| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Community
Forum: Teams
Thread:  Join "MyOnlineTeam" Today - Chapter 55 |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 445
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
RT, as I understand it (correct me if I'm wrong) the i5s and i7s may only be dual core or quad core, but the design allows for more than the number of cores to run simultaneously. Like a quad core may allow seven or more tasks (WUs) to run at the same time since a core is not used 100% of the time for any one WU. It's like MVS running on big iron with the operating system restarting a WU when an interrupt occurs.
|
|||
|
||||
|
RT
Master Cruncher USA - Texas - DFW Joined: Dec 22, 2004 Post Count: 2636 Status: Offline Project Badges: 
|
RT, as I understand it (correct me if I'm wrong) the i5s and i7s may only be dual core or quad core, but the design allows for more than the number of cores to run simultaneously. Like a quad core may allow seven or more tasks (WUs) to run at the same time since a core is not used 100% of the time for any one WU. It's like MVS running on big iron with the operating system restarting a WU when an interrupt occurs. All the I7s are quad cores with hyper-threading which is a really neat technology that allows one hyper thread for each core. So all my I7s run 8 WUs at once; and do so very effectively. That is they will run 8 WUs in about the same time as they will run 4 WUs. Now I do not know that much about the I5s. Both that I run are in laptops and I keep them down to 2 WUs each for thermal purposes. If I tell them to run at 100%, they start up 4 tasks meaning they are quads but apparently do not have HT technology. I think that all I5s will only run 4 WUs because they do not have HT. I could be wrong on that. Now I was a MFT, MVT and subsequently a MVS systems programmer way back when and did not mess with it after about 1977 but... when I was working on it, any job check pointing had to be done in the application. (We had some Geophysical work that did its own checkpoints) but the majority of the applications, generally written in COBOL, would just restart from the beginning when we had system failures (which were of course rare when I was the system programmer ) . Today with the BOINC applications, each WU does its own check pointing and restarts from the last checkpoint when it is restarted without having completed its run. I am sure that someone will correct me if I said anything wrong here but the above is the way I understand things now.edit> Oh, Now I see what you meant. Most of the Mainframes that I worked with were uni-processors and a few Multiprocessors. And yes, the tasks were kept up with by the Task Control Blocks and when tasks did I/Os, control was passed to the next task in priority order in the TCB. That was Multiprogramming whereas these modern processors (I7s for this example) are Multi-Processors allowing at least 4 instructions to be processing at precisely the same time..whereas with the old mainframes even with the TLAB and TLasideB there was but one instruction executing at one time. Sorry I misunderstood. ---------------------------------------- [Edit 1 times, last edit by RT at Feb 5, 2012 1:58:19 AM] |
|||
|
||||
|
darth_vader
Veteran Cruncher A galaxy far, far away... Joined: Jul 13, 2005 Post Count: 514 Status: Offline Project Badges: 
|
RT, as I understand it (correct me if I'm wrong) the i5s and i7s may only be dual core or quad core, but the design allows for more than the number of cores to run simultaneously. Like a quad core may allow seven or more tasks (WUs) to run at the same time since a core is not used 100% of the time for any one WU. It's like MVS running on big iron with the operating system restarting a WU when an interrupt occurs. All the I7s are quad cores with hyper-threading which is a really neat technology that allows one hyper thread for each core. So all my I7s run 8 WUs at once; and do so very effectively. That is they will run 8 WUs in about the same time as they will run 4 WUs. Now I do not know that much about the I5s. Both that I run are in laptops and I keep them down to 2 WUs each for thermal purposes. If I tell them to run at 100%, they start up 4 tasks meaning they are quads but apparently do not have HT technology. I think that all I5s will only run 4 WUs because they do not have HT. I could be wrong on that. Now I was a MFT, MVT and subsequently a MVS systems programmer way back when and did not mess with it after about 1977 but... when I was working on it, any job check pointing had to be done in the application. (We had some Geophysical work that did its own checkpoints) but the majority of the applications, generally written in COBOL, would just restart from the beginning when we had system failures (which were of course rare when I was the system programmer ) . Today with the BOINC applications, each WU does its own check pointing and restarts from the last checkpoint when it is restarted without having completed its run. I am sure that someone will correct me if I said anything wrong here but the above is the way I understand things now.edit> Oh, Now I see what you meant. Most of the Mainframes that I worked with were uni-processors and a few Multiprocessors. And yes, the tasks were kept up with by the Task Control Blocks and when tasks did I/Os, control was passed to the next task in priority order in the TCB. That was Multiprogramming whereas these modern processors (I7s for this example) are Multi-Processors allowing at least 4 instructions to be