Oc'ing Hurts Performance... [Archive]

3rd gen noob

10-05-2003, 01:48 AM

a friend of mine has just built up an uber PC
his spec is as follows:

Athlon 2800+ (Barton)
ASUS A7N8X-D
2x512MB TwinMOS PC3200
Radeon 9800 Pro 128MB
120GB S-ATA hdd
450W Antec PSU
Thermaltake Xaser Case
Lite-On 52/32/52 CD-RW
16/48 DVD-R

it's obviously a rather powerful machine, benchmarking in 3dMark03 at 5409 standard:

http://www.arap78.dsl.pipex.com/standard.JPG

however, he wasn't satisfied with that and overclocked the cpu and tweaked the ram to try and increase performance...
he dropped the cpu multiplier to 11 and increased the fsb to 200MHz ddr (400MHz effective), yielding a clock speed of 2.2GHz (effectively an Athlon 3200+)
he also tightened the timings from the standard 2.5-4-4-8 to 2-3-3-7, however, this overclocked setup actually caused a huge dip in performance:

http://www.arap78.dsl.pipex.com/OC'd.JPG

does anyone know what the cause could be?
firstly, you'd think temperature might be the issue here, however, his system is running pretty cool:

http://www.arap78.dsl.pipex.com/asusprobe.JPG

is it possible that the large increase in fsb (nearly 20%), has interfered with the AGP bus and caused the decrease in performance?
any help/advice would be greatly appreciated

adamp2p

10-05-2003, 02:42 AM

Hmmm, that is rather strange. What else is going on there, because that makes no sense. Overclocking should increase performance, but only hypothetically.

If the overclock somehow interferes with the devices in the machine, it may cause you some trouble; however, my experiences with overclocking have been *lovely*; and the only troubles I have ran into have been either the fault of the operating system or the fact that I had a program running (in one case flashget seriously hamered performance in real time gaming, it caused a lag) that I did not realize. However, when I stopped the program everthing went back to normal again. You might want to reboot that machine and see if you can find some more stable settings. Thanks for the post; I like this discussion, as all of those you engage in 3rd gen noob.

B)

3rd gen noob

10-05-2003, 03:09 AM

ok, a bit of an update...strangely enough, it seems to have been the agp frequency which was throwing things out...he's locked that to 66MHz and tried again.
this time he got up to 5861, but he's still not happy...:lol:

http://www.arap78.dsl.pipex.com/bestsofar.JPG

now he's oc'ing the gpu to try and crack 6000

@adam, thanks for the advice, this problem had us a little baffled too...we were lucky we remembered the agp frequency thing...i suppose overclocking isn't a very exact science, so it's hard to diagnose things sometimes...

LeGoMyFnLeg

10-05-2003, 03:13 AM

If you have mentioned the adjusted core voltage anywhere in this thread, I am not able to find it. Please don't tell me that you have attempted an overclock without a core voltage increase proportional to the expected clockrate.

3rd gen noob

10-05-2003, 03:16 AM

Originally posted by LeGoMyFnLeg@5 October 2003 - 03:13
If you have mentioned the adjusted core voltage anywhere in this thread, I am not able to find it. Please don't tell me that you have attempted an overclock without a core voltage increase proportional to the expected clockrate.
no cpu core voltage change...is this necessary?

james_bond_rulez

10-05-2003, 03:17 AM

power hungry bastards... <_<

LeGoMyFnLeg

10-05-2003, 03:18 AM

Absolutely! Where do you think the need for expensive cooling measures comes from? It's the heat generated by more electrons passing through the processor.

3rd gen noob

10-05-2003, 03:22 AM

Originally posted by LeGoMyFnLeg@5 October 2003 - 03:18
Absolutely! Where do you think the need for expensive cooling measures comes from? It's the heat generated by more electrons passing through the processor.
well, if the system is stable at the standard voltage, is there any point increasing the voltage (and therefore heat)?

he's up to 5997 now at 405/750 on the gpu

http://www.arap78.dsl.pipex.com/5997.JPG

adamp2p

10-05-2003, 03:32 AM

Wow, that is a powerful score/machine! What's that dude get on Aquamark 3?

james_bond_rulez

10-05-2003, 03:32 AM

you have to increase cup voltage to properly overclock otherwise your games would hang.

it's like you are running a high performance engine but the fuel injectors are not pumping enough fuel into it the car stalls

abu_has_the_power

10-05-2003, 03:32 AM

a really good place to find some info is forums.extremeoverclocking.com

this is a pirating forum. lol. no really, extreme overclocking will give u quite a lot of good answers. more than 70% of that forum's members have oced successfully before. try it. i go there for complicated hardware problems.

