I've got my 2600+ running @ 250x10 right now with AGP locked at 66. I know that the 250 FSB is helping things in the memory department, but would I be better off dropping the FSB some and increasing the AGP up to something like 70 so that the AGP and PCI busses are overclocked? Or is the highest FSB regardless of AGP/PCI setting the most important factor in performance?
High FSB/Locked AGP
- Thread starter thebiz
- Start date
This is the correct answer, and locking the AGP/PCI also insures the stability of your system and the safety of your data. Peace.
Thanks for your answers, both of you.
Is this true of the pci bus as well? (The pci bus is set to half of the agp bus, right?)
Is this true of the pci bus as well? (The pci bus is set to half of the agp bus, right?)
ive played with OCing the AGP speed on my NF7-S saw gains....but they were so small it was probably just a better run in the benchies I was running (which is what I fully believe) but as far as OCing your PCI bus goes. thats just crazy I learned the hard way of corrupting 2 hard drives...it isnt worth it
Intersting topic. I first wondered this also back when OC'ing a celeron from 66 to 75 fsb, lol. I wondered if the Voodoo3 3000 would gain performance.
Soulkeeper
Diamond Member
- Nov 23, 2001
- 6,732
- 155
- 106
haha
yup increasing the agp bus seems to have little effect with Current video cards and games
they just havn't reached the point yet where the use of the full theoretical 2.1GB/s of the 8x AGP bus (which btw the pci bus is at 133MB/s which is doubled then multiplied by 8 to get agp 8x )
no need to increase the agp bus
it's probably like a 10 to 1 comparison or something crazy when talking about the effect of increases in mhz when compared to the FSB.
This will change tho, aventually video cards will probably be pulling 20GB/s or more through their interconnect and people will be saying that it's not being used when the video cards are only pulling like 12 or 14GB/s of it. It's just a matter of maximizing the breathing room of future technologies by keeping the standards/protocols/interconnects/etc. updated well before the actual components actually need or have the ability to utilize such techs to their full potential, agp 8x might have offerred nearly no noticeable advantage over 4x, but it was a move that cost next to nothing to do (comparatively speaking) and left the door open for all companies in this business to use the bandwidth if necessary in their future products.
i think it's possible we could max out agp 8x in the next gen cards (what i mean by this is having a more sustained and constant supply of data at or near the theoretical max of the 8x's bandwidth being attained quite frequently in real world situations), these new ones that just came out are probably close if not doing it quite often during gameplay/benchmarks as it is. But you have to know that it's hard to keep an agp bus fully saturated for any long period of time when it is mostly only responsible for the transfer of data to and from one memory location to another usually only when the video card needs/requests a transfer, otherwise it is probably mostly idle or atleast only working at a fraction of it's potential during game play when textures, and other vital game files aren't being requested to work on and transfers aren't needed.
guess you could think of it like an ftp server sending homework to someone over the internet.
they do their homework only when they get it and they probably aren't requesting more data till they are done with a good chunk of the work, can clear up more space, etc,.
/me imagines a mythical agp bus with like 20GB/s bandwidth to a vid card with over 1GB of crazy vid memory from the 4GB or more of installed system memory
yup increasing the agp bus seems to have little effect with Current video cards and games
they just havn't reached the point yet where the use of the full theoretical 2.1GB/s of the 8x AGP bus (which btw the pci bus is at 133MB/s which is doubled then multiplied by 8 to get agp 8x )
no need to increase the agp bus
it's probably like a 10 to 1 comparison or something crazy when talking about the effect of increases in mhz when compared to the FSB.
This will change tho, aventually video cards will probably be pulling 20GB/s or more through their interconnect and people will be saying that it's not being used when the video cards are only pulling like 12 or 14GB/s of it. It's just a matter of maximizing the breathing room of future technologies by keeping the standards/protocols/interconnects/etc. updated well before the actual components actually need or have the ability to utilize such techs to their full potential, agp 8x might have offerred nearly no noticeable advantage over 4x, but it was a move that cost next to nothing to do (comparatively speaking) and left the door open for all companies in this business to use the bandwidth if necessary in their future products.
i think it's possible we could max out agp 8x in the next gen cards (what i mean by this is having a more sustained and constant supply of data at or near the theoretical max of the 8x's bandwidth being attained quite frequently in real world situations), these new ones that just came out are probably close if not doing it quite often during gameplay/benchmarks as it is. But you have to know that it's hard to keep an agp bus fully saturated for any long period of time when it is mostly only responsible for the transfer of data to and from one memory location to another usually only when the video card needs/requests a transfer, otherwise it is probably mostly idle or atleast only working at a fraction of it's potential during game play when textures, and other vital game files aren't being requested to work on and transfers aren't needed.
guess you could think of it like an ftp server sending homework to someone over the internet.
they do their homework only when they get it and they probably aren't requesting more data till they are done with a good chunk of the work, can clear up more space, etc,.
