Jump to content
ktkm

JackHammer in IIfx no boost in perfomance

Recommended Posts

I’m currently testing two identical JackHammer cards (3.2) to see in what condition they are. But I’m not sure if my set up is optimal. I thought an IIfx would be a decent candidate to test them with, but I’m not getting any increased performance (I get approximately 1,15 MB/sec R/W with a Seagate HD 160SC). I have tried different setups in the JackHammer control panel, and both cards behave precisely the same. I have followed all instructions in the manual (I think). However, I suspect some kind of termination issue since I’m getting error messages when testing the SCSI chain in Lido, and Snooper won’t recognise the name of the drive, and Mac Bench 3.0 can’t even put it to the test it (it’s greyed out).

Share this post


Link to post
Share on other sites

Well, I switched to a SCSI2SD 5.1 it worked for a while, but it lowered the performance further until I switched to 32-bit memory and started to get bus errors at startup. I also tried them in two different IIci:s, with and without a Turbo 601 and SCSI 4.3 activated, with varied results. Sometimes I got a grey screen at startup, and occasionally it would merely freeze halfway through.

 

What performance increase would you usually gain from a JackHammer?

Share this post


Link to post
Share on other sites

I've got a Jackhammer in my IIfx, connected to a 4GB 68-pin drive (DEC RZ29B). The other drive (Quantum LPS540S) is connected to the internal 50-pin bus. In ATTO Performance Utility, the 50-pin drive reaches about 1.4MB/s and the 68-pin Jackhammer drive reaches about 4.0MB/s. Boot time difference between the two drives with identical OSs is not nearly as dramatic, but I'd expect the reduced latency of a flash solution would make even better use of the wider pipe.

 

Hope this helps!

 

Edit: Also, those tests were conducted in Mac OS 7.6.1. Note that the IIfx does have onboard SCSI DMA (said to reach up to 3MB/s) but DMA is only implemented in A/UX (just like the WGS Pisces card.)

 

Jackhammer.jpg.52e8f47248d1136ad074c92197d0b225.jpg

Edited by jeremywork

Share this post


Link to post
Share on other sites

My IIfx and Jackhammer with SCSI2SD v6.  I initialized the drive with FWB Toolkit, and updated the driver to FWB's solution.  I also changed the max chunk size to 65535 in the FWB Disk Driver configurator to achieve these results.  I'm pretty happy, although I thought I'd see better write speeds.

 

Picture-10.png

Share this post


Link to post
Share on other sites

Absolutely, Jackhammer's a Fast/WIDE SCSI II card for drives of the same kind. IIfx/SCSI2SDv6 is more suited to a fast/NARROW card the likes of the ATTO SEIV in my IIfx. Save Jackhammers for late era Quadras and X100 PPC machines that can make proper us of them. If you're not striping a pair of fast/wide drives for VidCap or the like, you really aren't doing the Jackhammer any justice. YMMV [;)]

Share this post


Link to post
Share on other sites

What Unknown_K said, twice.

 

Your disk performance is limited by the slowest performing item in the system.   The bottleneck.

 

In this case, the bottleneck has probably been the storage device in every case.

 

Back when, I had some Seagate ST32550N and ST32550W drives.  These were some of the very first Barracuda drives which spun at an amazingly fast 7200 RPM.  :-)

 

The N variant had a narrow SCSI interface (5 or 10 MB/s depending on unenhanced vs. Fast SCSI).   The W drives had a Fast & Wide interface (20 MB/s).

 

They both delivered exactly the same 6MB/s performance when installed on a JackHammer or on the Fast bus of an 8100.

 

It didn't matter that the interface could do 20 MB/s;  the drives couldn't deliver data any faster than 6 MB/s.  

 

If you want a good test of the potential of your IIfx with JackHammer, you need a SCSI drive that can deliver data faster than the interface you're trying to test.   More modern SCSI drives probably can do this.

 

Of course the 68030 may not be able to make good use of the interface.   I'm not sure about that.

 

BTW, IBM, when they sold hard drives, before they sold their drive business to Hitachi, used to include "Media Data Rate" (I think that's right, the first word was "Media" in any case) in their datasheet for their hard drives.   This was the rate at which you could expect the platters and heads to actually deliver data, regardless of what the interface potential was.

