• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

I know you OC an Athlon/Duron with multiplier and not FSB, but WHY?

D

DesignDawg

Diamond Member
Hey all,

I was just thinking about this. Since the original athlons, it has been well-known that they don't OC via FSB worth a crap. We found the golden fingers, and now, with the durons and such, we have the pencil trick. We do all of our OCing via the multiplier. But why is this so? Why WON'T athlons overclock well via FSB adjustment? Is it because of the double-pumped FSB? Or is it the chipset? Obviously, the chips can handle higher speeds than they are rated for. But why are they limited to HOW they can attain that speed?
--And, could you extrapolate this and figure that, because Intel chips OC so well via FSB, that they would be BEASTS of overclocking if we could adjust the multiplier? I seem to think not, but I don't really have the answers....

Thanks,

Ricky
DesignDawg
 
B/C the EV6 bus is so much more complicated than Intel's 100/133Mhz buses. There's less of a tolerance for overclocking on the DDR bus. It will probably be even worse for Intel's "quad-pumped" bus.
 
Actually the EV6 doesn't have anything to do with it, but rather AMD and VIA's current implementation of it. Look at AMD's roadmap - they've announced the EV6 running at 400MHz next year, which is the maximum spec for the EV6. The Athlons can run at bus speeds up to and over 133Mhz. Seems like AMD and VIA got together and decided that instead of adding bus locking to the CPU, they'd add it to the system chipset and artificially limit bus speeds to around 110-115MHz max on the current chipsets. Over in Japan, there's a device called the TurboPLL that overrides the clock generator on many system boards. With one of these installed, Athlons have been seen to run on up to a 155MHz bus (310MHz DDR). Heck, BX boards with Pentium IIIs are hitting over 200MHz bus speeds with one of these things. Only downside is that there's soldering and moving of crystals on the system board and the TurboPLL costs around US$100 - not worth it for most people.

If you stop and think about it though, it makes sense. Why wait and pay for a DDR SDRAM system board if your current system board can run at 133MHz/266MHz DDR? Anand's review showed that DDR SDRAM doesn't provide a huge jump in performance. Also, consider all the rumored announcements of a 266MHz Athlon chipset from VIA that runs with regular SDRAM. If it's true, I'd suspect that VIA will simply "unlock" the current chipset. And isn't it curious that this chipset is being released right around when AMD's 760 DDR chipset is supposed to be out?
 


<< Actually the EV6 doesn't have anything to do with it, but rather AMD and VIA's current implementation of it. >>


Well if it's AMD and VIA's implementation, then it is still the bus 😉 They are just holding back for the 760 and VIA &amp; ALi's 133/266 offerings. They wouldn't want their 750/KX133/KT133 boards hitting 133/266 before the new stuff gets here 😉

But there are tolerance levels to deal with. Whereass about 95% of BX boards do 133+ with no problems, most 750/KT133/KX133 boards have trouble coughing up 115MHz. OTOH, some sites have demonstrated KX133 boards reaching 133MHz with no problems. So there is a lot of inconsistency floating around with these boards compared to Intel's BX. Also, Intel's BX was capable of 133MHz operation long before the first 133MHz processor hit the streets. IMHO, its just the fact that the EV6 is a complicated piece and its tolerance levels don't quite match mighty Intel (at this time) is what is holding the chipset back in its current form.
 
wetwilly..

i do not doubt what you have said however a bx board running at over 200Mhz you say?

a BX motherboard has a max agp divider of 2/3 so a video card would be running at 133Mhz. Most cards have trouble getting to over or at to 89Mhz (66Mhz is the standard).

if you use a pci card the the max divder would 1/4 (with boards that support this over 133Mhz) thus at 200mhz the pci bus would be running a 50Mhz (33Mhz standard). I find it difficult to believe that agp or pci card (or the ide controller ) would be runing this fast!

Not to memtion ram running this high also. (on a BX motherboard these speeds are all tied into togeather.)

also these things (to the best of my knowledge) are set into the chipset and can not be changed!

