LinkBack URL
About LinkBacksFair enough but I'd imagine the PWM's aren't very happy with that since they have no heatsinks.
Fair enough but I'd imagine the PWM's aren't very happy with that since they have no heatsinks.
I wonder what CPU? E6320 with 266FSB or E6550 with 333FSB? Try running a 200FSB CPU and see how "good" it is. Try BSEL mod (nope, Gigabyte barfs at that!)Originally Posted by Webby
On the G33 not being great overclockers I would tend to disagree the Gigabyte G33m-DS2R has been seen at 600MHz FSB and mine is happily plodding along at 429Mhz as I type this all at stock volts (no idea what my boards limit is, this is as high as I have taken it... so far)
Will be taking it a bit further tomorrow maybe, currently 3GHz may see if 3.2Ghz is achievable, will have to wait and see!
Yes, mine (same mobo) was crap, would not reliably keep my E6400 at 400FSB even when clocking 6x400 with any amount of tweaks. My IP35 runs it 400FSB at default and at 8X multi. I had an Asus P5K-VM which was just as "poor". Neither have an FSB Strap but the P5E does and the new one also... so should clock fine with 800FSB processors.
See "My System" for all my spec, I have an E6550.
Well as with all components mileage will vary my board is happily running at over 400 FSB and as I have said I have seen forum articles with people having the board running at 600 FSB. Not to say that theirs and mine are not the exceptions and the rest are rubbish of course!
I rest my case... from 333-400.. WOW.. not... Tell me its awesome when you can clock an 800FSB CPU (E2160/E4300 etc) to 400FSB (even with lower multi)... then I will be impressed. Maybe I was unlucky but neither of my G33 mobos did that reliably... my IP35 does 24/7....
Again, I shall say the G35/G45 have FSB Strap setting so are worth the trouble... but jeez they are overly expensive!!
So your saying that because the CPU is designed to run at a lower FSB that will have a detrimental effect on the FSB achievable by the chipset? I can not see why that would be (I can see that a lower clocked CPU may have more stability problems at higher front side bus which may be exacerbated by the chipset but thats all) and if the chipset is designed to run at 333MHz then if you run it at 400MHz independent of the CPU you are still only overclocking the motherboard by the same percentage.
If I am wrong then please explain it to me or point me in the direction of somewhere that I can read about it.
Oh and look here for some impressive FSB figures on a G33M-DS2R
E8200 doing 600x8 on air with 1.488V - XtremeSystems Forums
that would be a pretty heafty overclock you have to agree but again that is a 333MHz chip so maybe you wont be at all impressed.
Here is a 266 chip going over 400 (scroll down a bit)
http://forums.hexus.net/pc-memory/12...g33-board.html
Again not saying that they are amazing overclockers but they can overclock and dismissing them out of hand is a touch narrow minded.
I read your post several times and it kinda reads like a sulky child to me but thats just me!
Don't know if that was the intention but that's the impression it managed to give and if that wasn't intended well.. I didnt write the post!
Sure, I'm impressed with 600FSB - that is amazing but I'm just saying that my TWO G33 mobos didnt clock well with chips that are fine in P35 mobos. YMMV, but there is known issue with lower FSB chips (800s) - Google FSB Wall if you want info. Having a Strap Option like this new mobo is a BIG plus. Hence why I was saying this is a good candidate (in case you missed that in getting all defensive over the precious G33s - that's just as "narrow minded" and "dismissive" as you accuse me of being).
333 to 400 is not impressive.
333 to 600 IS impressive.
Was that clear enough for what does and does not impress me?
600FBS, thats bloody superb
I think the guy has an exceptional board. Even he seems to think that (hence why he won't get rid!!). IMHO he's a lucky git!
My intention was not to appear sulky I just don't understand where you are coming from in respect that if it is the chipset that is limiting the FSB in your case then why can mine easily exceed the 400 mark?
Now I'm not saying that in your cases the chipset didn't give up at or before 400MHz just that I don't understand why the chipset and its FSB wall should be effected by the CPU, that would imply that this wall was caused by the CPU and not the chipset or is determined by a combination of both. If I am missing something vital that differentiates 200MHz FSB chips and 266/333MHz chips (latency or something similar) then please point it out, other than the fact to get them both to 400MHz is obviously a different percentage overclock but only for the CPU not the chipset which is stock at 333MHz in both cases.
Also the 333 -> 600 FSB is an 80% overclock so for the same overclock with a 200 chip would only take it up to 360, but obviously the numbers don't seem as impressive.
The Q6600 example I linked to as well (266 -> 437 FSB) is a 65% overclock and since it is a quad core twice as much information is potentially being passed through the FSB so you would expect the "Wall" to come sooner would you not? Just something else that came to me not implying that it is the case just what seems reasonable to me.
Once again if you can explain to me why this should be the case or can link me to somewhere which explains it (rather than just telling me to google something, which I did and couldn't find any thing really useful) then I will happily read it and hopefully that will answer all my questions.
Hopefully this post does not come across as sulky or in any other way than that in which it is meant to and that is, I would like to understand the situation to further my understanding as the reasoning for your argument currently evades me. Once more I am not saying that in your cases the G33 chipset is not the issue I just want to understand why the Wall occurs earlier for CPUs with a lower base FSB.
