bus speeds and overclocking

tigernut

tigernut

tigernut

Associate
Joined
5 Oct 2007
Posts
8
hello all!

Question for you, to iron out a little row me and a mate are having...

does the bus speed of the motherboard limit the overclocking ability of the board to such an extent that it can make overclocking the components of the board (CPU, GPU etc) a totally pointless exercise?

he is under the impression that if he owns a 1333 capable board, everything on that board MUST be 1333 in order to get the PC to communicate effectively and make overclocking a future option for him.

i on the other hand, whilst recognising the advantages of using the fastest components possible for his motherboard, feel that its only one part of the puzzle, and that he can happily use 1066 or 800 speed parts in it and still be able to overclock them.

as far as i understand it, overclocking is speeding the internal parts of the CPU, or RAM or GPU etc is it not, and where as the bus speed of the boards nothbridge say will have a powerful bearing on how fast the motherboard can pass that information to the next part of the PC that matters, the overclocking of the individual parts will still have advantages as the calculations done within the component (be it CPU, GPU etc) will still be calculated faster. Boards limit cummuncation speed between parts, but dont limit the speed you can push those parts because of it. other factors on the boards BIOS may limit what can be done, but bus speeds arent really one of them, they only limit communications between parts

hopefully that makes sense. may general explanations may be a little off the mark or poorly done, but the jist of what i am getting at is right i think. am i right on this one as he has a bee in his bonnet over it and so for the time being i have been pleading the 5th, even though tbh i think he is waaaay off base on it all.

your thoughts please gents and ladies...
 
A single component e.g CPU having a clockspeed of say 3.6Ghz at 360x10 will give high performance in things like SuperPI where only CPU calculations are needed, although 3.6Ghz from 400x9 would give a higher external bus and arguably better performance.

The best thing would be to try several different settings and see what gives you the best performance in benchmarks.
 
Sorry, I missed the point, and thought that was the core, oops :D.



He can use a 800 or 1066 bit of kit with OCing being perfectly fine, I have a 1600Mhz capable board and am using a 1066Mhz CPU and I can overclock perfectly fine.


All speeds are "Up to" speeds. Although with OCing, you can go higher.
 
Voxination said:
Sorry, I missed the point, and thought that was the core, oops :D.



He can use a 800 or 1066 bit of kit with OCing being perfectly fine, I have a 1600Mhz capable board and am using a 1066Mhz CPU and I can overclock perfectly fine.


All speeds are "Up to" speeds. Although with OCing, you can go higher.
Click to expand...

so, looking at the last sentence then from you, the mHz speeds quoted on boards can still be pushed further than the quoted speeds? if say then, we had a 800mHz board, wacking in some 1066mHz would be pointless as it will only run in the board at 800mHz anyway, however, overclocking the ram you may well be able to take it to higher speed, perhaps to 1066 anyway??

its a slight detour off the origonal point for me that though. his arguement is because the northbridge bus speed on a motherboard is set at 800mhz say, ther is no point in overclocking the parts you put in that board any further. my counterpoint is that though it might be true that the northbridge will only pass information across at 800mHz, overclocking the parts attached to the board is still of benefit as the calculations done within the parts in question (be it CPU, RAM or whatwever) are still achieved faster, and so are of benefit.

...again i could prolly explain that better but hopefully you get the jist of it everyone
 
Ok, NOW I think I get it lol :p.


Couldn't you just increase the FSB to stop an FSB bottleneck? That would void both arguments, unless the FSB is locked at 800Mhz.


If someone who has an Extreme/Black edition CPU would be willing to help out (or simply drop the multiplier of the CPU on a normal edition) then the point could be solved by:

- Leaving all settings at stock with the 800Mhz FSB and run a benchmark.

- Use Memory dividers to increase the memory speed from say 800Mhz to 1066Mhz
- Increase the CPU multiplier to achieve a higher CPU clock speed
- Run a benchmark and see if the results improve



If I get some time free later I'll give it a go for you, however I think that having individual components clocked higher will still yield an improvement in areas where only that component is needed, for instance RAM is not used that much when running SuperPI or performing physics calculations on the CPU, so the FSB speed shouldn't make a difference.


I'm with you, it would still yield a benefit.
 
Your post nearly makes sense, but I fear it is worded rather abnormally. The issue is one of bottlenecks I take it? That if the processor and ram are both running at the same speed, then overclocking either one will not help?

The 1333 or 1066 speed boards is where the flaw lies. Components aren't rated like this, 677MHz ram is in no way an exact match for a 333 fsb processor. Motherboards are also not rated like this, you can't say it's not worth overclocking the processor because the board is cheap. You can say that overclocking it will be more difficult, and it wont reach the speeds it would with a better board.

