ong@d.cs.okstate.edu (ONG ENG TENG) (02/06/91)
Well, after reading my review on Knights of the Sky in rec.games.misc, one of our regulars asked me why I lowered by 386SX CPU speed down to 17.5MHz. Since I have discussed the CPU speed question on this net extensively, I felt obligated to post my answer. After upping my 386SX from 16MHz to 20MHz (without heatsink or fan), it worked fine, and still will if I choose to run it at that speed. But after many hours of continuous operation, my fixed-HD (I have a fixed-HD and a SyQuest removable where I run off most of my softwares, the fixed-HD contains most "standard" programs like system programs, word processors, graphics editor, terminal emulator, etc, only) my fixed-HD started giving disk-read error. To get rid of it, I simply do warm-boot (NO turning off the power). I don't think the problem is due to the 20MHz directly, but indirectly. I know that it takes more power to run at 20MHz, and my guess is that it somehow affects the disk. I think if I re-low-level format the disk, it would be all right. But since I don't have any extra SyQuest cartridge at this time to backup my fixed-HD, that will have to wait. I know I don't make sense, but bear with me. Computers are sometimes funny that way, and so must my reaction be. Thanks again for all the folks who contributed to the speed-up-my-CPU discussion. Some said they have upped it to 20MHz outright and has no problem since, some said I should not hold down my finger to the CPU (how pertinent to the ultimate aim of the discussion I don't know), some said hot-chips are the way they like them. Any how, I got as many *solid* encouragements as I got *solid* discouragements to go 20MHz. You can make a case both ways. If once again I encounter a game that demands 20MHz (yes, like Wing Commander 2), I will most likely switch it up to 20MHz again. I simply have to pop my casing, unplug the 35MHz oscilator, and pop in the 40MHz oscillator, whole thing takes 5 minutes. Right now I got 2 40MHz and one 35MHz oscillators and guess what... I am keeping all of them! E. Teng Ong (ong@d.cs.okstate.edu)
mlord@bwdls58.bnr.ca (Mark Lord) (02/06/91)
In article <1991Feb5.172330.27155@d.cs.okstate.edu> ong@d.cs.okstate.edu (ONG ENG TENG) writes: < <After upping my 386SX from 16MHz to 20MHz (without heatsink or fan), it <worked fine, and still will if I choose to run it at that speed. But <after many hours of continuous operation, my fixed-HD (I have a <fixed-HD and a SyQuest removable where I run off most of <my softwares, the fixed-HD contains most "standard" programs like <system programs, word processors, graphics editor, terminal emulator, <etc, only) my fixed-HD started giving disk-read error. To get rid of <it, I simply do warm-boot (NO turning off the power). ... guesses about cause deleted ... This problem is most likely due to the higher expansion bus speed due to the higher frequency clock oscillator (crystal). A 16-Mhz SX typically runs the bus at 1/2 cpu speed, giving a convenient 8Mhz bus to match the AT spec. When one speeds up the CPU by changing the crystal, as ONG has done, the expansion bus will also be run faster in most cases. Since it is run at 1/2 CPU speed, this means 10Mhz instead of the standard 8Mhz. Not all expansion cards can cope with faster bus rates. In particular, I once had a Western Digital disk controller that did not work reliably in my 10Mhz AT clone, which ran the bus at 10Mhz. In the non-turbo mode, everything worked fine. I suspect something similar is the culprit in this case. In my own souped-up system, everything is lucky enought to work fine on a 10Mhz bus (I have a souped up 16Mhz -> 20Mhz 386SX). -- ___Mark S. Lord__________________________________________ | ..uunet!bnrgate!mlord%bmerh724 | Climb Free Or Die (NH) | | MLORD@BNR.CA Ottawa, Ontario | Personal views only. | |________________________________|________________________|
timd@hotwheel.dell.com (Tim Deagan) (02/06/91)
In article <5540@bwdls58.UUCP>, mlord@bwdls58.bnr.ca (Mark Lord) writes: > In article <1991Feb5.172330.27155@d.cs.okstate.edu> ong@d.cs.okstate.edu (ONG ENG TENG) writes: > < > <After upping my 386SX from 16MHz to 20MHz (without heatsink or fan), it > <worked fine, and still will if I choose to run it at that speed. But > <after many hours of continuous operation, my fixed-HD (I have a > <fixed-HD and a SyQuest removable where I run off most of > <my softwares, the fixed-HD contains most "standard" programs like > <system programs, word processors, graphics editor, terminal emulator, > <etc, only) my fixed-HD started giving disk-read error. To get rid of > <it, I simply do warm-boot (NO turning off the power). > ... guesses about cause deleted ... > > > This problem is most likely due to the higher expansion bus speed > due to the higher frequency clock oscillator (crystal). A 16-Mhz SX > typically runs the bus at 1/2 cpu speed, giving a convenient 8Mhz bus > to match the AT spec. > > When one speeds up the CPU by changing the crystal, as ONG has done, > the expansion bus will also be run faster in most cases. Since it is > run at 1/2 CPU speed, this means 10Mhz instead of the standard 8Mhz. > > Not all expansion cards can cope with faster bus rates. Two points to consider: 1) The difference between a 16MHz CPU and a 20 MHz CPU is generally that the 16MHz chip failed when they tried it at 20 MHz. The process is called sampling. When INTEL does it they brand the results on the outside of the chip, when others do it you call it roulette. :-) The CPU is not the only chip speed sensitive, the entire chip set is speed rated (and generally not only sampled, but engineered for timing specs relevant to the rated speed). 