The Support Does Not End With Dual 604es: |
August, 14, 2002 1:56 PM |
webmaster814 |
Ryan, The support does not only end in 10.1.5, I am making sure that the predecessors of X are covered as well, in addition to continuing work on pet peeve projects that I have dreamt about for some time. I am working on new patches that will enable full dual 604e support in the following OSs: 7.6.1, 8.6.1, 9.0.1/0, and 9.1.l (in addition to the already completed support of 10.1.5, and 9.2.2). My long term goal would be both full support for single and dual 604e in NextStep 3.3.1. When dual 604e Support is complete on all the listed OSs (minus NextStep 3.3.1) my support will be included on a single CD (Mac Resolution), and will be sold for the reasonable price of $49.99 or $59.99 (depending on demand). Once PCIQuartz’XTREM’, I will bundle it in addition to the dual 604e support on a single CD (nicknamed MacRevolution) and sell it for the reasonable price of $99.99 or $129.99 (depending on demand). Lastly, once support for 604e based Macs in NextStep 3.3.1 is complete I will bundle by itself onto a single CD (codenamed Orion). The road ahead is long and uncharted who will be the guide. Finally, all those interested in beta testing must pay a nominal $5.00 donation, and in return will be sent either the 9.2.2, or 10.1.5 support CD. Michael P.S. I am open to donations which will only help in the development on Mac based 604e support. Please note, that my support will be fully packaged in professional level boxes, with professional level manuals and what not, in addition to entitling the buyer to one month of technical support over the telephone. And, no pre-orders are not yet available, but should be in the upcoming week(s). |
. |
RE: The Support Does Not End With Dual 604es: |
August, 22, 2002 5:15 PM |
webmaster814 |
. |
Thomas, I have been out of town in Palo Alto for a while now. Before I left I had to switch my mail over from a Daystar Genesis to my NEC OpenStep Laptop in order for me to get my email on the road. My guess is your mail was lost in the transaction. Please resend whatever email you sent me. Michael |
. |
RE: The Support Does Not End With Dual 604es: |
August, 22, 2002 3:40 AM |
twalter |
. |
Most of the concepts webmaster814 describes I haven't heard of anything before and many explanations are little bit vague, at least for me. So I am sceptical but eager to try it out. I tried to contact him (webmaster@minux.zzn.com) but I didn't get any answer. Did anyone get an answer? Has anyone received the software and can confirm that it is functions and isn't vaporware? Thomas |
. |
RE: The Support Does Not End With Dual 604es: |
August, 16, 2002 5:46 PM |
mjoecups |
. |
SOunds like a trolling fool. I think. Notice the XTREM reference? Marty PS Hope I'm wrong |
. |
RE: The Support Does Not End With Dual 604es: |
August, 16, 2002 10:56 AM |
webmaster814 |
. |
Will your software work in future releases and additions of the Darwin Kernel? Well, behind every major release, i.e. 10.1.x, 10.2.x there are a lot of different variances, meaning the ‘x’ in 10.1.x, and 10.2.x. My software will work as long as the kernel for the previous release of the software is based off the same ‘root’ version of Darwin, ie 10.1.x, and 10.2.x. If the Darwin file structure changes dramatically, in say 10.3.x, then I will most likely have to update my software to support it. However, as of now, it runs in 10.2.0 build C649b, and will continue to work in the next versions of 10.2.x. I’ll am sure I will have to update it a tad when the commercial release of 10.2.0 is out. So to answer your question, yes and no, testing will be done before I release any versions of my software and it will state explicitly what OSs are supported and which ones are not. Will the software have built in capabilities to directly install into X? Yes it has a MacBasic script that I have written which will execute the necessary steps, and install the necessary software which enables the computer to be detected during boot as a G4. “if the date…� That was a typo on my part, it is supposed to read “if the data…� Top quality ram My premise behind that statement is to stress the importance behind the way my software is utilized. My software puts A LOT of stress on the media and the ram of a computer. The RAM is used a lot more under my method. What do you mean? In current computers when the computer detects an error in the data it goes back and searches through the data until that error is found. But of course, there are a lot of problems with this, when forcing data