Chinese 68060 Rev 6

[Damion]

The most plausible explanation is that the chips are recycled, the date code on the chip in the pic says it was made in 99 (a year that mask was produced). Apparently some lesser rated chips were relabeled a while back, but I don't think that's the case here.

$55 isn't that expensive for Rev 6 '060s btw, although I've seen them even cheaper from sellers who don't know amiganerds are after them. ;-)

[Damion]

@Miloo

I just ordered one, when it arrives I'll let you know.

[Damion]

AFAIK the only "fake" was a picture of one relabeled to a full (non-existent) 75MHz part. I don't know if anyone was actually scammed. In this case, I don't think it would be worth the trouble just to sell the odd CPU to an amiga or atari user. Plus, anyone can check the real revision in a matter of minutes and post scathing feedback... hardly worth it over $50.

[Damion]

Hasn't arrived yet, but hopefully I'll get it in time for the weekend... (and hopefully it works, and overclocks, lol)

[Damion]

Well, mine showed up today... but unfortunately, the mailman was too lazy to walk to the door (which they are required to do btw), and left a slip in the mailbox instead... so it'll be Monday eve before I can check it out. :evil:

 

[Damion]

Had a chance to test mine, it looks new and is absolutely a Rev 6. (I'll try to get a pic and maybe a WhichAmiga screenie up sometime later.) According to the date code it was manufactured in '02. The seller is fine, packed everything well/etc, totally recommended.

Unfortunately, I can't really test overclocking, as it seems the TekMagic won't run an '060 at '040 BCLK like the Apollo cards (won't boot), and the Rev 6 in my Apollo 1260 is soldered. :/ It does run just fine at 64MHz 1:1, but I don't want to risk pushing the bus any further without research.

In a few days I'll compare the temperatures of the Rev 1 and Rev 6 in the same card. The Rev 6 feels a bit cooler, maybe 5c or so. (A small copper heatsink like the enzotech CNB-SL1 makes a massive difference in temperature, keeps them cool even at 80MHz.)

In regard to some of the pics from the seller where it looks like the face of the chip is damaged, it looks to me like some glue residue from the factory Motorola heatsinks, like seen on the VMEbus boards. Mine was in absolutely perfect condition though. :pint:

[Damion]

OK, as promised some temperatures.

Load temps after 1 hour of Quake (demo loops), smallest window and screensize. 50MHz, A2000 case, no additional cooling (just the PSU fan). Measured with my good 'ol 3M Scotchtrak, at the warmest spot on the chip I could find. Ambient 19C.

Idle/Load

Rev 1: 55/62C

Rev 6: 51/57C

Roughly a 5C difference, IMHO not bad for 2 revisions of the same chip (the process shrink obviously helped).

Then I added an enzotech cnb-s1l + ceramique to the Rev 6, and measured again:

44/51C - A decent improvement, load was now where it idled without a heatsink! Ambient was also up a few degrees from the earlier tests. Temp might even come down a degree or so more after the paste breaks in.. :p

The above plus a small fan, and amazingly the load temperature drops to between 28 and 35C, depending on the amount of airflow (doesn't take much, an undervolted 40mm will do). On my Apollo 1260, load is always under 50C at 80MHz, in the desktop case, with the same copper cooler and a 40mm fan at 9v.

[Damion]

miggy1974 wrote:
hi mate, thanks very much for the info, i will try and get 50ns compatible simm you mention on ebay and overclock in small steps to see what my blizzard will actually achieve

Thanks again

The Blizzard 1260 usually hits 66MHz OK, the SCSI kit may or may not work at that clock depending on the revision of the SCSI kit. Mine started to act funny around 70-72MHz.

As matthey stated, the external bus is a limiting factor. If you can set the clock divider to 040 setting (BCLK 1/2 PCLK), in theory 80MHz should be no problem, RAM would then run at 40MHz. On the Apollo cards the divider is easily set with a jumper. (Even with the RAM at 1/2 CPU, the thing flies at 80MHz, so it's worth exploring if you can figure it out.) Not overclocking the bus puts a lot less stress on the card IMHO.

[Damion]

DoogUK wrote:
Is the apollo a better choice for overclocking the 68060 on?

It's definitely the easiest, for sure. Running the bus at 1/2 CPU clock is easy on the card, and the Apollo has a nice RAM controller, so 40MHz RAM is still plenty fast.