 

As far as I know, IBM was the only company that included this bit of honesty in their hard drive specifications, and pretty much everyone overlooked it.  :-)

Share this post


Link to post
Share on other sites
9 hours ago, Trash80toHP_Mini said:

Absolutely, Jackhammer's a Fast/WIDE SCSI II card for drives of the same kind. IIfx/SCSI2SDv6 is more suited to a fast/NARROW card the likes of the ATTO SEIV in my IIfx. Save Jackhammers for late era Quadras and X100 PPC machines that can make proper us of them. If you're not striping a pair of fast/wide drives for VidCap or the like, you really aren't doing the Jackhammer any justice. YMMV [;)]

Don't the SEIV and the Jackhammer both support Fast/Narrow and Fast/Wide?

I opted to save the SEIVs for my newer machines, since only the AVs and 601 machines provide the SCSI manager 4.3 it needs to be bootable- the Jackhammer does this internally so it works in any Nubus machine. I realize boot support is not really the original intention of these cards, but for use with a SCSI2SD I'd imagine being bootable is a big plus. SEIV and a modern 10k drive will do 18MB/s on an 840av, and 16MB/s on an 8100/100, I can't imagine the Jackhammer would do dramatically better (but I guess I'll test it at some point.)

Share this post


Link to post
Share on other sites
16 hours ago, jeremywork said:

I've got a Jackhammer in my IIfx, connected to a 4GB 68-pin drive (DEC RZ29B). The other drive (Quantum LPS540S) is connected to the internal 50-pin bus. In ATTO Performance Utility, the 50-pin drive reaches about 1.4MB/s and the 68-pin Jackhammer drive reaches about 4.0MB/s. Boot time difference between the two drives with identical OSs is not nearly as dramatic, but I'd expect the reduced latency of a flash solution would make even better use of the wider pipe.

 

Hope this helps!

 

Edit: Also, those tests were conducted in Mac OS 7.6.1. Note that the IIfx does have onboard SCSI DMA (said to reach up to 3MB/s) but DMA is only implemented in A/UX (just like the WGS Pisces card.)

Definitely, thanks! Which Nubus slot do you use for the Hammer?

 

16 hours ago, joethezombie said:

My IIfx and Jackhammer with SCSI2SD v6.  I initialized the drive with FWB Toolkit, and updated the driver to FWB's solution.  I also changed the max chunk size to 65535 in the FWB Disk Driver configurator to achieve these results.  I'm pretty happy, although I thought I'd see better write speeds.

How many drives do you emulate? And how do you terminate the SCSI2SD v6 together with the Hammer? Have you done something extra with the SCSI2SD utility?

I’m gonna try it again and upgrade the drives with FWB Toolkit.

 

11 hours ago, Unknown_K said:

You still need some fast 68 pin wide drives to saturate the bus (or a pair of them striped).

Externally, internally or both?

 

9 hours ago, trag said:

Of course the 68030 may not be able to make good use of the interface.   I'm not sure about that.

But a Turbo 601 in an IIci should?

 

Thanks to everyone! I’m gonna do some more test with what I have learned today!

 

/K

Share this post


Link to post
Share on other sites
8 hours ago, ktkm said:

Which Nubus slot do you use for the Hammer?

It's in the slot next to the PDS-adjacent one (can't remember the order offhand- either #2 or #5)

I haven't tested whether there's a performance difference, just didn't want the 68-pin cable blocking more ventilation than necessary, and the 8•24 GC has SIMMs sticking out of it, so it's on the end.

 

8 hours ago, ktkm said:

But a Turbo 601 in an IIci should?

While the faster CPU will increase throughput capability, there are some differences in the Nubus itself through the model years:

• The Mac II series uses the original Nubus, which runs at 10MHz allowing an average of 10-20MB/s with bursts up to 40MB/s.

• The early Quadra series uses a partial implementation of Nubus 90, which runs at 20MHz for increased transfer speeds of around 30MB/s with bursts up to 70MB/s. The partial implementation here allows various cards on the bus to communicate between each other at Nubus 90 speeds, but they can only communicate with the CPU at original Nubus speeds.

• The AV Quadras and x100 PowerMacs completed the implementation of Nubus 90, so the CPU can also communicate with devices at Nubus 90 speed.

However, the original release of the x100 PowerMacs used a BART4 Nubus controller, which unintentionally disables Nubus block transfers on all cards if any one card cannot support the feature. The revised BART21 controller fixes this problem, and is present in all 8100/100, 8100/110, and 9150/120 models (rumored to be in some late model 7100/80 and 9150/80s, but not confirmed.)

 

*Nubus speed figures are from Wikipedia

 

In my experience with the ATTO SEIV, the complete Nubus 90 implementation makes a huge difference- compare 18MB/s - 16MB/s for the 840av - 8100/100 to 8MB/s in my 950 with a 601 @ 115MHz.

