Been reading up recently on the athlon 64 architecture and how to overclock it and feel I have a very good understanding now but have been confused by one issue. Athlon 64 mobo reviews tend to lower the cpu multi so that they can see how high they can get the HTT which makes sense as they want to test how far the mobo will overclock without being limited by the cpu. However I've seen articles suggesting to lower the cpu multi while increasing the HTT in order to get the max cpu speed but don't quite understand what the point of this is. Is it not best to leave the multi at stock and just up the HTT. I suppose having a lower multi gives you more granular overclocks as each HTT increase would be a smaller cpu clock speed increase than if a higher multi were used. What am I missing?
Those articles could possibly be saying the same thing as the mobo reviews, to see what the max is on the HTT. So you don't seem to be missing anything
you don't want to increase the HTT multiplier to high. for me when i am overclocking an A64 system you need to keep the HTT multiplier in spec with the motherboard while increasing the FSB multiplier. I see alot of people confuse the HTT multiplier with the FSB multiplier. HTT is for the hyper transport and for most boards the thresholds are usually either 800mhz or 1000 mhz. for example, take my current system. opteron 165 @ 2.4GHZ to achieve this i had to lower the HTT to 3x, the FSB to 265mhz and lock the ram base speed at 166mhz. if i leave the htt at the default of 5x it is unstable since the speed is way over the 1000mhz threshold. Each motherboard can handle differeing levels of oc on this bus but usually you wouldn't want to go more than 10% over it.
chill peeps All friends here. LCDguy your posts have illustrated what I have come across often in researching this topic and that is that the terminology used to describe various clocks etc in the k8 architecture are often used interchangeably which I must say until recently was very confusing. So to put it clearer I'm aware that the hypertransport bus (CPU to NB) should generally not exceed it's rated speed (800Mhz/1Ghz) by more than 10% and so when overclocking the "base frequency" it's neccesary to reduce the CPU-NB link multiplier from it's stock 4/5x value. In addition use of the term "FSB" is IMHO inaccurate and creates unneccesary confusion with regards to the k8 architecture. I believe the "base frequency" on which the cpu clock speed and hypertransport link speed are based has been refered to as HTT,HT and LDT which again underpins how one can get confused when reading articles from different writers. IMHO most of these articles could have easily been improved with some well constructed diagrams but alas we can't have everything. I'm beginning to think that Fr4nk was right and I simply misinterpreted one or more articles as indicating that lowering the cpu clock multiplier played a role in overclocking the cpu unless someone knows of a particular reason why lowering the cpu multiplier is beneficial in overclocking the cpu?
i think it depends on what kind of over clock your going for. but in my expeirnce i usually jusy bump up the base speed keep the multi at the max and lower the ram speed if needed.
How do you mean "depends on what kind of overclock you're going for"? Do you mean overclocking the ram without (or least not by much) the cpu clock speed in which case it would certainly make sense to lower the multi. For this thread I'm only interested in overclocking the cpu speed so I take that as another vote for me having misread the use of lowering the multi for cpu overclocking
Hehe, I was going to say, " have I read the thread question wrong? " As said I can't think of any reason to lower the CPU multi if you goign for a high CPU clock, sure you can lowe rit to get smaller steps but if your on a X10 multi with 290 HTT you'd have 2.9Ghz (Nice and stable), then you try 291 HTT so 2.91Ghz, it comes unstable, that could be the absolute limit of the cpu but it coudl be stable on 2.905Ghz (2905 Mhz), 5Mhz less. but finding out that would seem impracticle as its only 5mhz, which isn't going to make a signifficant differance against 2.9Ghz. So instead of faffing with the multi to find the max stable clock, just bump the vCore up (know your limit though). Hope this helps
Fr4nk, it seems like if you're going for max cpu clock then lowering the multi is only useful for the reason you outlined and I suggested earlier which is that a lower multi gives you more granular overclocks which allows you to wring out every last drop of performance. I certainly wouldn't care about finding an extra 5-15mhz so I will consider the case closed in that the multi should stay at max when overclocking the cpu. Thanks guys for clearing up the situation for me!
3 advantages from lowering multi - 1. allows you to find the max overall HTT ("FSB") speed for your board - Bit pointless really. 2. If you've got a crap clocking CPU but good clocking ram - Drop the multi and run a high/higher HTT speed and ramp up the ram speed. (ie ram does 300 HTT, CPU will only run 2.5ghz) 3. It makes things easier to "max" your overclock. Sometimes its necessary to run a lower multiplier (or even a 1/2 multi) to get the most out of the CPU and memory. (ie run a higher HTT speed, and lower multi maybe using a divider to drop the ram speed..)
Highland3r, points 1&2 make sense and are understood. Could you elaborate on point 3. How would lowering the multi allow you to "get the most out of the cpu"?
Say you had a 2.7 ghz capable cpu, 300mhz capable ram and your stock multi is x10, you could easily run 270 x 10 with the ram on 270 but your ram still has some juice left, so it woudl be ideal if you lowered the multi to x9 and knocked the HTT to 300; 300X 9 = 2.7ghz with the ram on 300mhz
I totally understand the issue of lowering the cpu multi with regards to maximising the ram and for getting that last 5-15mhz of cpu clockspeed as discussed above. Let me be clear for this thread I am ONLY interested in the effect of lowering the multi with regards to maximising cpu CLOCK SPEED and not the use of lowering the multi in general overclocking. I'd considered the case closed in that lowering the multi served no purpose for overclocking the cpu clock speed outside of getting the last few mhz due to the increased granularity of a lower multi. If the previous sentence is not correct then by all means I'm interested but other issues associated with the multi that don't concern cpu clock speed are for another thread
