tak@tcela.COM (Michael Takayama) (01/17/91)
Looking through the latest UNIX Review (Jan 91), I noted that the 4D/340VGX
is SPECmarked at 19.5 while the new 4D/35 is SPECmarked at 23 (pgs. 75 and
110, respectively).
OK, all you folks at SGI in ASD and ESD, given the price differential between
the 4D/35 and 4D/3xx workstations, what is the *REAL* difference in compute
performance (not graphics - it's pretty much a given that GTX and VGX
performance far exceed TG performance) between these systems (esp. 4D/35S
and 4D/310S)?
-------------------------------------------------------------------------------
Michael Takayama email: tak@tce.com
Technical Support Manager
TDI America
"Well, which DOES weigh more - a pound of feathers or a pound of rocks?"
- pre-law student in 1st year physics class
-------------------------------------------------------------------------------
Disclaimer:
All of the opinions herein stated are my own... blah, blah, blah...
------------------------------------------------------------------------------- jeremy@perf2.asd.sgi.com (Jeremy Higdon) (01/18/91)
In article <112@tcela.COM>, tak@tcela.COM (Michael Takayama) writes: > Looking through the latest UNIX Review (Jan 91), I noted that the 4D/340VGX > is SPECmarked at 19.5 while the new 4D/35 is SPECmarked at 23 (pgs. 75 and > 110, respectively). > > OK, all you folks at SGI in ASD and ESD, given the price differential between > the 4D/35 and 4D/3xx workstations, what is the *REAL* difference in compute > performance (not graphics - it's pretty much a given that GTX and VGX > performance far exceed TG performance) between these systems (esp. 4D/35S > and 4D/310S)? > The 4D/340 has 4 processors. The 19.5 mark is for a single processor only. I believe that the throughput number is 67.something, or 4 @ 16.something.
tj@merlin.asd.sgi.com (tj) (01/24/91)
In article <112@tcela.COM>, tak@tcela.COM (Michael Takayama) writes: > Looking through the latest UNIX Review (Jan 91), I noted that the 4D/340VGX > is SPECmarked at 19.5 while the new 4D/35 is SPECmarked at 23 (pgs. 75 and > 110, respectively). > > OK, all you folks at SGI in ASD and ESD, given the price differential between > the 4D/35 and 4D/3xx workstations, what is the *REAL* difference in compute > performance (not graphics - it's pretty much a given that GTX and VGX > performance far exceed TG performance) between these systems (esp. 4D/35S > and 4D/310S)? > > ------------------------------------------------------------------------ ------- > Michael Takayama email: tak@tce.com > Technical Support Manager > TDI America The performance differences between the 4D35 and 4D310 are correct. The price difference is not because of performance, rather because of configurability. The 310 is upgradeable through an eight processor system. It has more VME and disk capability as well as faster I/O with IPI drives and up to 256MB of memory instead of 128 on the 4D35. Tom Jermoluk ASD/SGI
tak@tcela.COM (Michael Takayama) (01/24/91)
jeremy@perf2.asd.sgi.com responds: >In article <112@tcela.COM>, tak@tcela.COM (Michael Takayama) writes: >> Looking through the latest UNIX Review (Jan 91), I noted that the 4D/340VGX >> is SPECmarked at 19.5 while the new 4D/35 is SPECmarked at 23 (pgs. 75 and >> 110, respectively). >> >> OK, all you folks at SGI in ASD and ESD, given the price differential between >> the 4D/35 and 4D/3xx workstations, what is the *REAL* difference in compute >> performance (not graphics - it's pretty much a given that GTX and VGX >> performance far exceed TG performance) between these systems (esp. 4D/35S >> and 4D/310S)? >> > >The 4D/340 has 4 processors. The 19.5 mark is for a single processor only. >I believe that the throughput number is 67.something, or 4 @ 16.something. No, no, no! This is not what I (and many others) am interested in! I guess that I have to be MORE SPECIFIC: In the heavyweight division, we have the new 4D/310S Power Center IRIS compute server from ASD in a big box with lots of slots for peripherals and heavy-duty power consumption. In the welterweight division, we have the new 4D/35S Personal IRIS compute server from ESD in a small box with few slots for peripherals and relatively low power consumption. These are both