Thread: Is there really any point?

I have a P4 3.4GHz HT (AMD equilivent 4500) overclockable to 4.08GHz (equilivent to AMD 5400) it is highly stable at 40'c and the performance difference is outstanding, without HT on 3.4 it is painfully slow, with HT it makes a great difference (33% increase i believe)

i dont know about the longterm effects but it hasn't done my system any harm, when overclocked to 4ghz i can convert 2 dvd's to divx, this is the only time when the cpu reaches 100% load outstanding performance

And by the way would anybody know how i would find the ram devider on my board?(see my sig for my board)!

Thanks.

Well i though sod it ill give it a bash.

Hyper transport frequency : x3
Ram timing mode: auto @ 333MHz
CPU frequency: 245
CPU multiplyer: x11
CPU voltage: 1.450v
Heres the results....

I take it that the MHz of the ram in the timing?
In my bios i have:
AUTO
200
266
333
400
And so on....
My ram is 400MHz but i have it in the bios on auto, any ideas what it would be best to set the ram at? the higer io overclock my cpu the more i lower the speed of my ram is this correct?

i believe the lower the ram timings the better the stability, the faster the ram timings obviously the more performance

you can work out what your ram timings will be by multiplying the multipler by the ram speed (BUS200mhz x MULTIPLIER14 = CPU2800MHz) you multiply the ram speed by 2 which gives your bus speed so the slower the ram speed the higher multiplier you will need to run at

i sense im wrong somewhere please correct me but its something like that

Originally Posted by newpaws

you can work out what your ram timings will be by multiplying the multipler by the ram speed (BUS200mhz x MULTIPLIER14 = CPU2800MHz)

I dont know if its just me but I have always referred to the RAM timings when talking about CAS, RAS etc.
Everything runs off the FSB ie FSBxCPU multiplier = cpu speed, like you put. The FSB also has its own multiplier nowadays giving you the 400mhz, 800mhz 1066mhz etc FSB speeds but your CPU and RAM doesnt use muliples of this "extended" FSB. They only run off the original FSB.

Originally Posted by newpaws

you multiply the ram speed by 2 which gives your bus speed so the slower the ram speed the higher multiplier you will need to run at

RAM dividers are used the same as multipliers, they are just multiples of the FSB only with a division thrown in!
ie 100mhz FSB with a 5:3 divider is 100/3 and the answer *5 = 166mhz

Ive personally always speak of 400mhz FSBs and DDR333 etc and not 100mhz FSB and 166mhz RAM.
You get people who say they have a 800mhz FSB and 200mhz DDR ram which to me is mixing standards. Yeah, the RAM is at 200mhz but performance wise its as good as 400mhz, and more to the point provides the bandwidth 400mhz RAM would give.

The Athlon 64 doesnt have a FSB like other processors, but I think the numbers should still stand.
Your RAM, because it uses a multiplier/divider of the FSB is also overclocked. You say its set to auto which is at 333mhz. Maybe your motherboard is more clever than mine?! but its DDR400 RAM which seems to be set to the DDR333 setting when at auto. This is the best way for your overclock. I know on my motherboard my DDR400 set to auto will always remain at the DDR400 divider even when the FSB is increased. If this was the case for your board your RAM would be set to 490mhz instead of 400mhz.

I should also add that HT on a P4 is different to HT on the AMD processors. Its Hyperthreading on the P4 and Hypertransport on the AMD processors. HT on a P4 allows multiple threads of code to run at the same time whereas HT on an AMD is , I THINK!, a way of increasing the speed of transfer from the CPU to the memory controller. Im not sure if Hyper transport is only available on Athlon 64 CPUs but the memory controller is on the CPU on Athlon 64s and not on the northbridge as in traditional systems, thus allowing quicker data transfer. HT on an Athlon 64 is effectively removing the need of a FSB, hence my earlier point about it not having one!

Ive found that a 20% FSB overclock has resulted in my benchmark scores increasing by 20%. If thats the same for everyone else its gotta be worth it. It does reduce the lifetime of your system but most silicon chips will probably out live many of us, certainly is own era of technology. I think the killer nowadays is over volting. My CPU (P4 northwood) was a nasty one for this. Overclockability (!) decreases then one day its knackered.

For me its a case of not being able to play BF1942 at 1024x768 if I dont overclock so I suppose it depends on how much demand you place on your system.

While, yeah, in the long run (say 10-15 years), overclocking probably is not good. However, who really has the same computer for 10-15 years? Then take those people from that small group who would overclock it.

Overclocking becomes more hazardous when people start applying too much voltage. A little extra voltage can help give the CPU or RAM that extra kick to keep firing, but going too far above that can cause more heat than the cooling can handle. In some cases, notably the Pentium 4 (specifically the Northwood core) may simply fail, regardless of cooling if a voltage threshold is exceeded after a point.

RAM can be tweaked and overclocked. They are linked together, but newspaws has the relationships mixed up with some FSB thrown it. RAM will run at a tighter latency (lower number) with a lower speed than at a higher one. Conversely, a looser (higher number) can allow for a higher speed. This assumes that you're being held back by the RAM in the first place. Overclocking can be fine-tuned by fiddling with these settings.

It's not mixing standards to say that you have PC3200 (DDR400) and an 800MHz FSB. Look at a Pentium 4 system with an 800MHz FSB. For optimal performance, it needs to be equipped with PC3200. You can go with higher-speed stuff, but you won't render any benefit unless overclocking. You can go lower, but then you have to run the CPU slower or run the memory at a divider. If the CPU loves memory bandwith, as is the case with the AthlonXP's (although the P4's did too), this will kill it. People that have a P4 sporting an 800MHz FSB are correct. However, in this case, there are a couple of things you need to understand.

1.)PC-DDRxxxx relation. I'm going to assume you've read that for the rest of my post.
2.)The Pentium 4 (and Celeron's based off the P4 designs) utilize a quad-pumped FSB. The base frequency is multiplied by 4. To achieve and 800MHz FSB, you start off with a base frequency of 200MHz. Tie this in with PC3200 (as explained in the link) and you get your matchup.

HyperTransport is usually denoted as 'HTT', while HyperThreading is denoted as 'HT'. HyperTransport is a bus, while Hyperthreading makes use of the long execution pipeline of the Pentium 4 design to process additional data when parts of the pipeline aren't being used by another process. It's a half-assed multi-CPU execution. The Core 2's design is about efficiency, whereas the P4 was about scaling to high clock speeds. Core 2 doesn't have the longer pipelines, so while Intel probably could implement HyperThreading, it wouldn't be as optimal. There's also a new CPU war, and that's about stuffing as many cores onto a die as possible.