into bytes it sometimes causes more errors. Now, in my method when the error is found it completely rewrites the complete block of data, thus putting more stress on RAM, which is why I recommend only top quality RAM. AKA RAM from OWC, Viking, Kensington, ect . Just do not use the good ‘ol ram packaged with you computer from 1994+. Any ways, yes, my software puts more emphasis on RAM then in current methodologies of ram usage that computers use today. How do CPUs operate? Currently, Apple’s strategy on CPU usage is to max out the first CPU to 100% then any overflow goes to the second processor. Now obviously this idles the main CPU, stalls the second one until the second one can find both the last operation that the main CPU was processing, and on top of that floods the data path/ system bus. There are many advantages of this methodology in faster clocked CPUs, and CPUs which have larges memory controller. However, in 604es this does not work, mainly because of the lack of L1/L2 cache, forcing the second CPU to go to RAM, which in returns slows down the computer’s operation. That is frankly why, I decided to do this. I decided that I thought the best way to solve this problem was by switching cycles. In my application the different chunks of data alternate between CPUs, and not forcing the CPU to go to RAM as often, I have only come across a few times in which both CPUs stalled. Like I said before, this greatly increases the speed of the 604es. |
. |
RE: The Support Does Not End With Dual 604es: |
August, 16, 2002 12:41 AM |
developer |
. |
>> Instead of the main CPU running at 100% my software enables the CPUs to toggle between clock cycles, sending every clock cycle to a different CPU, instead of the main CPU consuming 100% of the bandwidth. In the end this cools both of the CPUs down considerably << What does this mean exactly? That each instruction of any running thread is scheduled to a different CPU? That in effect only one CPU operates at any given time? Where will the speed increase will come from - to my understanding the registers have to be transfered too etc., and this would be the simplest problem compared to things like instruction pipelines within the CPUs itself. And doesn't MP mean, that all CPUs operate - in the ideal case at 100% - at once? You see - questions over questions... |
. |
RE: The Support Does Not End With Dual 604es: |
August, 16, 2002 12:14 AM |
developer |
. |
>> How long does it take to optimize a system under my software package? Depending on your type of hardware, and options slected, it may take anywhere from 45mins to 3hours. << Hmm? Does this mean that you go though every file within the darwin kernel (or even Core/Foundation/AppKit Frameworks) and optimize them for your MP support and replace them? your method sounds to me that you replace a set of instructions with an optimized version, maybe adding some support for the prevention of idling... along with the possible combinations of instructions this could be an explanation for the 50 MB you stated. But if this is the case, shouldn't this be the job of the compiler (gcc)? |
. |
RE: The Support Does Not End With Dual 604es: |
August, 15, 2002 11:48 PM |
developer |
. |
>> Well, in my software package about 5% of the processing used in the 604es is dedicated directly to rewriting the error check schemes. Meaning if date is missing all the data with errors in it will be rewritten instead of the CPU grabbing the error data and finding the portion of the data with errors on it. << Sorry but this is confusing me. What do you mean with "if date is missing"? Does this mean that some data in the L1 cache is outdated because of changes in the other CPU? And "all the data with errors" in conjunction with >> However, only top quality RAM is recommended with this option selected in my public release of the software package. << ?? So do you mean that the CPUs get wrong data from the RAM? No that couldn't be the case. Maybe you mean that your kernel will utilize the memory bus to a higher degree and such will stress the RAM more then the normal schemes? I know that some RAM errors on machines like the 8100 only show up in Photoshop using hefty filters like Gaussian blur etc. ... But anyway: shouldn't the synchronisation of data within the caches and concurrent access to RAM coordinated by the CPUs and the memory controller themselves? Just thought of MESA (maybe this is the wrong acronym, but I mean the four states for MP capable CPUs..) will do exactly this. Thanks! PS: sell it to Apple! |
. |
RE: The Support Does Not End With Dual 604es: |