This is ATTO Performance Utility again- I need to try out the benchmark in FWB toolkit to see how the results differ. @joethezombie has a faster read result than I would expect to see in a IIfx, but all of my drives are spinning rust, so it's hard to compare.

Share this post


Link to post
Share on other sites
On 6/2/2020 at 7:14 PM, jeremywork said:
On 6/2/2020 at 5:15 PM, Trash80toHP_Mini said:

IIfx/SCSI2SDv6 is more suited to a fast/NARROW card the likes of the ATTO SEIV in my IIfx.

Don't the SEIV and the Jackhammer both support Fast/Narrow and Fast/Wide?

I think you're right. My bad, meant to say ATTO Silicon Express Fast/Narrow card. Dunno the version offhand, but it's not IV.

Share this post


Link to post
Share on other sites
On 6/2/2020 at 3:27 PM, trag said:

BTW, IBM, when they sold hard drives, before they sold their drive business to Hitachi, used to include "Media Data Rate" (I think that's right, the first word was "Media" in any case) in their datasheet for their hard drives.   This was the rate at which you could expect the platters and heads to actually deliver data, regardless of what the interface potential was.

Wasn't this also known as the "internal transfer rate"? If so, most drives did have this published. Check out the pages of hard drive specs at stason.org: most of them have it listed in the first few lines (assuming they're accurate). According to the info there, for example, an HP C3325A (2GB, 5400RPM) supports a 10MB/s external SCSI interface but only transfers internally at 5.5MB/s. A Conner CFA-1080S (1GB, 4500RPM) supports the same 10MB/s externally but has only a 4MB/s internal transfer rate. A Conner CFP-1060AVW (1GB, 5400RPM, supposedly AV-rated) has a 20MB/s fast/wide SCSI bus for external transfers, but still only rated for 4MB/s internally. Ouch. Not buying that one. IBM's 0662-S1D (1GB, 5400RPM) has a 10MB/s external SCSI bus and transfers internally at 5MB/s. Surprisingly that HP drive is supposedly the fastest of these and the only one to exceed 5MB/s internally. I did not expect this. I actually have one of those, and I guess I'll have to stick that on a fast SCSI bus in a late-model Mac rather than use it to upgrade a stock Quadra or something. Is there anything that can saturate that SCSI bus though?

Ah, here we go: IBM DCHS-39100DEF20W (9GB, 7200RPM) has a 40MB/s Ultra Wide SCSI interface and a 15.4MB/s internal transfer rate. This drive is more than a match for the fastest SCSI buses built into Macs and would nearly saturate the 20MB/s Fast/Wide SCSI bus of a JackHammer, depending on file sizes and how well the caches were utilized.

Share this post


Link to post
Share on other sites

Don't forget that the drives internal transfer rate is slightly different for reading and writing. AV drives were optimized for writing data, consumer desktop drives are optimized for reading data. You also have different real world speeds depending on if you have data on the inner tracks or outer tracks (empty vs full drive).

Share this post


Link to post
Share on other sites

This is what happened last night. I re-formatted all the ’SCSI2SD disks’, four of them in total and 2GB each, with FWB HDT 2.0.6. Then I switched to the JackHammer. At first, nothing happened, grey screen with the mouse pointer. I removed two Digidesign cards, and the system would start. I then manage to install the JackHammer Control Panel, and then I did some bench tests in FWB HDT in two set-ups.

 

First, no options checked. (Please note, my SCSI2SD drive ID is named ‘QUANTUM FIREBALL*’)

A8faJQa.jpg h9Aej6b.jpg

 

Last, all options checked.

Wioe6W0.jpg 49mrnIS.jpg

 

The SCSI2SD 5.1 in my IIfx is not even close to @joethezombie performance with the SCSI2SD v.6, but at least it seems to be functional. However, I think its time to get a new PSU, because every time I cold boot, the system won't start (grey screen with the mouse pointer). It takes at least two start-up procedures, with the JackHammer installed, every time if I leave it off for a while.

On 6/2/2020 at 5:16 PM, joethezombie said:

I also changed the max chunk size to 65535 in the FWB Disk Driver configurator to achieve these results.  I'm pretty happy, although I thought I'd see better write speeds.

FWB Disk Driver is that something that comes bundled with FWB Hard Disk Toolkit? The closest thing I can find is ’FWB SCSI Configure’.

 

Cheers!

Share this post


Link to post
Share on other sites

What SD card are you using in your SCSI2SD?  A slow one will absolutely limit your performance.  I will pull out the IIfx and verify my settings perhaps this weekend. 

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×