A ddr2 motherboard rated to run 1066mhz ram only means that it is able to run memory at this speed. Whether the memory you're using can actually do this is a seperate issue, you can have a good set of 800 mhz ram which will happily run at 1066 your motherboard, or more likely you'll get 950 or so out of it.

Overclocking/underclocking is changing the speeds of the components from stock. If you have a very slow processor and very fast ram, overclocking the ram further will do no harm but also offer no benefit. In most cases it's difficult to tell which the slowest component/limiting factor/bottleneck is, and calculating it is impractical. This is why you run a benchmark, then overclock/underclock a component and repeat to see if it made any difference.

Say you correctly identify the slowest part, and proceed to overclock it. Performance will steadily improve, then start to level off, and reach a point where no more performance from this component makes a difference. Then you get to move onto the next component. If you tell us the components you're using and what you're using them for we can have a fair guess at where the slow point is, from experience as much as anything else.

The final point is that bottlenecks are use dependent. One task may be limited by cpu, another by ram, a third by graphics card. This points at running each part as fast as you can being the optimum, which kind of makes sense anyway.
 
JonJ678 said:
The final point is that bottlenecks are use dependent. One task may be limited by cpu, another by ram, a third by graphics card. This points at running each part as fast as you can being the optimum, which kind of makes sense anyway.
Click to expand...

That more or less sums up the whole argument I think. Having a limited FSB will affect certain tasks, but not others.
 
Originally Posted by JonJ678
The final point is that bottlenecks are use dependent. One task may be limited by cpu, another by ram, a third by graphics card. This points at running each part as fast as you can being the optimum, which kind of makes sense anyway.

Voxination said:
That more or less sums up the whole argument I think. Having a limited FSB will affect certain tasks, but not others.
Click to expand...

thank you you two. though i feel you both may of danced around my question rather than right on it (not least because of my somewhat dodgy description of what i was on about i know) i think you may of both pretty much hit the mark with the overall input from you.

most interesting was the discredit given to the scores (800, 1066, 1333 etc) that motherboard companies give to their products in the first place. marketing people would quickly point out that joe public likes the simplification of the situation most of the time and i suppose thats true generally speaking, but it can lead to misunderstandings of such things, something i feel i might of been guilty of myself a little on this one, and something my mate is certainly guilty of as he has thrown hundreds of pounds at components based purely on that magic 1333 number. lol

it reminds me of a similar situation with RAM. people take it that the mHz speed is the be all and end all, and yet if you ask joe public about latency (well, none forum PC world fan joe anyway), they often havent a clue what your on about. as we know, as an example 800 speed ram isnt always faster than 667 if the latency scores are beyond poor. life isnt as always as simple as they would have us believe is it hehe

anyways, side tracking somewhat now so i shall shut up. thanks a lot to both of you for some food for thought and i'm sure it will leave me mate scratching his head over it all - for thats its more than worth the time taken hehe. not only will it have him going away and putting his head in some more books/webbys addys, its had me reminded not to oversimplify myself, something i was treading near to in my origonal explaination to him.

incidently...

anyone wanna buy a still sealed brand new (i bought a copy and then someone got it me as a gift) game of ARMA II? if you ever want to see a game that will make a laughing stock out of a DDR3 no expense spared PC as easy if not worse than my humble DDR2 8800GTS ageing model, thats the game to do it with. it seems the better the system you own, the harder it will kick you in the ass!! £19 if anyone wants it.


oh lastly lastly and in the interest of saving me mate a few quid...

i cant seem to find a good way to explain this one as my start to this tread no doubt shows, and its near the same question as the origional anyway...

he now has 1333 everything pretty much, RAM CPU MB and so on. however, his GPU is running DDR5 and he is bothered about this fact to the point of selling off his DDR5 4870. can someone explain to him how its OK to have such a card in his PC and that it isnt ruining the communication speed between all the rest of his PC? i think he is running the following kit:

N board possibly a 790i
6850 CPU
DDR3 4gig
SATA2 connections for his 7800RPM drives
an unbranded PSU
saphire 4870

he wants to ditch the 4870 cos of the DDR5 on it.

:confused:


for me i would get a board with faster ports for his drives and ditch them if i was gonna get rid of anything rather than ebaying off a perfectly OK GPU. DDR5 on his GPU board wil have no harmful effect on his system will it, yet asking me to explain why is proving tricky. i have basicly said that the DDR5 speaks solely to the GPU and so isnt even playing a part it talking to anything else anyway, but he aint listening i dont think lol.
 
You must log in or register to reply here.
Back
Top Bottom