Forget about 20 nm tech, build your own 14nm design today. 8)

http://www.elektor.com/news/beat-intel-get-your-14-nm-process-development-kit.2108624.lynkx

_________________
The older and more respected a scientist is, the longer it takes to prove him wrong.

@Arko

OH FOR HEAVENS SAKES DECIDE WHAT YOU WANT. One minute it's not desktop unless it's above 3GHz, the next minute a Raspberry Pi is a desktop.

A laptop isn't a desktop just because you use it on a desk.

If the raspberry pi is a desktop, so is the minimig. I don't think you want to follow that path.

With the whole GFX vs GPU thing I just attempted preemptive action, because if I had suggested a device like the P1022 (which has internal graphics but no 3D acceleration) I assumed you would have said "but it doesn't have 3D acceleration".

@asymetrix

"please don't go the slow CPU route"

For everyone wishing a cheap board there is someone insisting that it must be faster. Funny how noone can accept that others might want a different product than themselves.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@asymetrix

Quote:

asymetrix wrote: @thread These are great ideas, I hope something gets made, even if people cant agree. Have a base system, then addon cards for extra features. Please dont go the slow cpu route. If you limit the CPU speed, you limit the creative freedom of a developer, it does not matter how fast and efficient your operating system is. Heavy duty applications and complex algorithms and games need GHZ speeds ! To give an idea what the PS4 expectations are for 2013+ Unreal 3 engine (minimum 2.5 teraflops requirement) http://www.youtube.com/watch?v=XgS67BwPfFY Quote: First Look at this Video http://www.youtube.com/watch?v=XgS67BwPfFY According to Epic that demo required 2.5 teraflops wich the ps3 is in capable of but we are already seeing Rapid increase in Teraflops making the next generation capable of such graphics (read Nvidia 2013 roadmap http://www.engadget.com/2010/09/21/nvidia-reveals-fermis-successor-kepler-at-28nm-in-2011-maxwel/)

Sounds like Epic's Samaritan demo has loaded UE3 with additional shader programs.

Epic has stated, Samaritan demo can run on a single GTX580 with optimization.

http://forums.anandtech.com/showpost.php?p=33107410&postcount=194

"The GeForce GTX 680 is about 10% faster than Radeon HD 7970, in Battlefield 3. In the same comparison, the GTX 680 is slower than HD 7970 at 3DMark 11. "

Driver optimization from AMD side can flip the situation.

Last edited by Hammer on 09-Mar-2012 at 08:55 PM.
Last edited by Hammer on 09-Mar-2012 at 08:36 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@thread

For our niche, we need affordable HW (but expandable) to make the community grow + SW.
On that end there is room for improvement, even after the netbook.

And with cheaper board there are more possibilities to get volume high enough.
I imagine a 1Ghz (dualcore) HW around 100...150eur would be pretty good and it could fit in a lot of places + it would be energy efficient (5W).
(onboard 2D GFX, enough RAM, SATA, PCIe x16(x4 speed),...)

Last edited by KimmoK on 09-Mar-2012 at 09:12 PM.
Last edited by KimmoK on 09-Mar-2012 at 09:06 PM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@KimmoK

Quote:

For our niche, we need affordable HW (but expandable) to make the community grow + SW. On that end there is room for improvement, even after the netbook. And with cheaper board there are more possibilities to get volume high enough. I imagine a 1Ghz (dualcore) HW around 100...150eur would be pretty good and it could fit in a lot of places + it would be energy efficient (5W). (onboard 2D GFX, enough RAM, SATA, PCIe x16(x4 speed),...)

Or truly innovative hardware that "blows your mind away" could work too.

Both roads are difficult, with different challenges.

@Smurfen

The T4240 would probably blow our mind IF Hyperion manages to ever enable 12 physical cores with hyperthreading (24 virtual cores). I really hope they're aiming for more than 2 way SMP

Many things we do in daily life benefits automatically. Any apps ported from Linux would already utilise it. And for anyone doing any compilation or 3D rendering, it sure would rock.

But since all CPU families are going SMP to add performance then the only thing I can think of other than SMP that could POSSIBLY blow our minds would be GPGPU. Now that would rock. Who needs a cell or even Altivec if you can offload HUNDREDS of small number crunchers to a chip you already have?

If there's ONE thing this community should focus on, it's how to get some sort of benefit from shaders in general computing.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@olegil

I'm pretty sure they plan to use that 24thread chip one day and tune AOS4.2 accordingly.

btw. what is the SPE unit capable of on the P1022 chip?
(perhaps like handicapped DSP/VMX?? similar to DSP instructions of PPC440?)