/me imagines a mythical agp bus with like 20GB/s bandwidth to a vid card with over 1GB of crazy vid memory from the 4GB or more of installed system memory
A SCSI array can max out the PCI bus easy, but overclocking the PCI bus leads to data corruption so your only choice is to go 64-bit PCI if you want more bandwidth there...
aren't thier HD controllers designed for 66mhz PCI slots? if so woulddn't one of these placed in a 33mhz slot thats been overclock be no problem for the card?
This also brings up another interesting question. PCI and AGP busses may be maxed out, but I wonder about the overclock potential of the PCI-E bus??? I guess we'll know before long.
A 64bit PCI card will NOT fit in a 32 bit slot...it's not an overclocking problem, it's hardware. There is no reason to overclock the PCI/AGP bus.
While your facts aren't entirely correct, I'll agree with your conclusion.
There are controllers (Promise FastTrak TX2000/TX4000 & Ultra133 TX2) which are 32 bits but capable of running at 66MHz. That being said, you'll get the same stuttering by underclocking a PCI card @ 66MHz that you would underclocking @ 33MHz. If the timing isn't exact between the PCI bus, the controller and the HD you'll get a miss much like a cache miss on a processor - you'll have to wait until the next cycle to attempt again. The difference is that a miss on a HD is on the order of milliseconds instead of nanoseconds as it is with processors.
Cliffs notes: Best peripheral performance is always achieved when the peripheral buses are sitting pretty at their specified clock speeds, be it 33MHz, 66MHz, or 133MHz.
what do you mean?
http://www.hardwarecentral.com/graphics/screenshots/1733pci.JPG
64 bit and 32 bit have different connections.
Different connectors, but you don't have to use the whole connector.
32 bit cards that are PCI 2.2 compliant will fit in PCI64/66 slots. It's a fact, and I've done it. I wasn't just blowing smoke out my butt when I mentioned the FastTrak TX4000; I used it in the 64/66 slot of an ASUS A7M266-D (760MPX) board for well over a year.
EDIT: You also seem to be associating 64 bit with 66MHz. That's not a good conclusion either. You can have cards that support any variation of 32 bit and 64 bit, and 33 Mhz, 66MHz, and 133MHz. As far as the bus speed goes, you card can even automatically fall back to another supported bus speed.
As an example, I could have a 32 bit 133MHz capable card. I could also have a 64 bit 33MHz card.
I said that 64-bit cards wouldn't work in 32-bit slots, NOT vice versa, and at no point did I confuse 64 bit with 66MHz. Stop putting words in my mouth.
Guy's asking about 33 vs. 66 MHz, you reply with something about 32 bit cards vs 64 bit cards. Can't get much plainer.
I said the only thing that stood a chance of maxing out a 32 bit connection (at 33 MHz) was a SCSI array, BUT that many of them used 64 bit cards, so it was a moot point. I then said that these cards weren't compatible with 32 bit, and that it was different hardware, not just a different clock speed that made them incompatible (hence my quote, "it's a hardware problem, not an overclocking problem")
A 64-bit PCI card will fit in 32-bit slot and actually work in 32-bit mode. That's the beauty of evolving standards.
Overclocking the AGP bus isn't good for much. But overclocking the PCI bus can give you substantial IDE hard drive performance increases.
not the full size ones
http://www.hardwarecentral.com/graphics/screenshots/1733pci.JPG
but I guess you're right, they've changed the spec a bit since I was familiar with them.
yeah, nice reply, that's exactly what I mean.
Gee, let's actually put some brain power into this. If according to you a higher PCI bus will give you data corruption, than the higher PCI speed must be changing the way the hard drive works. This is called "overclocking" and it doesn't just apply to processors and video cards. Up to the point before you corrupt data it gives you a substantial increase in IDE performance, just like I said.
Any overclocking can mess up your operating system, especially memory overclocking.
You either have the balls to try it or you don't....
Overclocking the PCI bus would only help if you were already saturating the bandwith of it. I have, for example, a dual athlon rig which, due to a somewhat inefficient 32-but PCI implementation by AMD, rather bandwith limited. If I up my FSB(no PCI lock on this puppy, so PCI speed overclocks proportionally), then I actually get better HD benchmark scores. However, this is with a 3 drive RAID 0 array, which is one of the few things that sits in your PCI bus that will saturate it.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter