Pegasos 2 today is nearly 15years old. It was great machine and still is not bad today. Only for web-browsing and video we need more power.
I overclocked Peg2 to 1.33MHz, equip with Radeon9800Pro and now test some HDD and SSDs:
( test was made with SATA300 HDD and SATA300 SSD )

Possibilities to connect HDD/SSD:
1. onboard IDE/PATA controller with SATA reduction
theoretical params: UltraATA 100, mode 5, transfer rate 100MByte/s

2. onboard FireWire400 controller
theoretical params: transfer rate 400Mbits/s = 50MByte/s
- usable only for backup HDD/SSD, or for connection of USB2.0 hub to save PCI slot

3. PCI card SII 3x1x - SATA I
theoretical params: transfer rate 150MByte/s SII3x1x, 133MByte/s PCI slot

4. PCI card Promise - SATA II
theoretical params: transfer rate 300MByte/s Promise TX4, 133MByte/s PCI slot

Here is graph from SCSISpeed Results:
srovnani-jpg

Best results - read speed, MB/s:
PCI card Promise - SATA II 114 MB/s
PCI card SII 3x1x - SATA I 84 MB/s
onboard IDE/PATA controller 55 MB/s
(FireWire not tested, too slow)

Unfortunatelly, Promise - SATA II card is not supported hardware and you must use third-party driver ulsata2.device. It mean, that you cannot boot MorphOS System: from Promise card, and it limit it's use only for Work: and Backup: volumes.

Why PCI card SII 3x1x - SATA I so slow? Theoretical limit shows that show stopper is PCI (133MByte/s), not SATAI (150MByte/s), and PCI is the same for SII and Promise, but:

Let's compare software drivers: (info from morphos-debug.log and SysMon)

MOS 3.11:
onboard IDE/PATA controller ide.device UDMA5=100MB/s S.M.A.R.T no
PCI card SII 3x1x - SATA I sata.device UDMA5=100MB/s S.M.A.R.T no
PCI card Promise - SATA II ulsata2.device UDMA6=133MB/s S.M.A.R.T yes

AOS 4.1FE:
onboard IDE/PATA controller peg2ide.device UDMA5=100MB/s S.M.A.R.T no
PCI card SII 3x1x - SATA I sii3114ide.device UDMA6=133MB/s S.M.A.R.T yes
PCI card Promise - SATA II no driver at all

It looks that sata.device is not ideal - recognize SATA HDDs only like UDMA5 devices, and have no S.M.A.R.T. feature. It is not hardware issue - AmigaOS recognized it correctly.

And what to do now? It leave us on Pegasos2 only this possibilities:
- System: (and complete AmigaOS) on IDE (55MB/s), other MorphOS/linux volumes on Promise(114MB/s)
- all volumes on SII (84MB/s)
- System: (and complete AmigaOS) on SII (84MB/s), other MorphOS/linux volumes on Promise(114MB/s). But it takes two of three PCI slots.

I don't know the priorities for next MorphOS, but improve of sata.device will be great, at least for Pegasos owners.

With the limitations of the PCI bus, all current systems that rely on PCI adapters will not reach the maximum transfer rate of SATA2 or SATA3.

So you either accept the limitations of your hardware, or you upgrade.

Right now, an X5000 would be a costly upgrade compared to a Peg2 (considering the original price of the Peg2).

Here is where late 2005 G5 support would be nice. We could at least upgrade to an SATA2 controller.

An SATA3 driver might also be possible, depending on the transfer limitations of PCIe Gen1.

Quote:

yes, it is true. I think that Promise - SATA II 114 MB/s read speed reach the maximal limit of Pegasos2 PCI bus. This number is really very high, 85% of theoretical maximum 133MB/s.

if I had no X1000, then yes. But X5000 power is nearly the same as X1000 (except for MorphOS). I am accept now that my Peg is old and slow, and wait for G5. Or maybe for Talos.