It seems that the blue camp has been pretty active in researching next PPC options:
http://www.morphzone.org/modules/newbb_plus/viewtopic.php?topic_id=6465&forum=11
http://www.powerdeveloper.org/forums/viewtopic.php?t=1739
http://www.morphzone.org/modules/newbb_plus/viewtopic.php?topic_id=6196&forum=11&start=20#62074

(it seems P1022 can deliver 5000MIPS, it would be nice upgrade from our 667Mhz low end)

Last edited by KimmoK on 11-Mar-2012 at 12:38 PM.
Last edited by KimmoK on 10-Mar-2012 at 12:50 PM.
Last edited by KimmoK on 10-Mar-2012 at 11:14 AM.
Last edited by KimmoK on 10-Mar-2012 at 11:11 AM.
Last edited by KimmoK on 10-Mar-2012 at 11:09 AM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@olegil

Quote:

OH FOR HEAVENS SAKES DECIDE WHAT YOU WANT. One minute it's not desktop unless it's above 3GHz, the next minute a Raspberry Pi is a desktop.

/sarcasm on
hmm, you must be new here :p
/sarcasm off

he's been like this since day 1.

_________________
See my blog and collection website! . https://www.blog.amigaguru.com

@olegil

Quote:

olegil wrote: @Arko OH FOR HEAVENS SAKES DECIDE WHAT YOU WANT. One minute it's not desktop unless it's above 3GHz, the next minute a Raspberry Pi is a desktop.

Learn to read:

A CPU is not a desktop, a Roadmap for a CPU is not a desktop and if you want to understand what I meant about the Rasberry PI read again what I had written about it.

_________________
AmigaONE. Haha. Just because you can put label on it does not make it Amiga.

I borrowed this comments from here (#27 & #28):
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=38873&forum=2&start=20&order=0

@KimmoK

Those threads are good resources and there are more threads about ppcs on those sites. We discussed these things quite a lot in recent years and discussion is still going on.

_________________
My programs: via.bckrs.de
MorphOS user since V0.4 (2001)

@olegil

Quote:

But since all CPU families are going SMP to add performance then the only thing I can think of other than SMP that could POSSIBLY blow our minds would be GPGPU. Now that would rock. Who needs a cell or even Altivec if you can offload HUNDREDS of small number crunchers to a chip you already have?

I agree, having the best CPU is not enough to be mindblowing in a broader sense.
If GPGPU could be integrated nicely, things could start to get interesting

@Smurfen

"If GPGPU could be integrated nicely"

Now there is a development HW for utilizing GPGPUs, the x1000 with it's insanely advanced I/O & bandwidth (PCIe x16).

btw. Is the x1000 PCIe v1 or v2 version?

Last edited by KimmoK on 11-Mar-2012 at 09:24 AM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@KimmoK

I honestly don't know. The full datasheet is under NDA. Farnell had the hardware specs freely downloadable, but that's not mentioned there.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@olegil

Quote:

olegil wrote: @KimmoK I honestly don't know. The full datasheet is under NDA. Farnell had the hardware specs freely downloadable, but that's not mentioned there.

Considering that PCI-SIG announced PCI-e 2.0 spec about same time than P.A Semi started to ship PA6T it's quite safe bet that it's v1.1. (Even Intel/AMD didn't ship v2.0 chipsets until 6 months later).

@olegil

Quote:

olegil wrote: ..... manages to ever enable 12 physical cores with hyperthreading (24 virtual cores)...... Many things we do in daily life benefits automatically. Any apps ported from Linux would already utilise it. And for anyone doing any compilation or 3D rendering, it sure would rock.

- apps ported from Linux

Why should those benefit from it ?? It really depends on wether a specific app is multithreaded or not.

- compilation

Unless you mean something completly different those tasks are more discbound, running it over in more cores would probraly even slow down things.

- 3D rendering

Well o.k. I give you that

There is a good reason why desktops/portable only have 2-4 core CPUs, and why these CPUs focus on getting the cores faster instead of just adding more (even to the point where they allow one core to be overclocked when the other(s) are idle).

There is also a reason why you'll only find multi-CPU/core systems with fast core in servers and workstations utitlized for easily threaded tasks.

And you won't find those slow-per-core 10++ core systems outside of routers and suchlike for a good reason too.....

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Rose

Olegil perhaps answered to the SPE question.
But most likely you are right about the v1.1 for PCIe.

@Arko

Large compilations with gcc get big benefits from multiple cores.
For example on my previous job we had compiling cluster. (killer feature at early 2000, but even the laptops around 2007 (Core2) started to catch up with our cluster (P3 & 100Mbit LAN based))

For cluster use, local cache/storage on the cluster node is essential, far simpler when you have several cores on single motherboard.

@SPE
P1022 reference manual:
"The SPE (which includes the embedded vector and scalar floating-point instructions) will not be implemented
in the next generation of QorIQ devices. Freescale Semiconductor strongly recommends that use of these
instructions be confined to libraries and device drivers. Customer software that uses these instructions at the
assembly level or that uses SPE or floating-point intrinsics will require rewriting for upward compatibility with
next generation QorIQ devices.
The e500v2 core implements SPE double-precision floating-point instructions.
Freescale Semiconductor offers a libcfsl_e500 library that uses SPE instructions. Freescale Semiconductor
will also provide future libraries to support next generation QorIQ devices.
Note that in the Power ISA, MSR[SPE] and ESR[SPE] are renamed to MSR[SPV] and ESR[SPV]."

