- Joined
- Feb 18, 2002
1. Why is it a better idea to shoot for [higher] FSB x [lower] multiplier to get the same MHz?
2. What damages RAM more: higher voltage or setting Memory settings in BIOS to aggressive instead of optimal?
3. If a mobo has a PCI lock protecting PCI and video cards, what about IDE drives, are they protected more with [lower] FSB x [higher] multiplier, or the other way around?
EDIT: More specific questions:
4. Is there any downside to
[higher] FSB x [lower] multiplier
over
[lower] FSB x [higher] multiplier to get the same MHz?
Particularily if the FSB is in the 200 range.
5. Do you ALWAYS have to 'load optimal settings' after updating BIOS (which resets your specific previous settings)?
Reasons for asking:
166 FSB x 13 multiplier = 2166MHz and bootup display of Athlon XP 2700+
[higher] FSB x [lower] multiplier = 2166MHz and bootup display of Athlon XP 2600+
Higher MHz are not displayed as PR ratings at bootup, but as MHz
2. Would like to extend hardware lifetime despite how quickly these things get obsolete.
3. If you’re lucky to own an nForce2 with PCI lock (Epox, Abit, Asus, Chaintech), your PCI cards are protected from negative effects of overclocking but what about other components?
THANK YOU VERY MUCH
HERE'S A SUMMARY OF THREAD ANSWERS TO REAL NON-STUPID OVERCLOCKING QUESTIONS:
1. Why is it a better idea to shoot for [higher] FSB x [lower] multiplier to get the same MHz?
Higher FSB gives you better all around performance because RAM is working faster. Higher FSB is almost always more beneficial than higher pure MHz. I don't know at what point it would negate the difference, but higher FSB means far, far greater performance. Increased memory bandwidth means that you relieve some of the bottlenecking happening between the components. There are no negative effects to doing this unless you have to volt mod your board to get it to run at 200 or so.
2. Setting Memory settings in BIOS to aggressive instead of optimal should not damage your RAM. Very high voltage will.
3. If a mobo has a PCI lock protecting PCI and video cards, what about IDE drives?
The IDE controller built into a motherboard operates off of the PCI bus, and when the bus is locked down, then the IDE controller is still running in spec. Hard drives are typically designed to run on a 33mhz PCI bus, and sometimes when they run too far off that they begin to stop functioning as they should. PCI locks prevent this. As for the rest of your components, your PCI cards and AGP card will still be running in spec too. You can't tell without special instruments if there is a PCI lock. Google search engine or discussion boards are the only way to find out if there's documented evidence if your mobo is PCI locked (good) or not (not as good). Unfortunately, most are not.
4. Is there any downside to
[higher] FSB x [lower] multiplier
over
[lower] FSB x [higher] multiplier to get the same MHz?
Not if your mobo has a PCI lock. However, if your mobo doesn't have a PCI lock that means that too high of an FSB may take your hard drive and PCI cards too far out of specs.
Think of the FSB as a hose of varying diameter. Think of the CPU as a firemen with buckets waiting to put out a fire. Think of the water as the data need to put out that fire. The fatter the hose, the sooner they get the water. How many men you have waiting with buckets could be thought of in terms of multiplier, sort of. Essentially, your processor can outwork everything else in your system, and frequently waits for data to crunch. With a higher FSB, it gets the data sooner (and can send it sooner, too) so overall system performance increases.
Example: It's better to have an increase of 7 FSB and end up with resulting FSB X multiplier of 20 MHz less.
5. Do you ALWAYS have to 'load optimal settings' after updating BIOS (which resets your specific previous settings)?
Some say no. However some say that it's good to even always clear the cmos after bios flash before the first boot/post. They say in newer bios there may be some new variables which will get their values from places which weren't used before or were used for other variables...
6. Although upping voltage may affect life span of RAM, it was said that aggressive settings to RAM do not affect its life span... then how would you recommend adjusting the most widely available type of RAM, the PC2100 whose defaults are:
T(RAS) 6
T(RCD) 3
T(RP) 3
CAS 2.5
1. Active (to) Precharge Delay (aka Tras, tRAS) - usually 5 or 6 or 7, smaller the better, (Tras >= Trcd + CAS)
2. RAS to CAS Delay (aka Trcd, tRCD) - 2 is good, 3 is OK
3. RAS Precharge Delay (aka Trp, tRP, Precharge to Active) - 2 is good, 3 is OK
4. CAS Latency (aka CAS) - use 2 whenever possible
5. Cmd Rate (some bios does not have this, set automatically) - 1T is better than 2T
CAS Latency is most important for memory bandwidth, or for bursting a large block/page of data like in 3D, video, gaming applications.
