Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

New MacPro predictions? Hopes? Fears?

It's out there Karim, it's just that dram chip making technology can't take very complex circuits (so no booting Linux from a dram normal cpu for instance). So, there are a number of bridging memory/processing technologies, but these are specialist, and apart from static memory, which consumes a lot of space compared to dram, they basically multiple packet, which aids in speed a little, as the lines don't have to be driven for off chip access. So, what you do get is simpler acceleration circuits in the modules, which is not much use in normal use (but would be in the server farm etc specialists markets, which I imagine are already tied into competing technologies fit a small venture to place a big enough order to warrant a dram makers interest, as they ate interested in extremely big runs). However, my proposal is about some of the simplest common purpose processor designs. We are talking about 1000 transistor plus cores, parasitic capacitance memory, and another handful of simplifying features, and a very progressive feature, if I can get it designed and working practically, it would change the way we do computing. However, I'm interested in magnetic computing, as I have skne wonderful deduhn propissks for that, and that last technique might not be practical on it. With all that stuff happening to me last year, everything has ground to a halt and piled up. I've had a number of ideas in recent months that dissapeared in the fight. One, just going through my head, that I was going to use for a dystopian script, is a nethord that could duplicate a living human. I still have the general idea but the details disapear. Still, not yje sort of thing to put in a script, ideas are dangerous, and someone could start working towards it. This is what I have to put up with, quickly establishing practical methodology and design for achieving even bazaar things, bit no way to to anything about it. Like a turbulent flow of a river that organises into practical realistic thinking that usually is proven right. The bubling jumbling bunch of lego that assembled into practical objects, like an evo's dream of blowing up a pile of building materials and having them self assemble into a house (I've come across Ken Ham a number of times).
 
Have a look at the link above on the AMD super computing project, that is similar to what you are talking about, and what I was proposing in the 1990's I think.
 
Karim D. Ghantous said:
Here's a thought: how about a MacBook Pro with fixed RAM - it's all on the CPU package. GPU, CPU, xGB RAM, all in the one package. Intel has been putting GPUs on the chip for years now. There's a limit to how powerful that can be, but it shames a lot of cheaper graphics cards. Having RAM onboard the CPU package would mean that you don't need a separate cache. RAM and cache become one.

Or, the opposite: put some processing power - just a bit - into the RAM SIMMs. This was explored 20 years ago, and I never heard much about it after that.
Click to expand...

I had an SGI O2 and OCTANE that was sorta doing the same tech.
 
Karim D. Ghantous said:
A photographer buys a base model Mac Pro, deciding against the technically faster iMac:

http://www.stevehuffphoto.com/2017/...d-mac-pro-as-my-new-desktop-computer-in-2017/
Click to expand...

Lol.

Eric Santiago said:
I had an SGI O2 and OCTANE that was sorta doing the same tech.
Click to expand...

Yeah. Remember something like that. GameCube used a smart memory technology with some processing technology, and I forget if Nintendo 64 had anything in its 3D pipeline.

Here's an idea, a ring of little cubes facing inwards for thermal cooling, the Mac Pro Mini, 1-2 inches high. Then Apple could feel safe that nobody could imagine fitting a bluray into the case. Now, all they have to do is get it down to around a millimeter thickness, then they can feel safe nobody can use a USB stick on it instead of cloud storage. :thumbup:
 
Nooo, don't give away my secrets...
 
Honestly, like many years ago I wanted to do smart watches. I saw a snap band and thought, yeah, a smart watch like that. Eventually Apple got around to the same idea, but haven't seen one. Even flexible oled it us an issue the bending. However, not so much I think, especially if you want to sell them for $20 and gave them last a year. Whoops, of yeah we were talking Apple, who might want to have them for $1000 and last for maybe 10 years, but be replaced every year, and maybe forgotten about after 5 years So, that's $1000 divide by say five secondhand ones, that's $200 profit a year, but you are spending maybe $5k per five years, and I want to only charge you $20-$100 every five years. You want Apple or me?
 
Anyway, let's have a look at what I was planning, many possible designs, but I'll stick to one hybrid, say an mechanical scanned projected display inside the watch band.