2) Due to the plethora of fried cards due to the increased bus speeds of some 10 & 12 MHz 286's a few years back many card manufacturers have a clause in their warranty which declares the warranty void if the card is run in a system with a bus speed higher than 8.33 MHz. Check your warranty if you care. Also, the multi-speed function is usually just a divisor function off the main clock, rarely is it a different oscillator. ------------------------------------------------------------------------ ------------------"Wake up! Wake up!" - Spike Lee --------------------- ------------------------------------------------------------------------ timd@hotwheel.dell.com timd@twaddle.dell.com Rev. Tim Deagan
schwalbe@pinocchio.Encore.COM (Jim Schwalbe) (02/07/91)
In article <1991Feb5.172330.27155@d.cs.okstate.edu> ong@d.cs.okstate.edu (ONG ENG TENG) writes: > >After upping my 386SX from 16MHz to 20MHz (without heatsink or fan), it >worked fine, and still will if I choose to run it at that speed. But >after many hours of continuous operation, my fixed-HD (I have a >fixed-HD and a SyQuest removable where I run off most of >my softwares, the fixed-HD contains most "standard" programs like >system programs, word processors, graphics editor, terminal emulator, >etc, only) my fixed-HD started giving disk-read error. To get rid of >it, I simply do warm-boot (NO turning off the power). > Ong, Is your fixed-HD controller on the motherboard? (i.e. IDE) If not, then maybe it is your bus speed which is causing the problem. As I recall you have the AMI BIOS and the C&T chip set. With the setup, you can configure the bus speed to be syncronously scaled with the processor speed (i.e. you'd be running the bus at 10 MHz with the processor running at 20 MHz) or you can make the bus speed be fixed at 8 MHz (asynchronous, independent of processor speed). This is what you should try, keeping the bus speed at 8 MHz when you increase your processor speed to 20 MHz if you HD controller is on the bus and not the motherboard. It has been noted before that some peripheral cards don't work reliably at 10 MHz. I had this problem with my old ST-506 controller. Didn't like to work at 10 MHz. I upgraded to a dual IDE hard drive/ 4 Floppy controller (save a slot too!) and the problem went away. (Of course I got a bigger hard drive too). I now run my bus at 10 MHz with no problems. Hope this might help. .---------------------------------------------------------------------------. : Jim Schwalbe .----------------. "Half of what I say is : : Hardware Research Group .--+-------------. | meaningless; but I say it : : Encore Computer Corp. | | E N C O R E | | so that the other half may : : Mail: | `-------------+--' reach you." : : schwalbe@encore.com `----------------' - Kahil Gibran : `---------------------------------------------------------------------------'
phil@brahms.amd.com (Phil Ngai) (02/07/91)
In article <1991Feb6.055458.7488@d.cs.okstate.edu> ong@d.cs.okstate.edu (ONG ENG TENG) writes: |> When one speeds up the CPU by changing the crystal, as ONG has done, |> the expansion bus will also be run faster in most cases. Since it is |> run at 1/2 CPU speed, this means 10Mhz instead of the standard 8Mhz. | |Well, you might be right, but there is a good (and obvious) reason |to contradict. Since my motherboard allows me to switch down |to 8MHz operation regardless of the "main" oscillators I have |(32MHz, 35MHz, or 40MHz for 16MHz, 17.5MHz, or 20MHz operations). |It would be more reasonable for them to use the 8MHz crystal for It might be, but what counts is how it actually works, which depends on chip set you have and how it's configured. |Nonetheless, I have a funny feeling that you are right somehow! (i.e. |computer is funny that way). Computers are "funny" only to people who haven't a clue about how they work inside. -- Would you trust a government that didn't trust you?
phil@brahms.amd.com (Phil Ngai) (02/07/91)
In article <14901@uudell.dell.com> timd@hotwheel.dell.com (Tim Deagan) writes: |Two points to consider: |1) The difference between a 16MHz CPU and a 20 MHz CPU is generally that |the 16MHz chip failed when they tried it at 20 MHz. The process is called Tim, I've already tried explaining this kind of stuff to him and he simply isn't interested in an engineering approach to the subject. He'd rather treat computers as creatures with a will of their own. To him, computers are "funny" and the laws of cause and effect, and the principles of engineering, are not applicable. I don't know if he only behaves this way with computers or if he lives his whole life like this. |2) Due to the plethora of fried cards due to the increased bus speeds of |some 10 & 12 MHz 286's a few years back many card manufacturers have a clause |in their warranty which declares the warranty void if the card is run in a |system with a bus speed higher than 8.33 MHz. Check your warranty if you care. Trust me, he doesn't. -- Would you trust a government that didn't trust you?