Set it to 2 whenever possible.
2nd important is the RAS to CAS Delay (Trcd).
It is the number of cycles between the row and column access. 2 is good, but 3 is OK.
The Precharge Delay (Trp) is the precharge time after an active access. 2 is good, but 3 is OK.
Active to Precharge Delay (Tras) is the minimum time for an active access
(to perform a single row and column access).
It is the least importance for memory performance. Usually 5 or 6 or 7 is fine. Tras >= Trcd + CAS
The difference between running GOOD memory at the most agressive timings ( which it CAN do ) vs avg memory at moderate timings ( which is the BEST it can do ) is about 5-7% in 3D Mark.
7. Quality heatsink and fan and high rated RAM are essential for overclocking but so are power supply voltages. If line voltages go 5% lower then rated lines, this may be a sign that you need a better PSU.
Also if you ever wonder why not to spend more on a better heatsink&fan instead of case fans:
The reason is because to have a truely effective cooling system you need to add BOTH a better heatsink AND better case cooling. If you just get a better heatsink and no extra case fans you case will just get hotter and hotter untill even with your good heatsink your CPU is still overheating.
8. To test the stability of the overclocked system:
Use Prime95 for CPU and RAM stability tests (Prime95 Options > Torture Test)
http://www.mersenne.org/freesoft.htm
SiSoftware Sandra (Burn-in)
http://www.sisoftware.co.uk/
3dMark with all the tests including DirectX testing features for Direct Draw and Direct3D
http://www.futuremark.com/download/
________________
2100+ AIUHB 0248
[206] FSB x 11 = 2266 MHz (2800+)
at 1.75 V @ 37(min) C to 46(max) C
Epox 8RDA+
Thermalright SK-7 with variable speed 80x25mm YS-Tech FD1281259B-2F
2x 256MB Kingston (Hynix chip) PC2100 7 3 3 3 @ 6 3 3 2.5
GeForce4 Ti4200 ; Antec SX-835II case ; Antec SmartPower SL350
EDIT: increasing to 1.8 V resulted in 6 2 2 2 stable
and [208] x 11 = 2288 MHz @ 37 to 46 C
EDIT2: But that's only because Memory Frequency was not changed from Auto to 100%, which it always should be because regardless of benchmark scores, real life performance is best with 100% insync instead of Auto:
[152] FSB x 15 = 2280 MHz @ 1.8 Vcore with memory frequency at 100% [152] @ 2.77 Vdimm
EDIT3: To test the stability of overclocked CPU, download the latest bug fixed version of Prime95:
http://www.mersenne.org/freesoft.htm
Double click on PRIME95.EXE to start the program.
Just Stress Testing (if prompted) > OK >
Options > Torture Test > Blend (if prompted) > OK
The program should not give any errors.
It may be necessary to run the program 12-24 hours to make sure an overclocked system is stable (no program errors displayed).
If you get errors only after several hours, this is a result of slight instability because the system is running with little or no margin. It's stable enough to boot and to be moderately stressed, but as soon as the system is under enough load to go over that critical point, it will freeze. To be 100% stable, Prime95 should run 12-24 hours without any errors.
The point of testing is to see if you get errors or not thus testing the stability. The meaning of errors themselves is not as important.
Running Prime95 Torture Test for 5 to 30 minutes is enough to get a feeling about general stability. If it seems to be stable, increase the FSB (with 0.025V Vcore increase if needed), or increase the multiplier by 0.5.
Reboot and run Prime95 Torture Test for 5 to 30 minutes and repeat this until the Program displays errors.
It's then time to back down and repeat the Torture Test until there are no errors for at least 12 hours.
It's OK to use the computer while Prime95 Torture Test is running in the background.
After finding your stable point you may want to then retest by going to the Advanced section, setting the password to 9876 and then setting Priority to 10. (You won't be able to use your computer while Priority 10 Prime95 Torture Test is running, but it'll confirm your system stability.)
Use 3DMark to test video card stability.
EDIT: Average stick overclocks:
PC2100s OC from 133 to high 140s or low 150s FSB 100% insync.
PC2700s OC from 166 to high 180s or low 190s FSB 100% insync
PC3200 usually way into 200s, depending on brand. There are PC3000 sticks that are excellent overclockers too.
