Articles: Mainboards
 

Bookmark and Share

(7) 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 ]

FSB Strap

Once we got our MSI P35 Platinum mainboard to run stably at 490MHz FSB and pass some tests, we decided to investigate how efficient FSB adjustment with the jumpers could be. You cannot really reduce the FSB frequency with the jumpers, as even if you set it to 200MHz, the board will still boot at the nominal 266MHz FSB. Once the jumpers have been set to 333MHz FSB, things remained unchanged at first glance, only the CPU bus frequency increased from 266MHz to 333MHz, i.e. got 67MHz higher. The CPU remained stable until we hit 490MHz, however, the results of the Everest tool showed that with the jumpers set at 333MHz the read speed was 1.5 times higher than at 266MHz setting, provided that all other parameters including the memory timings remained unchanged.

We decided to check out the read speed in the entire supported interval between 333MHz and 490MHz with 10MHz increment. Our tests showed that there is no performance drop whatsoever, the obtained results are lining up ideally. Almost the same results were obtained on Asus P5K Deluxe mainboard – no performance drop (for details see our review called Asus P5K Deluxe Mainboard: Second Encounter). After that we set the jumpers to reduce the FSB frequency to 266MHz and repeated the tests, this time in the interval between 266MHz and 490MHz. The results turned out astonishing.

Look, the change of the FSB frequency through hardware has no effect on the performance – both lines coincide ideally. However, between 440 and 450MHz FSB the system performance crashes down if the FSB jumper is set at 266MHz, although in case of 333MHz FSB jumper setting, the system performance keeps growing stably. More thorough investigation revealed that the breaking point is the transition from 447MHz to 448MHz. Read speed equals 8538MB/s at 447MHz, but drops to 5594MB/s at 448MHz.

Keeping in mind that hardware FSB switching only increases or reduces the starting frequency with 67MHz increment, we assumed that the maximum frequency when the performance of the memory subsystem drops down dramatically, may also be shifted with the same increment. In other words, it is possible that when you overclock CPUs with the nominal 200MHz FSB speed, the performance will also drop so dramatically at around 381MHz frequency (448-67=381). We confirmed our supposition during the tests with Intel Core 2 Duo E4300 processor. The read speed reached 7091MB/s at 381MHz FSB, but dropped down to 4732MB/s at 382MHz.

Actually, there is nothing catastrophic about the above described cases. For CPUs with 200MHz bus that can overclock to 382MHz FSB and higher, you just need to set the FSB frequency jumper to 266MHz in order to push the dangerous limit a little farther. The same is true for CPUs with the nominal frequency of 266MHz: you have to set the jumper to 333MHz FSB and you can stop worrying about overclocking beyond 447MHz frequency.

However, there is another danger that awaits us here. Our experiments proved that by changing FSB setting from 266 to 200MHz we lower the threshold from 448MHz to 382MHz. Therefore, by changing the FSB setting from 266 to 333MHz we can obviously expect this notorious threshold not to vanish anywhere but to get shifted to 515MHz (448+67=515). Unfortunately, at this time we don’t have a CPU at our disposal that could overclock to 515MHz and up, so we cannot confirm this supposition experimentally. However, the probability of such outcome is pretty high.

Overclocking on MSI P35 Platinum mainboard puts a lot of popular processors into the so-called risk group. These are the CPUs with 7x clock frequency multiplier, because they are the ones that can hit 515MHz+ frequencies, such as Intel Core 2 Duo E6300, Intel Core 2 Duo E6320and the new Intel Core 2 Duo E6550. This is a serious drawback of the new MSI P35 Platinum mainboard.

 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 ]

Discussion

Comments currently: 7
Discussion started: 07/07/07 09:05:29 AM
Latest comment: 10/22/07 03:14:09 AM

View comments

Add your Comment