The issue is that you unfurl the thing, it is thin. What use is that. So I did measurements to fund out what hieght I had to make it to get a descent ratio, making it a thick band. But it was very thick then. Hence, Steve liked things just right, so I imagined he looked at thevsame thing and rejected it.

However, it is still useful.

Like my idea from the 80's or nineties. Of a printable game system for $10-20. I forget, maybe in a tear off role. Heaps more speed capacity than a 500mhz Pentium. The games for like $10 would be on cardboard like tear off roles. I think the idea was to have games on tear off roles with screen, processor and controls on the card.

Who do you want, Nintendo?

These all used doable technology.

One of the greatest tricks that could save Red millions a yeas, or $100 a camera, is that you only need as Mich speed as you need. Once you realise this, you can go for cheaper chip making techniques, even in house if a suitable stamping process is available. Instead of competing for foundry space fie the latest chip processes. But of course, they want to go for more and more pixels, so higher and higher chip requirements from foundries or fpga's, they are not doing a 4k camera and stopping there I wonder if there is some weird material ultra ultra low powered 500mhz chip making process out there, that would cancel the power consumption advantage out. Magnetic computing is one thing being looked at.
 
What is interesting about Intel's CPU offerings is that they bottleneck the ability of the CPUs <-> Main memory bandwidth.

This is Intels chosen tool to limit how fast the CPUs can process (for data driven tasks) - and make us pay handsomely 'datacentre' prices to get our mits on their performant chips.

A number of the benchmarks of the latest AMD x86 chips outperformed Intel by 2x, simply because AMD doubled the memory bandwidth. With AMD returning to the fold - new Intel chips will probably pop-out-of-thin-air with ... double the memory bandwidth (which would be great for us).

Thus far - Apple have (on their mobile platform) implemented several architectural improvements over ARMs stock designs (usually ahead of ARM themselves). The A10 'Fusion' chip actually works more effectively in BigLittle context switching than the stock ARM version (that ARM tuned for LINUX).

But - there is a microtecture advance + ISA (Instruction Set Architecture = underlying native instructions used by the CPU) that ARM will be announcing circa October '17 that may will give adopters ways to support acceleration of individual threads.

What does this mean
+) ARM are moving from the BIG.Little CPU clusters (et 2 super power CPUs and 4 very low power CPUs) to
+) CPUs that dynamically tune themselves - OR allow the OS / compiler accelerate thread by Boosting CPUs that critical threads are running on.
+) Mobile phone makers using stock ARM chips may dispense with 'low power cores' in preference to more 'Bigger' cores, and no 'Small' cores = no messy context switching + more SoC space for other goodies (like machine learning & GPU)

Apple moving to TSMC 10nm in Sep'17 could mean that their mobile chips core central frequency finally overtake my 3Ghz 10core MacPro.
And if Apple are an early adopter of the new ARM ISA - there is no 'sensible' CPU core limit within a cluster (previously there to coherently switch between BIG and Little cores).

I say IF - because the identity of the early adopter has not yet been published (it could be Cray, Fujitsu. Not convinced it would be Qualcomm, Cavium or the others in pining for high core counts)

AJ

Sources : (as requested)

1) EETimes : http://www.eetimes.com/document.asp?doc_id=1331489
-> TSMC 7+ Process runs ARM cortex A72 at > 4GHz
-> TSMC 7+ Process is 10% faster than 7nm
And TSMC 10nm is ~20% higher clockspees than TSMC 16FCC

2) NextPlatform : https://www.nextplatform.com/2017/03/21/new-arm-architecture-offers-dynamiq-response-compute/
-> ARM DynamicIQ ISA
 