single-processor systems (apparently the SAME processor). If the 19.5 SPECmark is indeed for the single-processor system (i.e. 4D/310S), then it is significantly lower than the SPECmark of 23 for the 4D/35S. This is remarkable given that a 4D/35S configuration is cheaper than an equivalent 4D/310S configuration. The REAL question is: Is there any reason to purchase a 4D/310S vs. a 4D/35S from a compute performance standpoint? Please leave out arguments of graphics, expandibility, upgrades, etc. Incidentally, the price/performance ratio is just as bad for the 4D/320S dual-processor systems. With the 4D/340S (quad-processor), we might begin to reach parity; with the 4D/380S (oct-processor), we might come up to parity plus maybe a bit more. With quantity discounts, though, a farm of 4D/35S's could be a better solution than ANY multi-processor IRIS in compute server configurations. How do you justify 4D/3xxS compute servers vs. 4D/35S compute servers given the price/performance discrepancies? ------------------------------------------------------------------------------- Michael Takayama email: tak@tce.com Technical Support Manager TDI America "The 4D/35S or the 4D/310S? Well, which keeps better time - Timex or Rolex?" -------------------------------------------------------------------------------
scfisher@oasys.dt.navy.mil (Steven Fisher) (01/24/91)
In article <113@tcela.COM> tak@tcela.COM (Michael Takayama) writes: >The REAL question is: > >Is there any reason to purchase a 4D/310S vs. a 4D/35S from a compute >performance standpoint? > >Please leave out arguments of graphics, expandibility, upgrades, etc. > One good reason for getting a 310 over a 35 is the price of memory. You can get 3rd party memory for the 310 at $850.00/8MB (from Impediment (617) 837-8877), while the 35 memory is only available from SGI at $3K/8MB. If you need alot of memory, this can make a huge difference. Because of the way SGI did the memory on the 4D-35, there may never be 3rd party memory for it (so they won't be as pressured to lower the memory cost). If you want 64MB, the difference in cost is $15K ($10K for us government types). The 4D-35 has the advantage of a built in parallel port and is easier to add 3rd party disks and stuff since it has a built in SCSI port on the back (if it is like the 20 and 25). The 4D-35 should be quieter than the 310. The 310 requires 20amp 120 volt power (or a 20amp to 15amp cord converter like we did). Most of all, the 310 is bigger => more impressive => more serious! -steve
tj@merlin.asd.sgi.com (tj) (01/25/91)
In article <113@tcela.COM>, tak@tcela.COM (Michael Takayama) writes: > jeremy@perf2.asd.sgi.com responds: > > >In article <112@tcela.COM>, tak@tcela.COM (Michael Takayama) writes: > >> Looking through the latest UNIX Review (Jan 91), I noted that the 4D/340VGX > >> is SPECmarked at 19.5 while the new 4D/35 is SPECmarked at 23 (pgs. 75 and > >> 110, respectively). > >> > >> OK, all you folks at SGI in ASD and ESD, given the price differential between > >> the 4D/35 and 4D/3xx workstations, what is the *REAL* difference in compute > >> performance (not graphics - it's pretty much a given that GTX and VGX > >> performance far exceed TG performance) between these systems (esp. 4D/35S > >> and 4D/310S)? > >> > > > >The 4D/340 has 4 processors. The 19.5 mark is for a single processor only. > >I believe that the throughput number is 67.something, or 4 @ 16.something. > > No, no, no! This is not what I (and many others) am interested in! I guess > that I have to be MORE SPECIFIC: > > In the heavyweight division, we have the new 4D/310S Power Center IRIS compute > server from ASD in a big box with lots of slots for peripherals and heavy-duty > power consumption. > > In the welterweight division, we have the new 4D/35S Personal IRIS compute > server from ESD in a small box with few slots for peripherals and relatively > low power consumption. > > These are both single-processor systems (apparently the SAME processor). > If the 19.5 SPECmark is indeed for the single-processor system (i.e. 4D/310S), > then it is significantly lower than the SPECmark of 23 for the 4D/35S. This > is remarkable given that a 4D/35S configuration is cheaper than an equivalent > 4D/310S