I was tinkering at 186 point WU's (using the graphical client), which took me two and a half days (CPU at 50%).aristide1 wrote:SMP!djkest wrote:How do you guys do 1700 points a day?
Windows 2000/XP/Vista SMP client console version
http://folding.stanford.edu/English/Download
5.91 beta
Back up to my post from Thu Dec 20, 2007 12:08 am.
When you are starting to fold it's hard to get info on some specifics. There are good stickies here and over at F@H forums, but it doesn't help to put things into perspective.
Fortunately, Neil answered my PM asking how people get so many points. I actually though he had a network of at least 30 computers running (turns out he "just" has 6 of them...).
Maybe another sticky pointing out how to get the most out of your PC would get more folders over to SMP? I still see a lot of low point folders who are very loyal, folding every day.
A simple table of how much hours a day a PC is running, and the kind of CPU, and as a result the "best" client for that would help a lot of users (not all people are willing to run their PC 24/7: SMP is less attractive etc - the pros and cons need to be explained).
Hey Spook,
there is quite a bit of the specific sort of information that you mention in your last post on the folding forums and also some on the Extreme Overclock site folding forums. (That site has a great stats page, and a little friendly competition while watching those stats has caused more than a few of us to also purchase more hardware for folding.)
there is quite a bit of the specific sort of information that you mention in your last post on the folding forums and also some on the Extreme Overclock site folding forums. (That site has a great stats page, and a little friendly competition while watching those stats has caused more than a few of us to also purchase more hardware for folding.)
-
spookmineer
- Patron of SPCR
- Posts: 749
- Joined: Sat Nov 11, 2006 6:02 pm
Yes, I know (I linked to the SPCR stats).
But I only found out after seeing these stats, and trying to find info about how to fold "faster".
For new folders this info is not readily available here, even people folding for a few years might not even know about it (I had to dig deeper then needed).
Like most newcomers, I just read the "F@H beginners guide" (started in 2004) and downloaded the graphical client. As more and more PC's are equipped with dual or even quad cores, more people should at least look into getting the SMP client above any other client.
If people are not aware that the difference in ppd is this big, they will not even look into it further I'm afraid.
But I only found out after seeing these stats, and trying to find info about how to fold "faster".
For new folders this info is not readily available here, even people folding for a few years might not even know about it (I had to dig deeper then needed).
Like most newcomers, I just read the "F@H beginners guide" (started in 2004) and downloaded the graphical client. As more and more PC's are equipped with dual or even quad cores, more people should at least look into getting the SMP client above any other client.
If people are not aware that the difference in ppd is this big, they will not even look into it further I'm afraid.
Code: Select all
Unable to connect to 'Gamer:8676',
sock error: generic socket failure, error stack:
MPIDU_sock_post_connect(1228): unable to connect to gamer on port 8676, exhausted all endpoints (errno -1)
MPIDU_sock_post_connect(1275): unable to connect to Gamer on port 8676, no connection could be made because the target machine actively refused it. (errno 10061)
Press any key to continue.I tried disabling all the security features in Vista, and it still doesn't work. I'm seriously considering UPGRADING to windows XP again.
-
aristide1
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Re: Have you thought about using a VM?
In comparing 2 similar but not identical boards (the 6100 chipset is running Ubuntu) and the 7050 chip (both NVidia and the 2nd merely has HDTV plugs) I've had to OC the 7050, which is running W2K, far more to get into the same league as the 6100, ie about 22-23 minutes per checkpoint. The 6100 is running 230MHz FSB for about 2.367GHz, the 7050 is running 280MHz FSB 2.9+GHz. They both have an AMD X2 4000+ but I must have the newer stepping on the latter. With an Artic Cooler 64, an all aluminum job that's simply larger than stock it's lukewarm even at this speed. When I go home this weekend I will probably end up at 10*300MHz 3GHz (to maximize memory speed which is compromised with normal Brisbane .5 multipliers.) The memory speed will end up at DDR2-1000 (configed as DDR667 running 50% OC as well). At 2.2 volts the memory is barely warm. The processor is at 1.365 volts, barely above spec.Dutchmm wrote:The slight difference Ari reports in the performance of the W2K and Ubuntu folding is likely down to the fact the linux version is compiled to take advantage of the SSE extensions in the 64-bit CPUs. As reported in the the viewtopic.php?p=385094#385094 thread here, there is no 64-bit version for Windows, but using a VM enables you to host a 64-bit OS such as Ubuntu or Mandriva and run your folding in that "appliance". This is also, according to conventional wisdom, the best way to squeeze ppd out a four-core system, if Santa drops one of those down the chimney this year.
