[Info-vax] OpenVMS I64 V8.1 "Evaluation Release"?

[Johnny Billquist]

On 2012-03-26 15.27, VAXman- @SendSpamHere.ORG wrote:
> In article<jkpl8q$m3p$1 at Iltempo.Update.UU.SE>, Johnny Billquist<bqt at softjar.se>  writes:
>> On 2012-03-23 18.45, Fritz Wuehler wrote:
>>> glen herrmannsfeldt<gah at ugcs.caltech.edu>   wrote:
>>>
>>>> Fritz Wuehler<fritz at spamexpire-201203.rodent.frell.theremailer.net>   wrote:
>>>>> Johnny Billquist<bqt at softjar.se>   wrote:
>>>>
>>>>>> No, the z/Arch, even in its latest implementation, does not give you 64
>>>>>> physical address pins out of the CPU, so no way you could hook up 2**64
>>>>>> bytes of physical memory.
>>>>
>>>>> Do you have a pinout handy? If so, I defer to you.
>>>>
>>>> As far as I know, they don't sell the chips separately from whole
>>>> machines, and don't publish data sheets. There might some
>>>> maintenance manuals, but likely not.
>>>
>>> If so we need to agree it's all conjecture and the architecture may support
>>> 64 bit physical addresses or it may not. If you don't have the pinout and
>>> the block diagram or some other documentation then nobody knows.
>>
>> I'm still ready to claim that no machine produced today can connect 2^64
>> bytes of ram to it. I just don't see it as even probable that anyone
>> would be putting out 64 address pins.
>>
>> And that is ignoring possible architecture limits, where the machine
>> very well cannot form 64 bit physical addresses. After all, it is not
>> really required that a machine can form a 64 bit physical address, even
>> if you can form a 64 bit virtual address.
>>
>> All claims for maximum physical memory on all machines I've seen so far
>> are way below any 2^64 address range, and for some I have managed to
>> find definitive information. The one I totally failed on was the z196.
>> I'll totally agree that it's only conjecture, but based on the
>> information I have managed to find, it looks like the max would be 64
>> GB. That is way, way from 2^64...
>>
>> It would be interesting to hear if anyone with deep knowledge about the
>> architecture could comment on what the theoretical limits are on how
>> large physical addresses can be formed.
>
> Why not?  If you used 1GB DIMMs, you'd only need 16,000,000,000 of them.

Because the hardware might not actually form a 64-bit physical address 
out of the MMU.
The "old" VAXen, for instance, only forms a 30 bit physical address.
The "new" VAXen could form either a old 30-bit physical address, or a 
new form 34-bit physical address. It's decided by which format the PTE, 
which can have either a 21-bit PFN or a 25-bit PFN.

So, until the NVAX, and the new PTE format, all VAXen were restricted to 
a 30 bit physical address, even though you had a 32-bit virtual address. 
(But I suspect you actually knows this.)

Or maybe you just asked the rethorical question of why not have 2^64 
bytes of memory because you thought 16 billion DIMMs in a computer would 
look ridiculous? :-)

By the way, even assuming a IB z196 single CPU can address the current 
specified maximum of 3 TB, that would only requires 42 address bits. 
That is still a long way from 64. And why it would be 3TB on a single 
CPU totally beats me, and I don't think that is a per-CPU limit.

	Johnny