But - I wrote this, becouse we don't rech the maximum from Pegasos yet.
MorphOS native sata.device recognized SATA disks only like UDMA mode 5 = 100MB/s.
Even AmigaOS sii3xxxide.device recognized the same hardware correctly (UDMA mode 6 = 133MB/s) - there is space for speed increase.
Or upgrade our native sata.device, or maybe incorporate ulsata2.device onto Morphos boot.img

@MOS devs:
nothing against. This sata issue (and sdl drivers) are only two things which are better on AmigaOS. I prefer MorphOS over AmigaOS and I still use more my old Pegasos 2 over X1000.

In some ways the X1000 was superior to the X5000. For one, it has more PCIe lanes. And I believe the PA6T supports AltiVec.

> Pegasos 2 today is nearly 15years old.

Release was in December 2003, so it's 15½ years :-)

> I overclocked Peg2 to 1.33MHz

;-)

> Here is where late 2005 G5 support would be nice.
> We could at least upgrade to an SATA2 controller.

The 133 MHz PCI-X slot of older G5s is enough to not limit one SATA3 port even. While SATA3 cards may not exist for PCI-X, SATA2 cards definitely do, theoretically being able to use up to 3 SATA2 ports simultaneously.

> An SATA3 driver might also be possible, depending on
> the transfer limitations of PCIe Gen1.

3 PCIe v1 lanes are enough to not limit one SATA3 port. There are SATA3 cards for up to 8 PCIe v1 lanes, which "late 2005 G5" happens to have a slot for, resulting in the theoretical ability to use up to 3 SATA3 ports simultaneously.

> In some ways the X1000 was superior to the X5000. For one,
> it has more PCIe lanes. And I believe the PA6T supports AltiVec.

Indeed, it does. A while back, we worked out some more examples of areas where the X1000 is superior to the X5000:

https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=12151&forum=11&start=16

Quote:

maybe I buyed him in 2004? I have to look to archive.

sorry, of course 1.33 GHz
1.33GHz not bad, it outperforms SAM460 in Tower57 "benchmark" and have 3003 MIPS/core compares to 3960 on X1000 and 4750 on G5Quad

Last things what I can do for speedup is SATA (this topic), and overclock Radeon9800pro. I cannot reach the 9800XT frequencies, but yesteday I tested succesfully overclocking core 378->394.88, and memory 337.5->351 frequencies. Now I need change BIOS and try it on Pegasos
Maybe later, If I find some cheap HD4xxx I will also test it on PCI slot.

Quote:
I completely agree.
and MorphOS was opimised for AltiVec from beginning.

Quote:

How do you overclock a Peg2?

Quote:

on G4 CPU card you have to remove SMD resistor and sold here dip-switch. And then test maximal frequency without errors. Good cooler is a must.

I made it thanks to Prober and tlosmx. Deatiled manual is here:
Probers.cz

and here are links to my articles:
Amiga Klub Forever + Czech PowerPC User Group
Hyperion - Pegasos II

in case of interest I can try translate some parts to English. (Pidgin English, of course)

I did it according Probers manual, but later I changed cooler to ARCTIC Alpine M1-Passive Socket AM1 - because Thermalright HR-05 SLI blocked way to IDE connectors.
I made modifications on Arctic cooler (to fit in place) and supply with fan.

Quote:

Interesting. Do you think you could remove the MPC7447 CPU from a Peg2 G4 board and refit it with an MPC7447A?

That would guarantee 1.4 GHz operation (or higher).
I believe they are pin and voltage compatible.

The Peg2's big bottleneck is the MV64361 northbridge. Overclocking the CPU would result in rather poor gains for the effort, I would think.

Quote:

I don't think so.
With higher frequencies ( 1.4 GHz and above ) I had memory read errors. From my opinion is northbridge MV64361 responsible for this, not CPU.
And also remove SMD and sold dip-switch is easy task, if I replace CPUs - there can be much bigger chance to disrupt something.

Even if I will succeed replace CPUs and northbridge have no errors with MPC7447A - the max frequency is 1.464 GHz - about 10% more than I have today. It is only small surplus with great risk.

