Page 10 of 15

Re: kludge’s TF534 build

Posted: Sun Mar 31, 2019 8:03 am
by Justme14
Hi Kludge,

Excellent,another TF534 brought alive !

Persistence reveals ,or as you earlier said, Never give up :D
Just read that 8bitdreams had a succes with your A600 adapter with a TF534 booting, this looks promising !

I love to try my luck with one of your A600 adapters for testing as I have several A600 here ,is it possible I could buy one from you ?( I Know they are not Finished and need a patch)

Re: kludge’s TF534 build

Posted: Sun Mar 31, 2019 11:03 pm
by kludge
So... It boots and runs. 9667 Dhrystones in SysInfo. But I'm getting some weird stuff like some WHDload demos quitting with something about "NMI Autovector", and the raytracing part of the Hardwired demo looking lite utter crap.

Also I'm not getting any faster speeds than about 1 MB on my CF card, even with cpu fastrom.

Am I having "normal" errors here or is something still wrong?

Re: kludge’s TF534 build

Posted: Sun Mar 31, 2019 11:46 pm
by 8 Bit Dreams
You need to disable Autovector in WHDLoad.prefs

Re: kludge’s TF534 build

Posted: Sun Mar 31, 2019 11:54 pm
by terriblefire
kludge wrote:
Sun Mar 31, 2019 11:03 pm

Also I'm not getting any faster speeds than about 1 MB on my CF card, even with cpu fastrom.
Pretty sure your CPU doesnt have an MMU then

Re: kludge’s TF534 build

Posted: Mon Apr 01, 2019 12:07 am
by kludge
terriblefire wrote:
Sun Mar 31, 2019 11:54 pm
Pretty sure your CPU doesnt have an MMU then
The markings say MC68EC030RP40B and SysInfo says it’s got an MMU. But the CPU is from China, so I guess there’s room for doubt.

Re: kludge’s TF534 build

Posted: Mon Apr 01, 2019 2:40 am
by alenppc
kludge wrote:
Mon Apr 01, 2019 12:07 am
terriblefire wrote:
Sun Mar 31, 2019 11:54 pm
Pretty sure your CPU doesnt have an MMU then
The markings say MC68EC030RP40B and SysInfo says it’s got an MMU. But the CPU is from China, so I guess there’s room for doubt.
I feel that I need to put this as my signature, but... DON’T rely on SysInfo. It is not accurate. At all. Install mmu.lib v46 from Aminet and try using mufastrom. If your cpu has no mmu, it’ll tell you. Also the 534 really pushes the ram at 50 MHz, a lot of chips will lock up at that speed. Try 40 instead.

Re: kludge’s TF534 build

Posted: Mon Apr 01, 2019 6:02 am
by 8 Bit Dreams
kludge wrote:
Mon Apr 01, 2019 12:07 am
The markings say MC68EC030RP40B and SysInfo says it’s got an MMU. But the CPU is from China, so I guess there’s room for doubt.
It can't have MMU if it's "EC"!
Also try different CF Card - have one SanDisk card here that doesn't give any speedup even with CPU FASTROM command

Re: kludge’s TF534 build

Posted: Mon Apr 01, 2019 7:18 am
by kludge
8 Bit Dreams wrote:
Sun Mar 31, 2019 11:46 pm
You need to disable Autovector in WHDLoad.prefs
Will do, thanks!
terriblefire wrote:
Sun Mar 31, 2019 11:54 pm
Pretty sure your CPU doesnt have an MMU then
alenppc wrote:
Mon Apr 01, 2019 2:40 am
I feel that I need to put this as my signature, but... DON’T rely on SysInfo.
[...]
Also the 534 really pushes the ram at 50 MHz, a lot of chips will lock up at that speed. Try 40 instead.
8 Bit Dreams wrote:
Mon Apr 01, 2019 6:02 am
It can't have MMU if it's "EC"!
Also try different CF Card - have one SanDisk card here that doesn't give any speedup even with CPU FASTROM command
I thought at least you could count on it detecting the capabilities of the CPU. Won't trust it anymore! I promise! :)

It is a SanDisk card, but as the CPU doesn't have MMU support it won't tell me much switching to another card.

I think the RAM chugs along. I'm expecting quite severe crashes if the RAM is unable to keep up, and I'm not seeing any of that yet. Frontier is _really_ smooth at least. Might order a couple of 40 MHz oscillators later for testing.

The raytracing part of Hardwired is the only really weird thing still not solved, and I suppose the CPU might just be too fast for the algorithm. I don't have another Amiga that comes even close to this speed, so it's hard for me to test. The closest I have is an A4000/030 at 25 MHz.

Re: kludge’s TF534 build

Posted: Mon Apr 01, 2019 2:32 pm
by alenppc
kludge wrote:
Mon Apr 01, 2019 7:18 am

I think the RAM chugs along. I'm expecting quite severe crashes if the RAM is unable to keep up, and I'm not seeing any of that yet. Frontier is _really_ smooth at least. Might order a couple of 40 MHz oscillators later for testing.
On some of my builds, I get lockups after a while at 50 mhz, but it really depends on individual chips. Also, I don't want Stephen to add more wait-states to the ram because that defies the purpose. But let's just say that the difference between the 12ns ram on the 530 and the 55ns on the 534 shows.

Also, because the CPU Fastrom option speeds the A500 so drammatically, I don't think it makes any sense to use EC cpus on this one, however on the 330 the speedup is minimal so it makes more sense there.

Re: kludge’s TF534 build

Posted: Mon Apr 01, 2019 2:53 pm
by kludge
alenppc wrote:
Mon Apr 01, 2019 2:32 pm
On some of my builds, I get lockups after a while at 50 mhz, but it really depends on individual chips. Also, I don't want Stephen to add more wait-states to the ram because that defies the purpose. But let's just say that the difference between the 12ns ram on the 530 and the 55ns on the 534 shows.

Also, because the CPU Fastrom option speeds the A500 so drammatically, I don't think it makes any sense to use EC cpus on this one, however on the 330 the speedup is minimal so it makes more sense there.
I might just be used to my stock Amigas, but I think the EC is plenty fast right now. It's a bummer that I didn't realise the lack of MMC earlier as I've ordered some more CPUs from the same seller as soon as I saw this one working. I might sell a few 534's and the buyers can choose if they want the EC030 or if they want to use a CPU of their own. Does everything go faster with CPU FASTROM or is it mainly the IDE?

I could of course build TF330's with the EC CPU's, but then I'd have to buy a CD32 :D