Intel P4 800MHz FSB CPU Round-up - Overclocking
(Page 5 of 7 )
OVERCLOCKING
These new P4's have been said to be "the next c300a" or "the next P4 1.6a". You'll notice that both of the previously mentioned chips are the first chips released of a particular stepping or build. Basically Intel (and AMD for that matter) release a new core and then ramp it up. Fortunately for us overclockers Intel either speed bins or quota bins. If they're speed binning (by binning I mean runs a CPU off the line and places it in the appropriately rated speed bins) they perform a quick test and place it where it's performance matches. When they quota bin they simply place every CPU that rolls off the line in the bin for the CPU they need most. Bascially what I'm saying is that all cores of a particular stepping are alike. It's quite possible for a 3.0GHz chip to roll off the line beside a 2.4c. SOOO, to wrap up what feels like a very confusing paragraph, the slower the rated speed the more headroom for overclocking it should theoretically have!
Taking the above into consideration it would stand to reason that certainly the 2.4c and 2.6c should overclock respectably, and the 3.0GHz should stand a decent chance as well. Will that be the case? We'll find out.
The big thing that comes into play with these 200MHz FSB Pentium4's when overclocking is the memory you're using. With the PCI and AGP divider you can simply lock them at 33MHz and 66MHz respectively with no substantial performance hit and overclock the FSB until your hearts content. With the memory that isn't necessarily so. With the memory the higher DDR bus you have your performance scales right along with it. With a bus that is default at 200MHz you need some memory that can carry the mail to say the least. You can of course run your memory at a divider (typically 1/1, 2/3, and 4/5 are you divider options), but the name of the game is keeping your memory at the highest DDR bus possible while pushing your CPU to the maximum stable speed. We chose to use Corsair TWINX PC3200 matched memory (2x256MB modules) for our testing being run at 2.5/6/3/3.
So what did were we able to overclock to?
P4 2.4c Overclocking Results |
 |  |
CPUZ CPU Screen | CPUZ Memory Screen |
With the 2.4c we were able to push the CPU to a very respectable 3.522Ghz. The drawback to that was that we were forced to run our memory at the 2/3 divider. So while our CPU FSB was at a cool 293MHz, our memory was cruising along at a much tamer 196MHz. We still gain the sizeable bandwidth jump thru our system bus, our memory bus slowed us a bit. To have been able to run the 4/5 divider with this Corsair TWINX memory it would have been required to run at 234MHz and that is a bit too tall of an order even for this stellar memory.
P4 2.6c Overclocking Results |
 |  |
CPUZ CPU Screen | CPUZ Memory Screen |
The 2.6c did not let us down either, coming in at a comparable 3.536GHz. The big advantage here is that our bus was low enough to be able to run our memory at a more stressful 4/5 divider, leaving our memory at a damn respectable 217MHz.
P4 3.0GHz Overclocking Results |
 |  |
CPUZ CPU Screen | CPUZ Memory Screen |
The 3.0GHz checks in with the highest overall overclocked speed at 3.561MHz but pays the stiffest price of all in that the FSB to run at this speed is 237MHz which puts us just out of reach of running a 1/1 memory divider. That forced us into the 4/5 divider and left us at a less than impressive 190MHz DDR bus speed.
There is some memory on the market that can certainly handle a higher bus speed, and with the better memory we could have possibly reached a better divider/CPU speed match, but for the sake of this review we felt that the popular Corsair TWINX memory was a good choice. The TWINX we used was able to handle a 221MHz bus speed at these settings which in and of itself is an accomplishment for PC3200 memory.
It also warrants mentioning that the overclocked speeds in this round-up were attained at STOCK COOLING and the speeds listed are absolutely stable in this configuration. I don't mean stable as in "it will run 3dMark01 once", I mean stable as in it will run three loops of 3dMark03 while folding, and then will remain folding overnight with no lockups". I have run two of the three processors in a water cooling rig and in each have gained an additional 80-120MHz STABLE from each. So by no means take these as the end all be all, more as a representation of what's available.
Just because our CPUs (all store bought) hit these speeds doesn't mean that yours will, it also doesn't mean that yours won't hit higher, again, just a representation.
Ok, that's enough of this, none of this information means jack if the performance isn't there to back it up, let's get on with the benchmarks!
OVERCLOCKED BENCHMARKING
Alright, second verse same as the first! We're going to run you thru the exact same benchmarks we showed you for the default CPUs, but this time comparing them all at their overclocked speeds.
SiSoft Sandra 2003
SiSoft Sandra CPU Arithmetic
CPU Arithmetic Results |
 |  |  |
2.4c @ 3.522GHz | 2.6c @ 3.536GHz | 3.0GHz @ 3.561GHz |
This makes sense, with the SiSoft CPU arithmetic test it is relying solely on the raw CPU speed, so we see the ever so slight scaling of these similarly clocked CPU's.
SiSoft Sandra Multimedia Benchmark
CPU Multimedia Benchmark |
 |  |  |
2.4c @ 3.522GHz | 2.6c @ 3.536GHz | 3.0GHz @ 3.561GHz |
Again we see the same as with the arithmetic runs. SiSoft says they're benching the CPU standalone performance and they mean it. :)
I expect to see whatever gains we anticipate with the higher bus of the lower rated CPU's in the memory tests and the 3d benchmarks, let's go see if that holds water.
Next: Overclocking Continued >>
More Computer Processors Articles
More By Justifier
| Recommended by Dev Hardware |
|---|
|