August, 15, 2002 11:30 PM |
developer |
. |
What is about Apples update mechanism? Once they update something like the kernel etc., what will happen? Will the machine boot into X again? While I think the proclaimed speed gains could be fine - I would like to see a Dual approach in the XPostFacto sense. I've seen that there is already a littlebit source out there, but I can't understand what the actual problems are to get MP running... |
. |
Let's wait for the code |
August, 15, 2002 9:57 PM |
OSXGuru |
. |
I don't know if it's for real or not. It sounds as though webmaster814 is approaching the MP support problem in a creative way--I don't understand how what he describes would work, but that doesn't mean that it won't. As for whether it is appropriate for the forum, I think it has been appropriate so far. People ask me about 604 MP support all the time, so there is interest. But let's not argue about whether webmaster814 has actually accomplished something real or not. Let's just wait and see. |
. |
RE: The Support Does Not End With Dual 604es: |
August, 15, 2002 7:40 PM |
willschou1 |
. |
Is this for real or just so much BS making use of OWC's forum which could better be used for something else. Ryan should this thread be deleted?????????? Just a question not meaning to start a flame war ;-) I must admit you've got nerve trying to sell a product which may or may not be real here at someone elses forum! |
. |
RE: The Support Does Not End With Dual 604es: |
August, 15, 2002 1:33 PM |
webmaster814 |
. |
OS X not fully optimized? Yes, that includes X as well, including Apple's SMP support for the G4. Donate to Darwin? Pricing of my software? I understand your thinking behind it. However, if I were too donate my time and work to Darwin, then there would be no reason why other people could not implement the same features themselves. When I sell the versions, I will have different pricing for education, government, business, and personal. The education versions will be the cheapest, some where around 25% of the government cost. Any ways, I think I need to be compensated for the work that I have done on this project. And at release none of my source will be available, minus the portion of L2CacheConfig that I used in part of Ryans. Now that does not mean that I will not donate my optimization software for the 604e, that completely optimizes the 604e processors. In fact, I think this is something everyone needs who uses 604e based processors. This code I have written fully optimizes all 32k cache or more 604es in the supported OSs based on my memory addressing structure. However, in single CPU environments my processing structure may not be as fast as Apple’s in single based implementations of the 604e. Support for other OSs, excluding X? I know many users who prefer 8.6 over any other OS. I know users who prefer 7.6.1 over any other OS. LIke wise, there are those who prefer 9.1 over 9.2 and/or over X, and vice versa. The bottom line is for those of us who have multiple OSs we want every OS we install to be fully optimized for our computers. Something that Apple did not seem to care about. Apple code implemented? The answer to this is absolutely no. I have not implemented any code written by Apple into this project. In fact, I would be sued by Apple if I did. However, I have used Apple and Motorola's models for SMP, memory addressing, and optimization of the 604e processor to design my project. How big will the software package be? The software package currently is 50mb in size per OS, more or less depending on the particular OS. Which I’m sure many of you brains can not fathom now, but wait and see. How did you overcome the process of 604e CPU idling? Well, in my software package about 5% of the processing used in the 604es is dedicated directly to rewriting the error check schemes. Meaning if date is missing all the data with errors in it will be rewritten instead of the CPU grabbing the error data and finding the portion of the data with errors on it. This of course saves time and CPU bandwidth and causes no idling in both the CPU and the RAM. However, only top quality RAM is recommended with this option selected in my public release of the software package. How long does it take to optimize a system under my software package? Depending on your type of hardware, and options slected, it may take anywhere from 45mins to 3hours. What graphics cards will be supported under PCIQuartzXTREM? Only ATI Rage 128 based GPUs with 32mb or more will work, i.e. ATI Nexus PCI. Work on the Radeon has not started by me yet, because I can not locate anywhere how the GPU writes to the onboard VRAM of the card, and I can not determine the schematics of such a GPU. Hope This Cleared Some Smoke, Michael |