LeGoMyFnLeg

10-05-2003, 03:34 AM

3d mark is only 1 measure and not highly regarded as an indicator of total system performance in a lot of circles.
Check out some of the more mature overclocking sites and tutorials on the web and you'll see it's standard practice to increase core voltage.
I see from your image that CPU performance was not included in these results at all. Athlons love a core increase, with 1.70V being standard, some have had great success with voltages as high as 2.10V. Heat and common sense are the limiting factors.

3rd gen noob

10-05-2003, 03:36 AM

Originally posted by adamp2p@5 October 2003 - 03:32
Wow, that is a powerful score/machine! What's that dude get on Aquamark 3?
he's not tried aquamark yet...he's still on 31.2k modem (he lives more in the middle of nowhere than i do), so the large download would take him hours to get
i'm gonna burn it to disc and take it up to him sometime though, it'll be interesting to see what he can score in that...

@jamez_bond_rulez, my point was simply, why would you want to add more 'fuel' if you don't need it?
adding too much is just as bad as adding too little...if the system can run 3dmark03 numerous times, it should be considered stable

3rd gen noob

10-05-2003, 03:40 AM

Originally posted by LeGoMyFnLeg@5 October 2003 - 03:34
I see from your image that CPU performance was not included in these results at all.
the CPU core is shown in the asusprobe pic

http://www.arap78.dsl.pipex.com/asusprobe.JPG

when he comes back online, i'll get him to run some of the tests in Sandra...he's done a couple of the cpu load testers with no problems though

LeGoMyFnLeg

10-05-2003, 03:44 AM

I see it is just very slightly under powered. 5V, 3.3V, and core are all slightly below expected values while 12V is ok. At the temperatures indicated along with the low fan speed, something is not quite up to par. I think there might be a lot more potential being wasted with these figures and settings.

3rd gen noob

10-05-2003, 03:48 AM

Originally posted by LeGoMyFnLeg@5 October 2003 - 03:44
I see it is just very slightly under powered. 5V, 3.3V, and core are all slightly below expected values while 12V is ok. At the temperatures indicated along with the low fan speed, something is not quite up to par. I think there might be a lot more potential being wasted with these figures and settings.
i don't understand, what do you mean "not quite up to par"?
surely the slight differences in the 3.3 and 5v rails is negligible?

LeGoMyFnLeg

10-05-2003, 03:59 AM

Maybe but why not test that? All devices will have internal voltage regulation so that a slight increase will not hurt, but being limited in performance by a lack of available current would actually be harmful, possibly even resulting in more heat being produced as a result.
Much as a standard car battery is nominally rated at 12V, in operation it is more like 15.5V.
I would expect to see 3.3 at 3.5, 5 at 5.25, 12 at 12.50 and so on in a healthy system.

_John_Lennon_

10-05-2003, 06:57 AM

Originally posted by LeGoMyFnLeg@4 October 2003 - 22:44
I see it is just very slightly under powered. 5V, 3.3V, and core are all slightly below expected values while 12V is ok. At the temperatures indicated along with the low fan speed, something is not quite up to par. I think there might be a lot more potential being wasted with these figures and settings.
low fan speed? his CPU is running up there, so its not stock either.

But, is the power fan actually for the power supply? Or is it just plugged into the port it is reading as that, because 1500 seems more like a low end case fan number.

3RA1N1AC

10-05-2003, 09:20 AM

Originally posted by 3rd gen noob@4 October 2003 - 19:09
ok, a bit of an update...strangely enough, it seems to have been the agp frequency which was throwing things out...he's locked that to 66MHz and tried again.
this time he got up to 5861, but he's still not happy...:lol:
while reading the first post, i was just thinking "hmm... something else has got to be causing the problem, besides what he's mentioned." yeah, knocking the AGP bus speed outta whack oughta f*ck things up pretty good. :lol:

it's worth mentioning to your friend that 3DMark 2003 is essentially a video card benchmark. computers with low end CPUs achieve similar scores to computers with high end CPUs, when tested with the same video card. overclocking the CPU will not make nearly as much of a difference as overclocking the video card in that test. he should try other benchmarks, if he really wants to measure the effect of CPU overclocking. 3DMark 2001 is much more of an overall system test, for example.

bigdawgfoxx

10-05-2003, 02:20 PM

Is overclocking all done through the BIOS? Do you raise the AGP bus to increase it...and the CPU bus to increase that? is it really that easy?

lynx

10-05-2003, 04:04 PM

Very often, the pci and agp bus speeds are a division of the FSB clock.