If you have a link, i would like to be corrected on these things. 🙂

thank you. 🙂
 
NFS4,

Thanks for posting the reference to the 216MHz BX. Since you're being such a stickler 🙂 I'll correct myself - it has nothing to do with the EV6 spec but rather AMD/VIA intentional crippling of it as implemented on current chipsets.


<< But there are tolerance levels to deal with >>


It's not entirely a tolerance level issue either. People who'll cough up the money and solder for a TurboPLL aren't getting special system boards. They're taking off-the-shelf system boards and adding the special clock generator. There's no real way I've heard of to pre-test a board for use with a TurboPLL. Put it this way for the one I saw:

1) Off the shelf Epox EP-7KXA maxed out around 110-115MHz
2) Same Off the shelf Epox + TurboPLL maxed out at 155MHz

Nothing changed except for the addition of the TurboPLL.


<< Whereass about 95% of BX boards do 133+ with no problems, most 750/KT133/KX133 boards have trouble coughing up 115MHz. OTOH, some sites have demonstrated KX133 boards reaching 133MHz with no problems >>


Because most Athlon chipsets are artificially limited. The number of &quot;virgin&quot; Athlon boards that'll hit 133MHz is pretty small. In fact, my Abit KT7 can run my Duron 700 at 8.5x115MHz like a champ - unfortunately I can't keep it there because an older Maxtor drive chokes on the oveclocked bus. But speaking of the KT7, isn't it rather curious that when you look at Abit's FSB/PCI speed table that it doesn't kick the PCI divider back to 4 at 133MHz but rather at 136MHz? When was the last time you saw a system board that only had a PCI divider of 3 at 133MHz? More evidence for my conspiracy theory 😉 The other curious thing is that while a few of the KX133 boards can hit 133MHz all by themselves, I haven't heard of any KT133s doing the same. Perhaps that was an additional tweak from VIA during the KX133-to-KT133 revision?


<< Also, Intel's BX was capable of 133MHz operation long before the first 133MHz processor hit the streets >>


It's the same situation with the KX/KT133 now, only difference is that in most instances you currently need a TurboPLL to reliably hit 133MHz. Heck, with a TurboPLL, you can run old Athlons at 133/266MHz, and it's unlikely you'll ever see an official 133/266MHz Slot A CPU from AMD.

Cosmic Horror,

Ye of little faith ... 😉 There's a simple answer to your valid questions about the multipliers - I had them myself. With some clock generators there are two modes on control - through the pins and through the serial data interface, which is what SoftFSB does. Sort of like an Asus system board. You can configure it with jumpers or through their JumperFree mode. And much like the Asus boards, you get more options through JumperFree mode than the physical jumpers. The clock generator on the Epox I saw kicked the PCI divider back to 5 at 150MHz = 30MHz PCI bus. As for video, you'd either use PCI video or I'd guess some quasi-AGP card like a Voodoo3 that has an AGP form factor but doesn't use AGP features. I ran my Voodoo3 2000 at 100MHz in a BX board when I forgot to reset the AGP divider from 1:1 to 2:3. Didn't notice for quite a while and the board ran just fine.

The real problem with the TurboPLL is the price - by the time you pay to get thing and add the price of a system board, I'm not sure it's worth it. I suppose though, there's some bragging rights to be has from some of those absurdly high WCPUID screens on the 'net.
 
Those folks over in Japan are MAD crazy!! 🙂 They O/C and supercool like it ain't no tomorrow...I don't think the average American/Canadian is gonna spend an extra $100 for a PLL chip 😀
 
I have long had suspicions similar to WetWilly's. The info about the TurboPLL seems to coroborate his theory.

Modus
 
Yep, WetWilly is right. :Q It's artifically limited in some way. Which really sucks. But with boards running at 133(266) spec, we should be able to go to around 145-150(290)-(300), if you figure about a 10-15% tolerance.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top