Quote:

I overclocked from 1.0 to 1.33GHz
and speed increase is markant:

memory: +3.5% (stream) +13% (RAMSpeed)

HDD, onboard IDE +24%
HDD, SiI SATA +69%
(Promise SATA2 is even much better cca +200% - but it is extrapolation from different comparisons)

Linux Hardinfo +27%

SDLbench +80% (including 9200->9800PRO replacement)

the most important is, that my subjective feeling of speed increase is great, and I can again play youtube videos and most of movies.

[ Edited by sailor 31.05.2019 - 10:53 ]

Quote:
Kudos to your overclocking efforts! Did you perhaps measure the power increase under load? I also wonder if or how much overclocking of hardware of that period reduces the lifespan of the hardware?

Does not matter much if you still run an e.g. Athlon 64 as you can easily replace it with some other cheap bog standard PC - but it does matter if your want to keep your 'legacy' G4 7455 hardware around.

[ Editiert durch ernsteiswuerfel 31.05.2019 - 15:19 ]

Quote:

Thanks.
No, I don't measure the power. Maybe I have to build some testbed next time.
Completely agree in this isuue.
Overclocking may lead to reduced lifespan, and it is very unlikely with Pegasos.

Only things what I did against this is new power supply and proper cooling:
I used oversized passive cooler with fan, passive cooler on RAM, passive coolers on Northbridge and Southbridge.
Rear side of CPUcard now have much lower temperature than with original frequency and standard cooler. ( measured by finger - another thing for which is testbed good for)

> With higher frequencies ( 1.4 GHz and above ) I had memory read errors.
> From my opinion is northbridge MV64361 responsible for this, not CPU.

The northbridge being responsible for this doesn't seem plausible to me. The northbridge does not care for the CPU clock rate at all. It runs at the same clock rate no matter the CPU clock rate.

> the max frequency is 1.464 GHz

Is it a given that 11x is the highest possible multiplier? After all, the 5 binary digits of the PLL config can theoretically represent 32 different numbers, and with 0.5 steps this could amount to a 2.07 GHz range.

> I also wonder if [...] overclocking of hardware of that period
> reduces the lifespan of the hardware?

It does, always.

> it does matter if your want to keep your 'legacy' G4 7455 hardware around. ;-)

Agreed, apart from the fact that the Pegasos II uses newer generation 7447 ;-)

> Overclocking may lead to reduced lifespan, and it is very unlikely with Pegasos.

Why is it "very unlikely with Pegasos"? Or do you mean your Pegasos in particular, due to your extra cooling measures?

Quote:

no, I simply mean this my Pegasos, because it is my best MorpOS machine. If I burn it, I have to wait for Talos port

Quote:

maybe, I don't know and have no way how to test it. Higher frequencies then 1.333 makes memory error.
Of course, It can be done on processor side. But also some synchro issues on northbridge.

yes, in theory 5 pin = 32 possibilites.
But if I see the table obr-3.jpg I have no idea how exact set higher frequencies.
And for sure, I will try such things only with spare Pegasos or at least spare CPU card.

>>> Overclocking may lead to reduced lifespan, and it is very unlikely with Pegasos.

>> Why is it "very unlikely with Pegasos"?

> I simply mean this my Pegasos, because it is my best MorpOS machine.

Okay, so not "unlikely" (which means "improbable") then, but "unwanted" :-)

> If I burn it, I have to wait for Talos port ;-)

I'd prefer Blackbird :-)

Quote:

yes - unwanted! I said that my English is Pidgin
I am looking forward to decision: Blackbird with MOS or Talos with MOS

Quote:
I return to this topics:
Please, do somebody use sata.device on Morphos machine other than Pegasos2 ?
And if, please can you check the morphos-debug.log, if your drives recognized like UDMA5=100MB/s or UDMA6=133MB/s.
Many thanks.

[ Edited by sailor 03.06.2019 - 10:10 ]