Now I gotta tell you. I had all kinds of problems with the Crucial Memory I was using (probably not the culprit from what I read in NewEgg reviews) and the Antec EA380. I came home and installed a Corsair VX450, which I have never heard running because of the Artic Cooling hs, and Team Extreme, 8 layer PCB DDR2-800, with these huge screwed on heat sinks. I worked my way up 5-10 MHz at a time. I'm 40+% over now and I haven't seen a glitch or a hot temp amywhere. The NB heatsink warmed up after 4 days, something I attribute to the thermal paste finally getting around and doing it's job. Some AS will be installed on that bugger. I have a bigger replacement if it starts running hot, but so far this chipset and board (Biostar TForce) is a killer for the money, and I've flashed both my Biostar boards without any problems, which is a LOT more than I can say for my Asus P5B-E with AMI (yuck) BIOS.
You know the E2140 and those guys, those CPUs have multipliers like 12 for 2.4GHz. Which means if you can run them 333FSB then you got yourself a 3GHz dual core, but even more interesting will be if they start manufacturing them at 45nm. It wouldn't surprise me to see one of those reaching 4GHz. One guy on Rebels Haven (a great place for BIOS) has already OCd one of the E21xx series processors 100%. Yep, from 200FSB to 400FSB. Don't have the details but its quite an accomplishment in any case.
Last edited by aristide1 on Wed Dec 26, 2007 5:18 pm, edited 1 time in total.
-
aristide1
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
This is my TForce board with the 6100 chipset. I can bypass "Safe mode" in Windows but at the expense of slower processing. Instead the future for this setup will be a X2 Black Edition with has unlocked multipliers. Even at the slower FSB one can still setup this processor to run at 3GHz.aristide1 wrote:Upon researching what I am doing right and wrong I found out the following:
1. My motherboards BIOS has a "SAFE MODE" that is not going to let me boot the system with an FSB above 230MHz, unless I ramp the voltage up to 1.5V from 1.35, where it will give me 5 more MHz, which is nothing.
2. People with high FSB's set their FSB in Windows after they boot up.
3. There do not appear to be any tools to accomplish this task in Ubuntu, though I will waste some time by looking for some.
Once my little experiment is concluded, I will have 2 similar boards, both with an X2 processor, and both running around 3Ghz. The difference will be the one will achieve 3GHz by running 10* 300MHz FSB while the other will achive it by running say 13*230MHz FSB. Both will be running Ubuntu then and I will be curious if one runs dramatically faster then the other.
I will say this (do I ever stop talking?), as I was increasing the FSB on the 7050 board there were several instances where all of a sudden checkpoints started taking longer. I suspect, like the Intel NB, that divider frequencies must have changed, slowing some settings down in order to remain stable. I don't expect to see a 1:1 correlation with overclocking percentages and time saved per checkpoints. I will also say that slower memory will definitely show up in your times.
Here's a simple way to determine the actual speed of your memory with the AMD divider setup.
Lets say you have the bargain basement X2 4000+, now at $57 at the Egg, and DDR2-800 memory (actual speed 400MHz).
So
2100 MHz (processor) / 400 MHz (memory) = 5.25
The divider can only do whole numbers and must round up, so it does
2100 MHz (processor) / 6 (the divider above rounded up) = 350MHz.
So DDR2-800 will actually run as DDR-700. This calculation is done ONCE and only at stock speeds, so if you OC you add whatever the result whatever you increased the overall FSB, 10%, 20%, 50%.
Now here a kicker with the 4000+. As you see above the memory will run slow, but the 7050 motherboard allows you to set the cpu multiplier from 10.5 to 10, which is a great thing. So you ask, why? Well
2000 MHz / 400MHz = 5 ergo,
2000 MHz / 5 (no rounding up) = 400MHz. So DDR2-800 runs at, duh, 400MHz as DDR2-800.
Now suppose you want your processor to still run at 2.1GHz, well increase the FSB 5% (negligible) and now you once again have a 2.1 GHz processor but now the memory is running DDR2-840 instead of DDR-700.
FYI a processor running 2000MHz with DDR-667 (333MHz) results in a completely even 6 divider as well, no rounding on this pair of settings.
Goal - complete a 1760 WU in under 36 hours, or 2 WU every 3 days.
-
aristide1
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
The OC forums I've been wandering through mention some interesting items:
1. A 4 GHz club.
2. A 5 GHz club, along with a distinct possibility of severe frostbite.
3. A 100% club, typically Intel E2xxx processors.
4. A guy who took an Intel E2xxx processor 180% OC. Yeah, from a 200MHz FSB to 560MHz.
I wish I could bold this guy -->
By the way advice on Prime 95 is all over the place courtesy our own Ralf Hutter.
1. A 4 GHz club.
2. A 5 GHz club, along with a distinct possibility of severe frostbite.
3. A 100% club, typically Intel E2xxx processors.
4. A guy who took an Intel E2xxx processor 180% OC. Yeah, from a 200MHz FSB to 560MHz.
I wish I could bold this guy -->
By the way advice on Prime 95 is all over the place courtesy our own Ralf Hutter.