Here's a bustest benchmark from an Apollo 1260 at 80MHz CPU/40MHz RAM:

BusSpeedTest 0.19 (mlelstv) Buffer: 262144 Bytes, Alignment: 32768
========================================================================
memtype addr op cycle calib bandwidth
fast $01458000 readw 53.7 ns normal 37.2 * 10^6 byte/s
fast $01458000 readl 95.5 ns normal 41.9 * 10^6 byte/s
fast $01458000 readm 95.6 ns normal 41.8 * 10^6 byte/s
fast $01458000 writew 70.5 ns normal 28.4 * 10^6 byte/s
fast $01458000 writel 140.9 ns normal 28.4 * 10^6 byte/s
fast $01458000 writem 141.1 ns normal 28.3 * 10^6 byte/s
chip $00060000 readw 968.1 ns normal 2.1 * 10^6 byte/s
chip $00060000 readl 970.0 ns normal 4.1 * 10^6 byte/s
chip $00060000 readm 970.3 ns normal 4.1 * 10^6 byte/s
chip $00060000 writew 613.1 ns normal 3.3 * 10^6 byte/s
chip $00060000 writel 614.0 ns normal 6.5 * 10^6 byte/s
chip $00060000 writem 614.1 ns normal 6.5 * 10^6 byte/s


Now, here's one from a CSMK3 at 50MHz CPU/RAM:

BusSpeedTest 0.19 (mlelstv) Buffer: 262144 Bytes, Alignment: 32768
========================================================================
memtype addr op cycle calib bandwidth
fast $08718000 readw 64.2 ns normal 31.2 * 10^6 byte/s
fast $08718000 readl 107.8 ns normal 37.1 * 10^6 byte/s
fast $08718000 readm 108.9 ns normal 36.7 * 10^6 byte/s
fast $08718000 writew 72.3 ns normal 27.7 * 10^6 byte/s
fast $08718000 writel 144.9 ns normal 27.6 * 10^6 byte/s
fast $08718000 writem 150.2 ns normal 26.6 * 10^6 byte/s
chip $00130000 readw 1005.6 ns normal 2.0 * 10^6 byte/s
chip $00130000 readl 1006.3 ns normal 4.0 * 10^6 byte/s
chip $00130000 readm 1005.7 ns normal 4.0 * 10^6 byte/s
chip $00130000 writew 612.7 ns normal 3.3 * 10^6 byte/s
chip $00130000 writel 612.9 ns normal 6.5 * 10^6 byte/s
chip $00130000 writem 614.4 ns normal 6.5 * 10^6 byte/s  


As you can see, even with a 10MHz disadvantage in RAM speed, this setup on the Apollo is still damn fast (as far as classic miggys go, LOL).

 

[Damion]

matthey wrote:
I got my 75MHz oscillator today and so far it's been rock solid in my CSMK3. Here's my bustest...

BusSpeedTest 0.19 (mlelstv)   Buffer:     262144 Bytes, Alignment: 32768
========================================================================
memtype   addr       op         cycle     calib         bandwidth
fast      $08AB0000  readw      35.7 ns   normal      56.0 * 10^6 byte/s
fast      $08AB0000  readl      58.6 ns   normal      68.3 * 10^6 byte/s
fast      $08AB0000  readm      59.7 ns   normal      67.0 * 10^6 byte/s
fast      $08AB0000  writew     39.6 ns   normal      50.5 * 10^6 byte/s
fast      $08AB0000  writel     79.3 ns   normal      50.5 * 10^6 byte/s
fast      $08AB0000  writem     82.8 ns   normal      48.3 * 10^6 byte/s
chip      $00020000  readw     859.6 ns   normal       2.3 * 10^6 byte/s
chip      $00020000  readl     860.1 ns   normal       4.7 * 10^6 byte/s
chip      $00020000  readm     859.8 ns   normal       4.7 * 10^6 byte/s
chip      $00020000  writew    565.3 ns   normal       3.5 * 10^6 byte/s
chip      $00020000  writel    565.6 ns   normal       7.1 * 10^6 byte/s
chip      $00020000  writem    565.8 ns   normal       7.1 * 10^6 byte/s

 :-o  That's quick. I can match your chip numbers if I boot without startup (or use a low-res screenmode), no way to come near the fastram speed though, LOL.

Any chance you could to a Quake timedemo sometime (no hurry), default settings with AGA/NTSC? It would be interesting to see what kind of effect the fast memory has on the result.