Last edited by KimmoK on 11-Mar-2012 at 02:43 PM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@Kronos

Compilation disk bound? What does you world look like, really? Ever do any compiling at all? I sometimes even use DistCC just to get a few extra threads when I'm compiling.

A modern solid state device can read the whole Linux source code tree in less than a second. Within the same second it could easily store back every object file and the resulting elf. Good luck finding a single-threaded CPU that'll keep up with that.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@Arko

Obviously you are not reading OUR posts here at all. Me and KimmoK are discussing what to put on a board WE WANT TO MAKE OURSELVES. I dunno about KimmoK, but I for one have done it before. Got a stack of protos here to prove it.

The questions from us are therefore TO BE CONSIDERED AS MARKET RESEARCH.

We get that you think an non-existent board is useless. We feel the same way. What we don't understand is why you one minute talk about 3GHz and the next minute talk about Raspberry Pi. Maybe if you could give us some USEFUL input?

Question we should make a poll for:
If given a choice, what would you buy:
A: 1GHz single core, cropped flex-ATX/mini-ITX/something like that anyway based on the P1013, 160EUR before tax (where applicable) and OS4
B: 1GHz dual core, same board but with a P1022 for 180EUR before tax (where applicable) and OS4
C: Neither, I need more power
D: Neither, too low spec for that price
E: Neither, I need a laptop
F: Neither, I need more expandability

One would hope people then took 10 minutes to research the P1022/13 before answering.

Unfortunately Hyperion would require four arms and a leg to even consider supporting it, so until they change the terms for porting to a new target don't expect anyone who isn't rich, stupid or fanatical (Trevor would hopefully be in the first category but more likely he's in the third one and will end up bankrupting himself) to try and get OS4 on their hardware.

(If it was a case of guaranteeing payment for the investment it would be fine, but afaict it's a one time up-front payment of all development cost AND a steep price for license per board).

So the problem here is actually not whether or not there are chips or even boards. The problem here is Hyperion. We could make 200 boards and sell them for the price mentioned above, earning back any investments done during prototyping. But adding 100EUR on top of that per board just to have the OPTION of running OS4 on it and THEN adding 100EUR on top of that per board to ACTUALLY have OS4 running on it quickly becomes rather silly.

Back in the days Hyperion were pissed off at Bill McEwen for not letting them support more boards. Irony?


(Me too angry? Nah. Just a mildly justified monday morning rage )

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@olegil

B (P1022 info)
(and very small with expansion possibilities like PCIe x4 / PCIe external / PCIe mini / cPCIserial / etc...)

And I think we really should not think about other niche OS before prototype boots Linux.
Then I would go for another round collecting investors and buyers to see how big production run could be made. And how interested ACube, Varisys or bplan is to do the production vs some new partner. (I would keep AROS+MOS+AOS4 informed)
If the run is big enough I think there would be more interest amongs Amiga niche OSs.


"I dunno about KimmoK"
I'm absolute newbie in HW design (and I have not lead any project), but as interested as any engineer.
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=35273&forum=4&0


Then for future:
-busboard modules
-I/O modules
-NatamiGFX module
-A1200 trapdoor adapter
-anything

Last edited by KimmoK on 12-Mar-2012 at 09:50 AM.
Last edited by KimmoK on 12-Mar-2012 at 09:31 AM.
Last edited by KimmoK on 12-Mar-2012 at 09:30 AM.
Last edited by KimmoK on 12-Mar-2012 at 09:30 AM.
Last edited by KimmoK on 12-Mar-2012 at 09:26 AM.
Last edited by KimmoK on 12-Mar-2012 at 09:19 AM.
Last edited by KimmoK on 12-Mar-2012 at 09:16 AM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@KimmoK

Since it's an SoC, chances are you won't spend a lot of time before Linux boots on it. Since you only need to make the bootloader report a sensible amount of peripherals with the mapping you want, and then you're there.

I for one would be interested in doing this even without any hope of seeing OS4 on it.

But the point still stands. If you are only gonna sell 200 of these in the Amiga market, paying 20000 EUR to Hyperion to have the OS working and then paying 100EUR per board on top of that for the license isn't going to cut it. Not if it's a community project where we're not exactly becoming millionaires ourselves.

Have you tried getting hold of any SODIMM sockets for DDR3? I just tried and came up with N O T H I N G. Digi-key have 5 pages worth of hits and it's all "call us for quotations" and "minimum order quantity 4500" and stuff. What's up with that? All other connectors on my list are available from singles up.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.