2. What damages RAM more: higher voltage or setting Memory settings in BIOS to aggressive instead of optimal?
3. If a mobo has a PCI lock protecting PCI and video cards, what about IDE drives, are they protected more with [lower] FSB x [higher] multiplier, or the other way around?
EDIT: More specific questions:
4. Is there any downside to
[higher] FSB x [lower] multiplier
over
[lower] FSB x [higher] multiplier to get the same MHz?
Particularily if the FSB is in the 200 range.
5. Do you ALWAYS have to 'load optimal settings' after updating BIOS (which resets your specific previous settings)?
Reasons for asking:
166 FSB x 13 multiplier = 2166MHz and bootup display of Athlon XP 2700+
[higher] FSB x [lower] multiplier = 2166MHz and bootup display of Athlon XP 2600+
Higher MHz are not displayed as PR ratings at bootup, but as MHz
2. Would like to extend hardware lifetime despite how quickly these things get obsolete.
3. If you’re lucky to own an nForce2 with PCI lock (Epox, Abit, Asus, Chaintech), your PCI cards are protected from negative effects of overclocking but what about other components?
THANK YOU VERY MUCH
HERE'S A SUMMARY OF THREAD ANSWERS TO REAL NON-STUPID OVERCLOCKING QUESTIONS:
1. Why is it a better idea to shoot for [higher] FSB x [lower] multiplier to get the same MHz?
Higher FSB gives you better all around performance because RAM is working faster. Higher FSB is almost always more beneficial than higher pure MHz. I don't know at what point it would negate the difference, but higher FSB means far, far greater performance. Increased memory bandwidth means that you relieve some of the bottlenecking happening between the components. There are no negative effects to doing this unless you have to volt mod your board to get it to run at 200 or so.
2. Setting Memory settings in BIOS to aggressive instead of optimal should not damage your RAM. Very high voltage will.
3. If a mobo has a PCI lock protecting PCI and video cards, what about IDE drives?
The IDE controller built into a motherboard operates off of the PCI bus, and when the bus is locked down, then the IDE controller is still running in spec. Hard drives are typically designed to run on a 33mhz PCI bus, and sometimes when they run too far off that they begin to stop functioning as they should. PCI locks prevent this. As for the rest of your components, your PCI cards and AGP card will still be running in spec too. You can't tell without special instruments if there is a PCI lock. Google search engine or discussion boards are the only way to find out if there's documented evidence if your mobo is PCI locked (good) or not (not as good). Unfortunately, most are not.
4. Is there any downside to
[higher] FSB x [lower] multiplier
over
[lower] FSB x [higher] multiplier to get the same MHz?
Not if your mobo has a PCI lock. However, if your mobo doesn't have a PCI lock that means that too high of an FSB may take your hard drive and PCI cards too far out of specs.
Think of the FSB as a hose of varying diameter. Think of the CPU as a firemen with buckets waiting to put out a fire. Think of the water as the data need to put out that fire. The fatter the hose, the sooner they get the water. How many men you have waiting with buckets could be thought of in terms of multiplier, sort of. Essentially, your processor can outwork everything else in your system, and frequently waits for data to crunch. With a higher FSB, it gets the data sooner (and can send it sooner, too) so overall system performance increases.
Example: It's better to have an increase of 7 FSB and end up with resulting FSB X multiplier of 20 MHz less.
5. Do you ALWAYS have to 'load optimal settings' after updating BIOS (which resets your specific previous settings)?
Some say no. However some say that it's good to even always clear the cmos after bios flash before the first boot/post. They say in newer bios there may be some new variables which will get their values from places which weren't used before or were used for other variables...
6. Although upping voltage may affect life span of RAM, it was said that aggressive settings to RAM do not affect its life span... then how would you recommend adjusting the most widely available type of RAM, the PC2100 whose defaults are:
T(RAS) 6
T(RCD) 3
T(RP) 3
CAS 2.5
1. Active (to) Precharge Delay (aka Tras, tRAS) - usually 5 or 6 or 7, smaller the better, (Tras >= Trcd + CAS)
2. RAS to CAS Delay (aka Trcd, tRCD) - 2 is good, 3 is OK
3. RAS Precharge Delay (aka Trp, tRP, Precharge to Active) - 2 is good, 3 is OK
4. CAS Latency (aka CAS) - use 2 whenever possible
5. Cmd Rate (some bios does not have this, set automatically) - 1T is better than 2T
CAS Latency is most important for memory bandwidth, or for bursting a large block/page of data like in 3D, video, gaming applications.