processing at precisely the same time..whereas with the old mainframes even with the TLAB and TLasideB there was but one instruction executing at one time. Sorry I misunderstood. As I've mentioned in some of the team's earlier chapters, hyper-threading is a way to deal with the fact that access to DRAM is much slower than the processor. Rather than stalling the processor waiting for memory to respond, some amount of logic is duplicated and another thread can be dispatched. All of this is transparent to the OS -- it looks like there are more cores than there actually are. You are correct that currently all I5 processors have 4 real cores and do not support hyper-threading. I would not be surprised if that changes at some point since Intel has not been able to stay with one naming convention for very long, at least lately. Now as for mainframes ... there were, of course, multi-processing systems in the 1960s and maybe even in the 1950s. IBM only had 2-way systems, back then if you don't count the processors that were in the I/O channels, but there were other systems that had far more. There was a supercomputer (yes, really just a large mainframe) at Ames Research Center that had 100 processors in the 1970s. Darned if I can remember the name of the system though. IBM had systems with 4 cores in the early 1980s. Current IBM mainframes have up to 80 cores (again, not counting the I/O processors, etc)... but it's more complicated than that as sometimes an individual core may be executing multiple instructions for a single thread at once, i.e., it's superscalar. To bring this back to WCG and systems you might have at home, some of the newer AMD processors that support their version of hyper-threading have only a single floating point processor for two "cores". That will hurt performance for many WCG workunits. AMD is continuing to move into the value space and seems to have ceded the performance crown to Intel as far as x86 processors are concerned. Edit: I finally remembered the computer at Ames Research Center was called the ILLIAC IV. It was the world's fastest computer for a few years. My high school math class went on a tour and we got to see it. Apparently parts of it are now at the computer history museum only a few miles away from Ames. - D [Edit 1 times, last edit by darth_vader at Feb 5, 2012 6:18:57 AM] |
|||
|
||||
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges: 
|
MyOnlineTeam Daily Statistics for 02/04 - All Members:
----------------------------------------Team rank movement report =========================
Points milestones report ======================== YellowAV reached 2,000,000 points  Runtime milestones report ========================= No runtime milestones found.  Results returned milestones report ================================== No results returned milestones found. New members report ================== No new members found. Retired members report ====================== No new retired members found. For the week as a team: Statistics Total Run Time Points Results Team Records: Results Returned: 12/15/2011 4,598 Points: 12/15/2011 1,578,741 Runtime: 12/15/2011 1:125:23:14:23 Good crunching folks!!!!! |
|||
|
||||
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges:
|
MyOnlineTeam Daily Statistics for 02/04 - Active Members
----------------------------------------Active team members report ==========================
Note: Active members are those who earned points in the prior 30 days. Top Twenty active members returning points today: 01: RT - 242,449 points 02: judson Somerville MD - 159,235 points 03: GeraldRube - 140,720 points 04: Coingames - 78,798 points 05: xroule - 55,332 points 06: David Autumns - 32,478 points 07: NiceMedTexMD - 29,977 points 08: dkt - 25,934 points 09: Daeloan - 25,710 points 10: brown chris - 24,894 points 11: parmesian - 21,658 points 12: finman - 18,237 points 13: darth_vader - 17,298 points 14: Vuj - 17,103 points 15: pramodp - 14,033 points 16: Jonathon Wright - 13,439 points 17: keithhenry - 13,354 points 18: lawrencehardin - 13,251 points 19: smcclarigan - 12,786 points 20: dubhain - 9,168 points Total points returned today: 1,014,921 Active members returning points today: 40 Average points per member active today: 25,373.025 |
|||
|
||||
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges:
|
.............................CONGRATULATIONS YellowAV ON REACHING 2,000,000 MOT POINTS !!!............................. |
|||
|
||||
|
RT
Master Cruncher USA - Texas - DFW Joined: Dec 22, 2004 Post Count: 2636 Status: Offline Project Badges:
|
 Good Morning MOT Congratulations YellowAV! Nice Round Number.I think the first Multiprogramming Operating System on IBM mainframes was when they first came out with MFT (Replacing PCP). I am not sure of that year but I am pretty sure it was in late 60s. My first exposure to it was in 1969 on a 360/50 IIRC. The first IBM multi-processor I ever saw as a 370/168 AP (Attached Processor) and shortly after that the MP and this was in the mid-late 70s (the AP and MP). Remember the aluminum bar that was mounted across the top of the front face of the 168s with "IBM 360 168 painted on it (well everything but the letters are painted black) Picture . I have one of those up in the cabinet above where I sit at my main computer as I type this. Those were fairly short lived and replaced by the 3033 in 1977-78. In any case the multi-programming was a function of the operating system and the multi-processing was a function of the hardware - along with operating system support..that is having more than one