configuration. > > The REAL question is: > > Is there any reason to purchase a 4D/310S vs. a 4D/35S from a compute > performance standpoint? > > Please leave out arguments of graphics, expandibility, upgrades, etc. > > Incidentally, the price/performance ratio is just as bad for the 4D/320S > dual-processor systems. With the 4D/340S (quad-processor), we might begin > to reach parity; with the 4D/380S (oct-processor), we might come up to > parity plus maybe a bit more. With quantity discounts, though, a farm of > 4D/35S's could be a better solution than ANY multi-processor IRIS in compute > server configurations. How do you justify 4D/3xxS compute servers vs. 4D/35S > compute servers given the price/performance discrepancies? > > ------------------------------------------------------------------------------- > Michael Takayama email: tak@tce.com > Technical Support Manager > TDI America > > "The 4D/35S or the 4D/310S? Well, which keeps better time - Timex or Rolex?" > ------------------------------------------------------------------------------- The answer is simple for performance. If all you are looking at is single process performance with no concern for expandibility of VME, memory, etc then the 4D35S is a better buy than a 310S. Many of our customers do in fact create compute farms of many 4D35S. It is entirely dependent on your application as to wether 8 4D35S or 1 4D380 is a better buy. Remember that with 8 4D35S you have to buy 8 times as much memory (on the 380S it's shared). You also have the network overhead of crossmounting disks with NFS (once again on the 380S they are shared). The 380S is also capable of running jobs in parallel mode to speed up a single process, not just more throughput. Tom Jermoluk ASD/SGI
sgf@cs.brown.edu (Sam Fulcomer) (01/25/91)
In article <1991Jan24.164248.9291@odin.corp.sgi.com> tj@merlin.asd.sgi.com (tj) writes: >In article <113@tcela.COM>, tak@tcela.COM (Michael Takayama) writes: >> >> server configurations. How do you justify 4D/3xxS compute servers vs. >4D/35S >> compute servers given the price/performance discrepancies? > >The answer is simple for performance. If all you are looking at is single >process performance with no concern for expandibility of VME, memory, etc >then the 4D35S is a better buy than a 310S. Many of our customers do in fact >create compute farms of many 4D35S. It is entirely dependent on your >application as to wether 8 4D35S or 1 4D380 is a better buy. Remember that Well, somebody please correct me if I'm wrong, but doesn't the 3x0 use the same memory (arch.) as the 2x0? If so, then it's seriously max'd out on a 380. If you want to buy a 3x0 with the thought of upgrading to a faster machine down the road, then definitely don't load the thing up with 3rd-party memory unless you've got some other machine to dump it into later. The 35 will, without doubt, be much less expensive to upgrade in view of the fact that the tricked-up memory architecture will support a much faster processor (ie, your memory will survive the upgrade). The main performance issue is memory utilization and cache size. The data/instruction cache sizes grew from 8k/16k on the 4D20 to 32k/64k on the 4D25. I would hope that the data cache size on the 35 is 64k. That would bring it into line with the IP7 64k cache size. A small data cache will hurt performance on computations using traditional memory access. "block-mode"-style algorithms can improve a cache-machine's performance on big problems, however once the run queue starts getting big the cache can still thrash. That's why the 4D20 was such a pig. So, what are the data cache sizes of the 35 and the 3x0?... It's always been true that esd-type people have had a certain pressure to not produce a too-versatile product. The typical result has been a smaller cache and limited i/o bandwidth. It's true that to a limited extent this reduces cost, however the main result is a machine less suited to a time-sharing environment. I'd have to say that the 3x0S really isn't an attractive option right now (for most people). I'd rather buy a couple of 35's and ride out the inveitable new product announcements. Of course, I could be completely wrong in my assumptions about the 3x0 memory architecture...