Set it to 2 whenever possible.
2nd important is the RAS to CAS Delay (Trcd).
It is the number of cycles between the row and column access. 2 is good, but 3 is OK.
The Precharge Delay (Trp) is the precharge time after an active access. 2 is good, but 3 is OK.
Active to Precharge Delay (Tras) is the minimum time for an active access
(to perform a single row and column access).
It is the least importance for memory performance. Usually 5 or 6 or 7 is fine. Tras >= Trcd + CAS
The difference between running GOOD memory at the most agressive timings ( which it CAN do ) vs avg memory at moderate timings ( which is the BEST it can do ) is about 5-7% in 3D Mark.
7. Quality heatsink and fan and high rated RAM are essential for overclocking but so are power supply voltages. If line voltages go 5% lower then rated lines, this may be a sign that you need a better PSU.
Also if you ever wonder why not to spend more on a better heatsink&fan instead of case fans:
The reason is because to have a truely effective cooling system you need to add BOTH a better heatsink AND better case cooling. If you just get a better heatsink and no extra case fans you case will just get hotter and hotter untill even with your good heatsink your CPU is still overheating.
8. To test the stability of the overclocked system:
Use Prime95 for CPU and RAM stability tests (Prime95 Options > Torture Test)
http://www.mersenne.org/freesoft.htm
SiSoftware Sandra (Burn-in)
http://www.sisoftware.co.uk/
3dMark with all the tests including DirectX testing features for Direct Draw and Direct3D
http://www.futuremark.com/download/
________________
2100+ AIUHB 0248
[206] FSB x 11 = 2266 MHz (2800+)
at 1.75 V @ 37(min) C to 46(max) C
Epox 8RDA+
Thermalright SK-7 with variable speed 80x25mm YS-Tech FD1281259B-2F
2x 256MB Kingston (Hynix chip) PC2100 7 3 3 3 @ 6 3 3 2.5
GeForce4 Ti4200 ; Antec SX-835II case ; Antec SmartPower SL350
EDIT: increasing to 1.8 V resulted in 6 2 2 2 stable
and [208] x 11 = 2288 MHz @ 37 to 46 C
EDIT2: But that's only because Memory Frequency was not changed from Auto to 100%, which it always should be because regardless of benchmark scores, real life performance is best with 100% insync instead of Auto:
[152] FSB x 15 = 2280 MHz @ 1.8 Vcore with memory frequency at 100% [152] @ 2.77 Vdimm
EDIT3: To test the stability of overclocked CPU, download the latest bug fixed version of Prime95:
http://www.mersenne.org/freesoft.htm
Double click on PRIME95.EXE to start the program.
Just Stress Testing (if prompted) > OK >
Options > Torture Test > Blend (if prompted) > OK
The program should not give any errors.
It may be necessary to run the program 12-24 hours to make sure an overclocked system is stable (no program errors displayed).
If you get errors only after several hours, this is a result of slight instability because the system is running with little or no margin. It's stable enough to boot and to be moderately stressed, but as soon as the system is under enough load to go over that critical point, it will freeze. To be 100% stable, Prime95 should run 12-24 hours without any errors.
The point of testing is to see if you get errors or not thus testing the stability. The meaning of errors themselves is not as important.
Running Prime95 Torture Test for 5 to 30 minutes is enough to get a feeling about general stability. If it seems to be stable, increase the FSB (with 0.025V Vcore increase if needed), or increase the multiplier by 0.5.
Reboot and run Prime95 Torture Test for 5 to 30 minutes and repeat this until the Program displays errors.
It's then time to back down and repeat the Torture Test until there are no errors for at least 12 hours.
It's OK to use the computer while Prime95 Torture Test is running in the background.
After finding your stable point you may want to then retest by going to the Advanced section, setting the password to 9876 and then setting Priority to 10. (You won't be able to use your computer while Priority 10 Prime95 Torture Test is running, but it'll confirm your system stability.)
Use 3DMark to test video card stability.
EDIT: Average stick overclocks:
PC2100s OC from 133 to high 140s or low 150s FSB 100% insync.
PC2700s OC from 166 to high 180s or low 190s FSB 100% insync
PC3200 usually way into 200s, depending on brand. There are PC3000 sticks that are excellent overclockers too.
Last edited: