Jump to content

WGS 9150: Crescendo + HPV possible?


Recommended Posts

edit: Page 2, HEH!

 

11 hours ago, jeremywork said:

9150s are rare enough that it's a pipe dream, but I would take the shot if an opportunity arose.

 

Rare enough doesn't quite say it. Here's Cameron Kaiser's WGS 9150 page: https://www.floodgap.com/retrobits/ans/9150/

He hasn't gotten back to me about posting a few of his pics here for illustration purposes, maybe later, meanwhile check 'em out.

 

Quote

If one could retain four Nubus slots in addition to HPV/G4 that'd be pretty sweet. The /120 model will also have BART21, so potentially the second fastest/most complete Nubus Apple shipped, behind the 840AV's. The 8100/100 I already have isn't far behind, and would be theoretically identical if I clocked it at 40/120MHz, minus one Nubus slot.

 

I fail to see how the stock 9150/80 could be considered slower than an 840AV? If you're saying the code base was not yet up to making full use of the 601 while they were barely contemporary, that's one thing. But the two PPC Macs are clearly better packages and faster if not equal in overall machine0. Even your 8100/100 should fare better running proper code. The Quadra 840AV ain't really all that outside of the 68K arena.

 

WGS9150vsQ840AVvsPM8100-100.thumb.JPG.1c7dad524adf1147726f7c64eca9edc5.JPG

 

VRAM is 1/2 or 2/4 on the HPV variants so that's no contest for your 8100 though not a standard feature on the 9150. System clock is equal for the 9150 and Processor clock doubles, 9150 has four times the 840AV's L1 Cache and 512k of L2 where the Quadra has none. Same is true of you 8100 with the fastest CPU clock of the three. The only place your 8100 lags would be System Clock and the 4x L1 bump and addition of 256k L2 should make up for or slightly better the Quadra's best in overall performance. The 33MHz-40MHz jump that made the 840AV so fast as compared to its 33MHz predecessors wasn't all that important once you moved out of the realm of unified bus architecture and into the world of Front Side/Back Side buses. The DevNotes call that the Processor Bus and Slow I/O Bus which debuted in PowerBooks/Duos. I don't recall a stationary 68K machine with a split bus offhand? Does anyone?

 

Now, unfortunately I'm even more interested in the 9150. With that fourth slot available for an extra striped JackHammer array, did the 9150 even have a need for the likes of the Apple Workgroup Server 95 PDS Card, given performance of contemporary HDDs?

 

Neat stuff! Who's keeping an eye out for one? :lol:

 

Edited by Trash80toHP_Mini
Link to post
Share on other sites
  • 68kMLA Supporter
1 hour ago, Trash80toHP_Mini said:

I fail to see how the stock 9150/80 could be considered slower than an 840AV?

Not the machine's performance as a whole, but the maximum Nubus performance. Apple's BART chips tie the 601 bus to Nubus, and it's not done quite as efficiently as the 840AV's Nubus 90 implementation. ATTO SE IV scores are more like 16MB/s on my 8100, where 840AVs will get over 18MB/s. 

 

Radius VideoVision documentation calls this out, claiming the 9150/80 is identical to the 8100/80, and both are slower than the 840AV: http://archive.retro.co.za/mirrors/68000/www.vintagemacworld.com/radius/vvision.html

 

Furthermore, there's a bug in the initial PPC BART chip (BART4) which disabled slave block transfer on all cards unless every card supports it. BART21 corrects this flaw, allowing fast cards to be fast while slow cards exist. The only Apple-shipped machines confirmed to always have BART21 are the 8100/100, 8100/110, and 9150/120: http://radius.vintagebox.de/Support/dv/Bart.html

 

1 hour ago, Trash80toHP_Mini said:

With that fourth slot available for an extra striped JackHammer array, did the 9150 even have a need for the likes of the Apple Workgroup Server 95 PDS Card

The WGS pisces card wouldn't install in a 9150 without an 040 PDS slot, however the 8100 introduced the second Fast-SCSI II internal-only bus (10MB/s over 5MB/s standard- not quite fast & wide but better.)

The 7100 didn't have this but I'd be quite surprised if the 9150 also didn't, especially considering the 950 before it already had two SCSI busses (though not different speeds.)

 

1 hour ago, Trash80toHP_Mini said:

Here's Cameron Kaiser's WGS 9150 page: https://www.floodgap.com/retrobits/ans/9150/

This was a fun read, thanks!

Link to post
Share on other sites
1 hour ago, jeremywork said:

The WGS pisces card wouldn't install in a 9150 without an 040 PDS slot, however the 8100 introduced the second Fast-SCSI II internal-only bus (10MB/s over 5MB/s standard- not quite fast & wide but better.)

 

Yep, that PDS is card is 33MHz in the 950/95 040 PDS whereas the HPV/G3 PDS bus is 40MHz.

 

8 minutes ago, Unknown_K said:

The WGS 95 card with cache is nice but there is a rare PDS SCSI card for the 950 that is probably much faster then anything Nubus.

 

I wonder what that gem looks like? PDS is a fatter pipe, but a pair of JackHammered Fast/Wide RAID setups (with redundancy) would have be fine for a 9150 with the HDDs of the era on that pair of NuBus slots? If the SCSI Controller ICs on the PDS card were no faster than the JackHammer's, and there's little reason to think they would be, PDS might not make that much a difference? Dunno, I can only fit the one JackHammer in my Radius 81/110 along with VideoVision and Thunder IV GX/1600. With the G3/HPV combo tucked in there as well I'm very happy indeed. Besides, who really needs that fourth 9150 slot and a second JackHammer? Anyway, I've pretty much been a Radius collector from day one when the spankin' new SE/Radius16 arrived and I started drooling on the Radius FPD ads in MacUser. [:)]

Link to post
Share on other sites
  • 68kMLA Supporter
40 minutes ago, Unknown_K said:

The WGS 95 card with cache is nice but there is a rare PDS SCSI card for the 950 that is probably much faster then anything Nubus.

Are you referring to the MicroNet Raven Pro or Storage Dimensions Data Cannon?

IIRC MicroNet claimed 17MB/ throughput in their marketing, which should be even faster with modern drives. Theoretically still in the same class as an 840AV with SE IV or Jackhammer, but without needing to share the bus the real world performance should be noticeably better. Too bad the 950's un-cached memory bus seems to only move 10-12MB/s in my experience. With your 50MHz/128k unit it should be much happier.

 

I dropped a bunch of 950 memory benchmarks over here the other day:

 

 

 

Edited by jeremywork
Link to post
Share on other sites

Well after a bit of under the bed diving, I'm suddenly back in the 10-ring of the topic!

 

WGS_9150-CrescendoG3-HPV_Adapter_Mockup-00.thumb.JPG.f7532a4d116b48c519dd132a9cd746e9.JPG

 

We're talking silly millimeters in width and height for the possibility of 4-Slot 9150 glory. Oh so very very close a run. Hot air rework not done, but necessary. Height should be good, the PCI risers slots are sitting higher than the PDS/NuBus backplane.

Link to post
Share on other sites
  • 68kMLA Supporter
2 minutes ago, Trash80toHP_Mini said:

Well after a bit of under the bed diving, I'm suddenly back in the 10-ring of the topic!

 

WGS_9150-CrescendoG3-HPV_Adapter_Mockup-00.thumb.JPG.f7532a4d116b48c519dd132a9cd746e9.JPG

 

We're talking silly millimeters in width and height for the possibility of 4-Slot 9150 glory. Oh so very very close a run. Hot air rework not done, but necessary. Height should be good, the PCI risers slots are sitting higher than the PDS/NuBus backplane.

Height looks very promising, but yes that initial bend at the Crescendo will be brutal. 

 

Luckily after doing some rough measuring on that 9150 prototype photo it seems that the PDS center is slightly further than the nubus cards are to each other. Good news for this idea...

 

I just copy-pasted this green line three times, so all of them are the same length, touching center to center of each Nubus, but only reaching the outer edge of the PDS.

uldz542tkm211.thumb.jpg.9cc2de39f68d9d52950fe4db8e80c2ce.jpg

Link to post
Share on other sites

I looked at the same thing, but you're measuring connector to connector. You need to estimate PCB to PCB spacing, which makes them more or less equidistant to my eye so I didn't go to the OS9 graphics bubble of the QS02. PDS is centered in its edgecard connector while the NuBus card PCBs would bea at the very top edge of the outside of the EuroDIN connectors. Proof of that is to be found in the even spacing of the five Backplane slot covers.

Link to post
Share on other sites
  • 68kMLA Supporter
3 minutes ago, Trash80toHP_Mini said:

I looked at the same thing, but you're measuring connector to connector. You need to estimate PCB to PCB spacing, which makes them more or less equidistant to my eye so I didn't go to the OS9 graphics bubble of the QS02. PDS is centered in its edgecard connector while the NuBus card PCBs would bea at the very top edge of the outside of the EuroDIN connectors. Proof of that is to be found in the even spacing of the five Backplane slot covers.

Ah I overlooked that, which would perfectly explain that spacing. At least it's not *less* than standard width :)

Link to post
Share on other sites
  • 68kMLA Supporter

Okay- super wacky idea...

Can the PDS riser be used before the Crescendo? Stick the HPV directly into the Crescendo upside-down in slot 1 and run the ribbon back to motherboard PDS?

 

________________

|

|

|

|        _...-----.._

|---""  ________| (G3)

| (Slot 2) (^HPV Slot 1)

| (Slot 3)

| (Slot 4)

| (Slot 5)

""""""""""""""""""

Link to post
Share on other sites

Dunno, not sure my eyes are tracking together on your nice ASCII graphic. Pretty sure I blew up one my fancy Sonnet Cables trying to us it as a generic PDS extender? Memory is fading out due to attitude adjustment/manic wind down session for the night. Spent much of the evening opening up, going through piles of mail, sorting and organizing the storage container layers into a wonderfully compacted format. IOW, what I use for a brain's mostly offline ATM. :blink:

Link to post
Share on other sites
  • 68kMLA Supporter

Unfortunately ASCII's a bit low res ;)

 

Picture the Crescendo and HPV installed properly in a 6100, then imagine flipping the whole assembly upside down and inserting the top of the HPV into the bottom of the 9150's HPV slot- then use the riser to connect the Crescendo to the motherboard PDS (unless your experience with this is bad news)

Link to post
Share on other sites

Yea, I tried something like that, I don't think it's healthy for cable or accelerator given the logic on the cable. We'll have to wait for @Bolle to chime in on that one, but it's just a matter of curiosity. Methinks I've been overthinking the problem.

 

The NewerTech cable positions the HPV overtop and at a right angle to the NuBus cards in the generous headspace of the 81/110. The Sonnet kit can place the HPV card in the generous floorspace of the case parallel to the third NuBus card. Cable length precludes that solution for the 9150 because of the fourth NuBus slot even if it had the required floorspace. The NuBus card's component sides face down in the 9150, so it doesn't appear to have the required floorspace in the exploded diagram.

 

So I've missed the most simple solution for the 9150 while target fixated on adapting known solutions for my clone. The answer was sitting in the pictures above all along as a rough reference point for slot spacing. The WGS 9150 needs only to be equipped with a machine specific Twin PDS slot riser card. ::)

 

WGS9150-TwinSlot-PDS-Riser-03.thumb.JPG.88667f31ed4faa8cab1673c0703e6e10.JPG

 

Du'oh! [:I] The way I'm seeing it now, the setup has some fairly simple aluminum angle and bar construction to hold the two PDS cards in position. Provision can also be made for holding the NuBus cards in place in the manner of the removable bar in my 81/100. Add power headers to the riser and mounting points on the aluminum support structure and a fan is positioned inboard the riser for cooling. Such might approach the most elegant of Rube Goldberg bodgery.

 

I've got riser like the single slot card with a second slot above the first. That would have made for a clearer example, but couldn't find it. The cabled riser is an off the shelf PCI plaything. A pair might be hacked to do the job, which would be truly Rube Golbergian. But it roughly approximates the height of a WGS 9150 PDS TwinSlot Adapter.

 

WGS9150-TwinSlot-PDS-Riser-04.thumb.JPG.6e238a1a2b8d632ddc923e56d8228f04.JPG

 

There, now I can start storing this cruft back in place.

Edited by Trash80toHP_Mini
Link to post
Share on other sites
  • 4 weeks later...

So I actually have a 9150/120. It's pretty beastly, just as imposing as a Q950 but with a few external changes that modernize it to more closely match the styling of the 6100 and 7100, including the relocated and now manual-inject floppy drive whose original home is occupied by a new CDROM (the former tape drive bay is still available for use if not occupied; mine's empty). These were one of those rare machines where it was available only as a WGS without a workstation counterpart. You could sort of count the WGS 95 in the same category since its special WGS cache card was unique to the server model but otherwise it could be configured as a workstation whereas the 9150 could not. Incidentally, as was the custom at the time, the 9150 logic board was offered as a drop-in replacement for the WGS 95, which makes it the only NuBus PPC model to use a DB15 monitor connector in addition to the fact that it keeps the Q950's audio in/out ports. Both the 80 and 120MHz versions ran the bus at 40MHz which was really fast at the time, with the 120MHz version being that much faster than the "flagship" 8100 which topped out at 37MHz in the 8100/110. Plus these could be configured with up to 1MB of L2 cache, as could the WGS 8150. Also, these are the only machines I've seen where the PDS terminator is required if the slot is unoccupied whereas all of the other NuBus PowerMacs don't care if the slot is empty or not. Why Apple didn't sell a PM 9100 is beyond me but they did all kinds of crazy crap for stupid reasons back then so who knows? They already had to build the "Fat AMIC" to accommodate the extra NuBus slot so I'm not sure why they didn't make a few more improvements to perhaps double the number of RAM slots; they had more than enough empty board space to do it.

 

Anyway there's really no way to use both the G3 PDS card with an HPV card and all 4 NuBus slots without either hacking up the case or using a custom cable that passed through the case and was connected externally; there's just no room in the case to do it cleanly and you'd need to build a new cable anyway. Thus, with either option you'd need a custom high-grade cable and precision workmanship to prevent signal problems since this is a high-speed bus connected directly to the processor. Maybe it would be possible to repurpose an Ultra320 SCSI cable or two? Those are built to handle high-speed data.

Link to post
Share on other sites

Yep. AFAIK all of them did since there were no workstation counterparts like the WGS 95 had. Mine had its original two 1GB IBM hard drives installed on the really long 7-connector SCSI cable. I think it was supposed to use a software RAID or something.

Link to post
Share on other sites
4 hours ago, Franklinstein said:

Yep. AFAIK all of them did since there were no workstation counterparts like the WGS 95 had. Mine had its original two 1GB IBM hard drives installed on the really long 7-connector SCSI cable. I think it was supposed to use a software RAID or something.

Yea, figured they were standard but just checking. I have only one from my wgs95 and have not seen any around. Would be cool to fab some up for normal 950's.

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...