|
"Turbo" StrongArm - is overclocking acceptable? |
|
(20:19 2/7/2000) The Doctor (23:19 2/7/2000) The Doctor (23:23 2/7/2000) Tony (15:44 3/7/2000) alib (12:46 4/7/2000) Matrix (18:02 4/7/2000) ToiletDuck (21:05 22/7/2000) ams (20:09 24/7/2000)
|
|
Spyder |
Message #1516, posted at 20:19, 2/7/2000 |
Unregistered user
|
Hi all Having seen the news statement that APDL are overclocking SA processors and calling it a "Turbo" upgrade, I was wondering what the acceptability of overclocked SA processors were? I remember back in the good old days of Destiny, Robert Templeman used to boast about his 298Mhz SA. Now it seems that APDL are claiming that this can match a Kinetic board, so I was wondering what the issues were, and where people stood if the overclock failed. Also, would an overclocked Kinetic board run faster than this, and are APDL prepared to do the same for Kinetic owners? I know that I'd be mightily annoyed if I found that you could get the same performance increase for just forty quid, plus with Viewfinder, could this mean that the Kinetic board is undercut and outdone? |
|
[ Log in to reply ] |
|
The Doctor |
Message #1517, posted at 23:19, 2/7/2000, in reply to message #1516 |
Unregistered user
|
I have to say, if all they are doing is overclocking the CPU then this will be of limited use in most cases. Example: When overclocking my 233 to 280mhz, there was virtually no difference in performance running the Iron dignity demo. I know this isn't a definitive test, (it was one of a few) but my point is that the 16mhz Memory bus is the main bottleneck. The SA spends a lot of it's time twiddling it's thumbs waiting for numbers to crunch.Like you said though, Are APDL prepared to do this to a Kinetic card? I imagine THAT would make a difference!
|
|
[ Log in to reply ] |
|
The Doctor |
Message #1518, posted at 23:23, 2/7/2000, in reply to message #1517 |
Unregistered user
|
As far as the Warranty goes, I detected only a slight increase in CPU temp when overclocked flat out, so surely an agreement with Arm PLC can be reached if a fan is supplied to keep the CPU super cool. |
|
[ Log in to reply ] |
|
Tony |
Message #1519, posted at 15:44, 3/7/2000, in reply to message #1518 |
Unregistered user
|
Has anyone seen the turbo mod? Does it have a heat sink attached to the SA surface? You could probably still fit a tiny I2C thermometer to it and write a module that checks the temp every 10 seconds. |
|
[ Log in to reply ] |
|
alib |
Message #1520, posted at 12:46, 4/7/2000, in reply to message #1519 |
Unregistered user
|
Firstly, take a look at the show photos page (now linked to from the news item). This shows the upgrade in all its glory with epoxy resin glue, heat sync and fan arrangement. It's basically a socket 7 jobbie bolted onto the top of an SA chip. With regard to warranty, APDL take on all liability. Lastly, you can't overclock the Kinetic because there's no allowance for it on the daughterboard. It would also mess things up severely with timings, I'm told. Cheers, a |
|
[ Log in to reply ] |
|
Matrix |
Message #1521, posted at 18:02, 4/7/2000, in reply to message #1517 |
Unregistered user
|
In PC world, the new boards at 100 Mhz are faster than the old to 33 but i saw that the difference is not more... the overclock of a procesor is ever (ABOUT A PROGRAM EXECUTION) better than a new fast board but the DMA process NOT, so if kinetic would support a good DMA we will see a faster file transfer rate but the problem is that kinetic are pluged on a old board so his benefit about DMA (if kenetic a day will support) we never see, kenetic improve only the program and data translation OUT of the processor (this is important of course) but not like a fast file tranfer or a new bigger chace l1 or maybe l2... i want remember to all peoples that if you use cache off you strong ARM will be very slowly than and old ARM710 etc... also if the clock is more high. so i would like to see new ARMs and i hope that intel will put in it an FP and also an SMP system so we can support more than one ARM on a board... i want reember that now ONLY PENTIUM XEON can work more than 2 processor in parallel thank to the SMP tecnology that intel put in it... a new board redraw at 66 Mhz will be very good not need a 100 Mhz board.... |
|
[ Log in to reply ] |
|
Mark Quint |
Message #1522, posted by ToiletDuck at 21:05, 22/7/2000, in reply to message #1521 |
Quack Quack
Posts: 1016
|
you could try using !ArcQuake as a test for seeing how oc'ing the strongarm affects its speed as the game relies heavily of the processor as there is no fpu. if you could find the console commad, then you could compare the fps. |
|
[ Log in to reply ] |
|
ams |
Message #1523, posted at 20:09, 24/7/2000, in reply to message #1521 |
Unregistered user
|
As far as I know the ARM does support multiprocessing. It has a model which supports 14 coprocessors (ARMs or other device types). Processor 15 is on the ARM itself and controls the cache and memory management. The ARM10 supports a vector FP processor that (on paper) looks to be very fast and some ARMs are already available (ARM9) that support DSP functions in hardware and can easily handle things like MP3 at even lower clock rates than the ARM SA-110. The ARM 10 by the way (again on PAPER) is faster per megahertz than Intels proposed SA-2. One thing Paulo remember we had multiprocessing on Acorns as far back as 1983 (the BBC Model B with a 6502 or NS16032 or Intel80186 processor) and a then radical I/O bus called the TUBE. The first ARMs (ARM-1's) were debugged and software was developed on BBC model B's with an ARM as a coprocessor (in 1985-87) and there was an ARM development board for the PC (Springboard) from Acorn available in 1987. The expertise is there to do a really good MP system based on ARMs. ARMs have atomic swap operations built in (for Semiphores for example) a necessary part of multiprocessing. So why not a multiprocessing ARM system, the only limitation is the addition of MP support in RISC OS and more bus bandwidth (the latter will come with Imago). |
|
[ Log in to reply ] |
|
|