All I was saying (or trying to) was that a 333-400 overclock wasnt impressive.
Additionally I was saying that a Strap setting helps overcome the the FSB Wall that many people encounter (particularly with low FSB chips (E2xxx/E4xxx)). It may be simply (not that I know for sure) that the lack of FSB Strap is why I find these G33 mobos "so poor".
Nope, your post was definitely sulk friendly!!i wasn't trying to offend you (really)!!
I get that but it was 333 -> 429(29% OC rather than 333 -> 400 20% OC)
Oh and out of interest at what % does an OC in your opinion become impressive? Just curious
Anyway that aside, I have read a bit about straps and they appear to change the latencies of the chipset depending on the FSB (apparently these are predetermined by Intel). So as you increase the FSB it slackens the latencies (just like you would for memory) now these changes occur at set intervals so for an 800 strap it would be at those latencies until your FSB reached 266 then it would change to the 1066 Strap, 1333 Strap, 1600 and so on. Maybe 800 CPUs require tighter latencies and the change to the 1600 strap is a step to far for them on a G33 chipset? If that is the case it would suggest that the 266 and 333 chips have more tolerence for slacker latencies (as they start at them to begin with) and so do not suffer from this problem until a higher strap is reached(?). That would explain my 333 CPU (and the 266) being able to exceed the 400 FSB mark while your 800 can not on a G33 chipset.
This is of course all speculation and assumption based on what I have read but it kind of makes sense.
Oh I dunno... lets say .. 50% ??I get that but it was 333 -> 429
Oh and out of interest at what % does an OC in your opinion become impressive? Just curious
Anyway that aside, I have read a bit about straps and they appear to change the latencies of the chipset depending on the FSB (apparently these are predetermined by Intel). So as you increase the FSB it slackens the latencies (just like you would for memory) now these changes occur at set intervals so for an 800 strap it would be at those latencies until your FSB reached 266 then it would change to the 1066 Strap, 1333 Strap, 1600 and so on. Maybe 800 CPUs require tighter latencies and the change to the 1600 strap is a step to far for them on a G33 chipset? If that is the case it would suggest that the 266 and 333 chips have more tolerence for slacker latencies (as they start at them to begin with) and so do not suffer from this problem until a higher strap is reached(?). That would explain my 333 CPU (and the 266) being able to exceed the 400 FSB mark while your 800 can not on a G33 chipset.
This is of course all speculation and assumption based on what I have read but it kind of makes sense.
As to Straps and how they work.. well I dunno.. but I do know that being able to set the strap in BIOS does help the overclocking of lower FSB CPUs.
Now with Intel launching E7200/E7300 (266FSB version of Wolfdale) I would be looking for a mobo that helps it RELIABLY get 400+ FSBs. Given previous experience with G33 I will be looking to G35/G45 or Nvidia's new version of 630i (with GF8200 VGA onboard)
Well Back to this again
I know its not quite 50% (only 44%) but there we go still all at stock voltages.
It booted into windows fine at 500 FSB but wouldn't run Orthos for any amount of time would still surf the web etc though. Think the limiting factor is probably my RAM rather than the chipset or CPU 800MHZ Corsair XMS2 running at 1000MHz is I think a touch optimistic!
Anyway still not a bad overclock and it could possibly go higher with some voltage changes and some different memory all I need now is the £40 for better RAM to test it with.
Well after some fiddling about I discovered that my memory is indeed my limiting factor but at 2.2v reaches DDR 998. So I played around and am currently testing my CPU running at 499 FSB had to increase the CPU voltage (went up 0.05v) and has been stable for 35mins now will lower the voltage a bit later to get it as low as it will go while still stable.
So I guess that is a 50% overclock
OK, you got me.. that's good work, and I will take all credit for motivational push
Interesting comments about FSB speed and NB strapping. I hadn't heard of that before. This issue probably affects gamers (I assume, most of you on this forum) more than multi-tasking pros (such as myself.)
With multi-tasking, faster FSB's have been shown to increase performance slightly. This has something to do with the memory manager, and it helps even if the latencies must be higher. (note that Intels' new workstation chip uses a 1600 FSB specifically to overcome memory bottlenecks)
Anyway, I just joined this forums to comment on the chipset OC issue. I think that a large part ofthe limitation some may have found, with typical mATX chipsets, is not the chips themselves but rather the build quality of the mobo. For instance, the G33m-DS2R is indeed known as a great OC mobo (mine is very happy at 400) but some folks had terrible problems with them. It wasn't the NB, it was the thermal pad Gigabyte strangely chose to put under the NB chip. Several folks removed the pad, lapped & used compound, and dropped NB temps 15-20 degrees. (I improved mine 22 degrees, by also adding a better heatsink)
I imagine that up until recently, the manufacturers haven't expected a lot of OC'ing on mATX, so they cut costs and did some things a little sloppily.
this should change now, as more gamers and pro musicians / video guys are using mATX (which happily fit in road cases.) The Asus P5Q-EM HDMI, for instance, is being specifically advertised as an overclocking mobo.
There are currently 1 users browsing this thread. (0 members and 1 guests)
SCAN.care@HEXUS
SCAN.care@HEXUS
Posting Permissions
Reply With Quote