So for instance if the nominal FSB speed is 133MHz, the pci speed would be FSB/4 (22MHz) and the agp speed would be FSB/2 (66MHz). If the FSB speed is increased to 150MHz, this would change the pci speed to 37.5MHz and agp speed to 75MHz. This can be enough to cause numerous faults on pci and agp buses, and consequently cause instability. This is often called the southbridge multiplier.

Often the memory speed is linked to the FSB speed too, this is called the Northbridge multiplier.

Some motherboards have the ability to use a separate clock for the Northbridge and Southbridge speeds, which is useful in determining maximum stable processor speeds, but the result is that memory and peripheral transfers are out of sync with the processor, which reduces performance slightly, but as we have seen from 3GN's posts the end result is still better than standard.

And while this is usually done in bios settings, motherboard manufacturers often supply software to adjust FSB and multiplier speeds 'on the fly', which can be extremely useful for testing system stability - if the settings have been locked into the bios, and the system won't boot because the speed is too high, the only solution is to clear the cmos and start again, and hope you can remember what all the settings were.

bigdawgfoxx

10-05-2003, 05:12 PM

ok thanx that helped a lot....but whyd you use 133..arent FSB at 800Mhz now? ...so would the PCI be 200mhz and the AGP 400 mhz? or is 133 not the fsb? a lil confused on that one...but thanx...would you just set the agp and pci to stay at 66 and 22mhz and just raise the CPU clock?

lynx

10-05-2003, 06:05 PM

It was just an example.
If the FSB clock was 200MHz then default agp would be FSB/3 (still 66) and default pci would be FSB/6 (still 33)

Note, FSB800 is actually 200MHz with a 4x pre-multiplier, confusing isn't it.

That's why it is important to look at base clock speeds rather than quoted rates, because of the multipliers/dividers involved.

If you have the ability to lock the agp and pci (and preferably memory too) it is a good idea to do so initially, to establish the maximum speed the processor is capable of sustaining. But it is also a good idea to find out the maximum speeds you can run the other devices at without exceeding the processor capabilities, and get the best of everything.

For example, if you determine that your processor becomes unstable at 1880 MHz (188x10), and the maximum speed for your agp/pci (with dividers) is 144MHz, you could set the fsb clock to 144MHz and the multiplier to 13, giving 1872MHz (a tiny drop in cpu speed) which would bring everything back into line and synchronise the north and south bridges with the cpu.

bigdawgfoxx

10-05-2003, 06:49 PM

oh...so do the agp and pci have to be set the same? and how do you know how to do FSB\4 or FSB\3...and what is the 200Mhz multiple of 4? what is the multiple of 4?

3rd gen noob

10-08-2003, 12:44 AM

lynx just reminded me about this topic two minutes ago in another topic, i had forgotten about it
well, he's up over 6000 now, the new drivers for video card seem to have done the trick...he's also getting approx 5800 just on the cpu overclock, so he's happy with that...
i don't have any ss's for proof just now, but i'll put them up when i get them

abu_has_the_power

10-08-2003, 12:50 AM

how do u lock down the agp to 66mhz?

3rd gen noob

10-08-2003, 12:51 AM

Originally posted by abu_has_the_power@8 October 2003 - 00:50
how do u lock down the agp to 66mhz?
look in your bios...you should be able to find an option in there

3RA1N1AC

10-08-2003, 12:58 AM

Originally posted by 3rd gen noob+7 October 2003 - 16:51--></div><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td>QUOTE (3rd gen noob @ 7 October 2003 - 16:51)</td></tr><tr><td id='QUOTE'> <!--QuoteBegin-abu_has_the_power@8 October 2003 - 00:50
how do u lock down the agp to 66mhz?
look in your bios...you should be able to find an option in there [/b][/quote]
also depends on which motherboard you own. locking the AGP/PCI/RAM bus speeds is a pretty new feature... up till a certain period, motherboards just couldn't do that at all so overclocking the FSB would screw up all sorts of things. hurrah for progress. ;)