instruction and execution units. I pretty much lost track of these technologies along in the early 80s as my job took me into a completely different "world". Now where I worked was not at IBM but we always had many IBM mainframes (as well as UNIVACs with custom built array processors) and so were not too far behind the bleeding edge of both companies. We had a few AMDAHL systems thrown in to keep everyone on their toes. Most of our commercial work was on IBM and Amdahl computers and most of the scientific work was on UNIVAC/SPERRY/RAND so it is entirely possible that IBM scientific processors/systems had features that I was either unaware of or have forgotten by now. In many ways the array processors we had built for us were similar to the GPU cards today. Of course the compute capacity is vastly different. We had massive copper grids under the false flooring to ground all the stuff to and huge rotating drums along with what IBM called LCS (Large Core Storage) boxes holding a million bytes of core memory (little metal doughnuts -- something like 9 million of them strung on copper wires - amazing stuff). It was an astounding site to see with each mainframe having up to 32 large tape drives and banks of 2314s and later 3330s. Armies of people moving tapes and occasionally disks between drives and storage. All of that and it is likely that I personally own more computational power now than we had as a corporation back then. Well enough of this trip down memory lane. That was then and this is now. If you are going to buy a CPU processor for WCG tasks, you are much better off with an I7 than you are with an I5 because you can get something approaching twice the throughput. As darth points out, you are likely better off with Intel. I was not aware of the single floating point processor for two cores he pointed out but it makes sense that this fact alone would make a dramatic difference in WCG throughput. I just looked it up and found this and this about the old IBM systems. I suspect it is boring to all but those of us that made a living working with them. Everyone have a great day this Sunday. I don't much care about the Super bowl. NY vs. Boston makes a big difference to some, but not to me. Now if it were the Cowboys playing the Texans, my tune would be different. One of your friends in Texas  |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You are pretty much right in your recollections and I don't want to go into much more detail, but let me give a very small description of the IBM Systems. These begin with the mod 30 which had only real iron core memory and PCP was the operating system in use.
As we go further in years we reach all of the other software operating systems including one called MVM which was used until the current MVS system was available. So a short description of each of these follows: PCP - the first IBM 360 OS provided for a single job single task. MFT - provided for multiple tasks but no subtasling. MVT - provided for a variable number of tasks with subtasking. MFT2 - an extension of MFT that allowed for subtasking. MVM - multiple virtual memory was, as I recall, the first virtual memory operating system. MVS - multiple virtual storage was, and is, the virtual storage operating systems where each address space is controlled by an address space control block (ASCB) with multiple task control blocks (TCBs) for controlling all of the subtasks within a job running under an ASCB. When I retired from IBM, more and more of the system functions were operating in their own address space. |
|||
|
||||
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges:
|
MyOnlineTeam Daily Statistics for 02/05 - All Members:
----------------------------------------Team rank movement report =========================
Points milestones report ======================== judson Somerville MD reached 81,000,000 points NiceMedTexMD reached 34,000,000 points Runtime milestones report ========================= smcclarigan reached 7 years of runtime Results returned milestones report ================================== David Autumns reached 45,000 results YellowAV reached 4,000 results New members report ================== No new members found. Retired members report ====================== No new retired members found. For the week as a team: Statistics Total Run Time Points Results Team Records: Results Returned: 12/15/2011 4,598 Points: 12/15/2011 1,578,741 Runtime: 12/15/2011 1:125:23:14:23 Good crunching folks!!!!! |
|||
|
||||
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges:
|
MyOnlineTeam Daily Statistics for 02/05 - Active Members
----------------------------------------Active team members report ==========================
Note: Active members are those who earned points in the prior 30 days. Top Twenty active members returning points today: 01: RT - 280,042 points 02: judson Somerville MD - 173,106 points 03: GeraldRube - 126,262 points 04: Coingames - 74,948 points 05: xroule - 47,289 points 06: Daeloan - 35,119 points 07: dkt - 35,098 points 08: David Autumns - 34,006 points 09: parmesian - 25,868 points 10: NiceMedTexMD - 20,106 points 11: brown chris - 20,028 points 12: finman - 17,771 points 13: smcclarigan - 17,763 points 14: darth_vader - 17,245 points 15: pramodp - 16,669 points 16: Vuj - 15,211 points 17: keithhenry - 11,932 points 18: Jonathon Wright - 11,507 points 19: lawrencehardin - 11,398 points 20: Tomwp - 10,633 points Total points returned today: 1,066,216 Active members returning points today: 38 Average points per member active today: 28,058.3158 |
|||
|
||||
|
|