Last edited:
Extra, after all this time the industry hasn't caught up to the design I started in 1985/86. All the time I see stuff, and think of the previous designs that solve it, knowing despots have given the wrong people fame and money. By despots, I mean technical domineering consumers that back the wrong races. It's alright dominating the market when you are doing the right thing, sort of thing. We'd probably be more than to the planets right now. If we had only just launched a plasma reactor drive ship and or Laser, we could be mapping the gravitational fields between star by now. Once the drive runs out of solar wind particles to propel off of it could convert over to laser. My idea was to effect the laser output to relay a high quality signal, and to communicate back we could use powerful earth based equipment. High bandwidth communications from the satellite. I don't know what the calculations would be, but imagine it would be within 40 years to Alpha Centauri (and that planet they just happened to miss up till recently around Proxima Centauri). And that's the slow way.

http://www.space.com/33834-discovery-of-planet-proxima-b.html

The reason Space x and Tesla are where they are is Elon Musk, they could he a lot further. Same with Apple and whatever else you complain about, it is the people in charge who often sets the pace, and his things are. Not saying they didn't do great things, but greater could he done. However, looking at.the PC space, it is miserable alright.
 
Antony Newman said:
What is interesting about Intel's CPU offerings is that they bottleneck the ability of the CPUs <-> Main memory bandwidth.

This is Intels chosen tool to limit how fast the CPUs can process (for data driven tasks) - and make us pay handsomely 'datacentre' prices to get our mits on their performant chips.

A number of the benchmarks of the latest AMD x86 chips outperformed Intel by 2x, simply because AMD doubled the memory bandwidth. With AMD returning to the fold - new Intel chips will probably pop-out-of-thin-air with ... double the memory bandwidth (which would be great for us).

Thus far - Apple have (on their mobile platform) implemented several architectural improvements over ARMs stock designs (usually ahead of ARM themselves). The A10 'Fusion' chip actually works more effectively in BigLittle context switching than the stock ARM version (that ARM tuned for LINUX).

But - there is a microtecture advance + ISA (Instruction Set Architecture = underlying native instructions used by the CPU) that ARM will be announcing circa October '17 that may will give adopters ways to support acceleration of individual threads.

What does this mean
+) ARM are moving from the BIG.Little CPU clusters (et 2 super power CPUs and 4 very low power CPUs) to
+) CPUs that dynamically tune themselves - OR allow the OS / compiler accelerate thread by Boosting CPUs that critical threads are running on.
+) Mobile phone makers using stock ARM chips may dispense with 'low power cores' in preference to more 'Bigger' cores, and no 'Small' cores = no messy context switching + more SoC space for other goodies (like machine learning & GPU)

Apple moving to TSMC 10nm in Sep'17 could mean that their mobile chips core central frequency finally overtake my 3Ghz 10core MacPro.
And if Apple are an early adopter of the new ARM ISA - there is no 'sensible' CPU core limit within a cluster (previously there to coherently switch between BIG and Little cores).

I say IF - because the identity of the early adopter has not yet been published (it could be Cray, Fujitsu. Not convinced it would be Qualcomm, Cavium or the others in pining for high core counts)

AJ
Click to expand...

Where is this, this Antony? I've read often less than top of the line Pentium. If you compare to i3, you get what you ask for, higher comparison of results.

Caching tends to greatly level out computational did difference an caused by bandwidth, but if you make caches or internal pipes to small or slow, or have really high bandwidth tasks, you get hit In the case of the i3, I imagine some of this is the case. So, multiple tasks at once hogging cache bandwidth and frequency with the 3D also gigging, with much mrmorynand dusk access, is probably going beba lit more woedul than on a top level chip.

But if this is possible, I trust you Antony to know it.

A new ISA is definitely what they need against RiscV, as they have the money and experience to analyse instruction execution paths from modifying ISA's. Looks like RiscV is already paying off, like DIY cinema camera project paid off.

Arm has just announced their new multiprocessor scheme. Whatever is the present one, careful use of it would likely maximise performance. They had this problem with the Sega Saturn, it seemed it was slow multiprocessing, but as they refines his they did it, it was much better. This is the issue with people, they barge in and just do it as they see fit and the results are down. The industry say, then try to refine the hardware to fit the behaviour, when the hardware and the behavour weren't that good to start with :( where at the first design stage or the behavior stage, they could have dine it better.
 
Antony. How do you think the 36 core AMD chip could do in real cinema workflow applications?
 
You must log in or register to reply here.
Back
Top