rcollins@altos86.Altos.COM (Robert Collins) (02/08/91)
In article <1991Feb6.055458.7488@d.cs.okstate.edu> ong@d.cs.okstate.edu (ONG ENG TENG) writes: >From article <5540@bwdls58.UUCP>, by mlord@bwdls58.bnr.ca (Mark Lord): >> >> When one speeds up the CPU by changing the crystal, as ONG has done, >> the expansion bus will also be run faster in most cases. Since it is >> run at 1/2 CPU speed, this means 10Mhz instead of the standard 8Mhz. > >Well, you might be right, but there is a good (and obvious) reason >to contradict. Since my motherboard allows me to switch down >to 8MHz operation regardless of the "main" oscillators I have >(32MHz, 35MHz, or 40MHz for 16MHz, 17.5MHz, or 20MHz operations). >It would be more reasonable for them to use the 8MHz crystal for >i/o bus clock instead. Remember, I "change" oscillator by putting one >in the optional slot, the original 32MHz (for 16MHz operation) oscillator >remains on the motherboard (it is soldered in). > Didn't anybody read my posting on this subject? I've seen ONG quote many people in his subsequent postings, but he missed mine! Who cares? How many of you have ported a BIOS over to the NEAT or LEAP chipset (ONG has the NEAT chipset)? How many of you have read the NEAT spec? The one thing lacking in this whole conversation is a discussion of the hardware design of the motherboard. The NEAT and LEAP chipsets have the ability to run the BUS CLOCK (BCLK) synchronously or asynchronously to the CPU CLOCK. If it is run in async mode, then you better guarantee that your hardware was design with a crystal dedicated to running the peripheral BUS. If the hardware wasn't designed to run the bus asyncronously, then BCLK is a function of CLK. CLK / 2, CLK / 3, CLK / 4, CLK / 6 are some, but probably not all the divisors that can drive BCLK. If ONG's computer was designed to drive the bus syncronously, then increasing CLK will increase BCLK, and as many of you have mentioned, is probably the cause of the HD failure. However, regardless of a secondary crystal socket, and maybe a jumper to enable the socket, this does not imply, nor guarantee that BCLK can run asyncronously. The BIOS must program NEAT for ASYNC mode for it to work, AND there must be a secondary crystal specifically for this purpose. So, now let's discuss the hardware...ONG? How was it designed? -- "Worship the Lord your God, and serve him only." Mat. 4:10 Robert Collins UUCP: ...!sun!altos86!rcollins HOME: (408) 225-8002 WORK: (408) 432-6200 x4356
ong@d.cs.okstate.edu (ONG ENG TENG) (02/09/91)
From article <4710@altos86.Altos.COM>, by rcollins@altos86.Altos.COM (Robert Collins): > Didn't anybody read my posting on this subject? I've seen ONG quote many > people in his subsequent postings, but he missed mine! Who cares? How > many of you have ported a BIOS over to the NEAT or LEAP chipset (ONG has > the NEAT chipset)? How many of you have read the NEAT spec? The one thing > lacking in this whole conversation is a discussion of the hardware design > of the motherboard. > > The NEAT and LEAP chipsets have the ability to run the BUS CLOCK (BCLK) > synchronously or asynchronously to the CPU CLOCK. If it is run in async > mode, then you better guarantee that your hardware was design with a > crystal dedicated to running the peripheral BUS. If the hardware wasn't > designed to run the bus asyncronously, then BCLK is a function of CLK. > CLK / 2, CLK / 3, CLK / 4, CLK / 6 are some, but probably not all the > divisors that can drive BCLK. If ONG's computer was designed to > drive the bus syncronously, then increasing CLK will increase BCLK, and > as many of you have mentioned, is probably the cause of the HD failure. > However, regardless of a secondary crystal socket, and maybe a jumper > to enable the socket, this does not imply, nor guarantee that BCLK can > run asyncronously. The BIOS must program NEAT for ASYNC mode for it > to work, AND there must be a secondary crystal specifically for this > purpose. > > So, now let's discuss the hardware...ONG? How was it designed? Woo... touchy, touchy. I probably missed your post since our machine goes down twice a week, and I sometimes find followups that has no original post. My CPU is running on CLK/2, while the bus is running on ATBUS. I believe they are asyncronous. I think the CLK is the main crystal at 32MHz and 35MHz after I changed the position of the shorting block on the MB. I looked around but could not find the spec for ATBUS. Does anyone have any idea?