. |
RE: The Support Does Not End With Dual 604es: |
August, 15, 2002 10:39 AM |
developer |
. |
Another question: you stated that SMP support is not finished in ANY OS Apple made. Does this include OS X? If so - I understand that selling your solution sounds somehow appealing. But: I think that it might be more helpfull for all and the platform itself, if you could put your code into opendarwin so it might be included into apples darwin. you might get even more money by supporters who think that a more advanced solution compared to Apples should be included into OS X itself. And if the pressure is high enough on Apple, who knows if they will employ you anyway? Michae |
. |
RE: The Support Does Not End With Dual 604es: |
August, 15, 2002 10:30 AM |
developer |
. |
First of all: I think that supporting pre OS X is just wasted time. I think that due to the nature of the cooperative multitasking there might be so many glitches hidden, that it simply makes no sense to debug all those. Maybe you've applied magic and it will work flawlessly... I also think that most of OS 8/9 users with dual/quad CPUs will jump on X if it will work there correctly. which leads to my question: I'm wondering if you are using darwins mach-scheduler or if you wrote your own. If you wrote your own - what does this mean to the rest of the OS? Will running programs act as they will on a supported Apple MP hardware? Will drivers (PCI bus?) work stable? Michael |
. |
RE: The Support Does Not End With Dual 604es: |
August, 14, 2002 2:53 PM |
webmaster814 |
. |
Sorry I omitted part of your first question on accident. Instead of the main CPU running at 100% my software enables the CPUs to toggle between clock cycles, sending every clock cycle to a different CPU, instead of the main CPU consuming 100% of the bandwidth. In the end this cools both of the CPUs down considerably, and in return allows the user to then over clock the CPU (20-25% of the clock speed) stably. Of course, this then increases the bandwidth of the computer considerably. In the case of the Daystar Genesis MP 932+ that I was allowed to use at my place of employment, the amount of bandwidth gained was unbelievable, my benchmarks were in the fascinate of a 733mhz G4 in OS 9.2.2. The over clocking feature that I am talking about can be done easily from my software, and is theoretically 100% compatible with any 604e with more then 32k Cache Hope This Helps, Michael |
. |
RE: The Support Does Not End With Dual 604es: |
August, 14, 2002 2:43 PM |
webmaster814 |
. |
Dual CPUs are not fully optimized in ANY OS that Apple has made. If you notice in the case of the 604es, usually one of them runs at 100% while the other one runs usually at anywhere from 0% to 10% (max), which is the main reason why dual 604e benchmarks are incredibly low, in retrospect to a bottom of the line G3. It is not like I am magically making this support. Apple never finished the work for SMP of 604es, the support that Apple wrote is only partially there. In generic terms I am simply riding the bear, and in return delivering the support that Apple promised to its’ customers in 1997+. To your second question, to install X I am using the old technique that was discovered sometime about a year and a half ago. My software writes data into many places (most notably the PRAM, the NVRAM, the system, and extensions folder) which in returns fouls OS X into thinking that the machine is a dual G4 at boot/OS X installer. However, once X gets through loading the kernel, and checking the ram, the extension of the application (which was compiled, optimized, and saved into the extensions folder) that I made then loads the extension, and fully optimizes itself for OS X, and SMP 604e support. Hope this cleared things up, Michael |
. |
RE: The Support Does Not End With Dual 604es: |
August, 14, 2002 2:09 PM |
bralston |
. |
I guess I'm missing something. Since when are dual 604e's not supported in 8.6-9.2.2? My dual processor 604e card has always worked and both processors recognized with 8.6-9.2.2. Granted, OS 8/9 aren't nearly as multithreaded as X, multiprocessing is supported. I suppose then that your software somehow enables "automatic" use of dual processors even for apps that aren't written for duals? Also, you claim 10.1.5 support, but you say it doesn't work with Xpostfacto? How are you even booting into X then??? -Ben |
|
|