A1222 For real??

[Iggy]

I think the issue is overblown. FPU's aren't useful for most regular computing, and even then, emulating instructions shouldn't be a big deal if done right.

I guess we'll find out once the NDA's are lifted, since no one is allowed to talk about AmigaOS on the A1222 at the moment. I don't know why they have NDA's for such a niche platform...I guess it makes the 4 people that are beta testing it feel important, when they can keep secrets from the 7 people that are planning on buying it :lol:

Unless a solution is implemented in the OS, software packages using floating point instructions will either have to be recompiled or they will fail to work.
Emulating the instruction with code based on the alternate fpu is not that big a deal, but trapping and translation IS.
At least initially, I don't anticipate that this will be a particularly efficient process, so we are likely to see specific binaries being compiled for the A1222.

And this is not THAT big a deal, but it IS a PITA.

[amiadudeorwat]

Unless a solution is implemented in the OS, software packages using floating point instructions will either have to be recompiled or they will fail to work.
Emulating the instruction with code based on the alternate fpu is not that big a deal, but trapping and translation IS.
At least initially, I don't anticipate that this will be a particularly efficient process, so we are likely to see specific binaries being compiled for the A1222.

And this is not THAT big a deal, but it IS a PITA.

This was more what I was wondering about.  If software still runs but it's just theoretically slower than it otherwise would be then that's ok-assuming it's not horribly slow.  But if everything needed recompiling then that's a big problem and I might not be interested.  I wasn't looking for benchmarks necessarily.

[Iggy]

This was more what I was wondering about.  If software still runs but it's just theoretically slower than it otherwise would be then that's ok-assuming it's not horribly slow.  But if everything needed recompiling then that's a big problem and I might not be interested.  I wasn't looking for benchmarks necessarily.

No one knows what the impact of the work around will be, so once OS4 is released for the A1222, THEN we will finally have a answer to this well worn topic.

[Vlabguy1]

Yes, it is very real! It has been shown at several shows and online!

BTW, nicholas "reckons" wrong for me and several others.

Wow, very cool.

[Vlabguy1]

I saw it at Amiwest back in 2016. Here's a photo I took

Well I figured the board was real..but I was more curious if the A1222 "folding" machine was real.. I think that is pretty cool, imo.  

R.

[amiadudeorwat]

What is the reason for going with the P1022 which has the E500v2 core in the first place, just price?  Is it that big of a deal to switch to one of the other QorIQ chips with the E500mc core or just too expensive which would negate the low price target of this board?

[dooz]

What is the reason for going with the P1022 which has the E500v2 core in the first place, just price?  Is it that big of a deal to switch to one of the other QorIQ chips with the E500mc core or just too expensive which would negate the low price target of this board?

If you want e500mc core than you must go to quad core P3041 or P2040. There are no dual core e500mc CPUs. Another "advantage" of P1022 is that it has integrated sound card into the SoC. Otherwise you have to add PCIe slots on the board or integrate sound chip on the motherboard which will result with more complexity and development cost of the board. If you add PCIe slots then you cannot have mini-ITX format of the motherboard any more.

All above can increase cost of the board and then it will not be "low cost" any more. There are for sure other factors that will influence the cost that we do not know about. If I remember correctly the mentioned price for motherboard was 400 EUR.

Only the price change for CPU from P1022 to P3041 will cost more than 100 EUR more for the motherboard itself without anything else.

P1022 = 1000 @ US$54.29 each

P3041 = 1000 @ US$181.70 each

And all of this because of problem that we still do not know if will affect performance in a real world applications. In fact P1022 have double precision FPU integrated. In this kind of "low cost" product everything counts! Maybe next mid level cost product will have more powerful CPU.

Unfortunately "T" series of CPUs was not available for delivery at the time when design was started.

-Dooz

[amiadudeorwat]

If you want e500mc core than you must go to quad core P3041 or P2040. There are no dual core e500mc CPUs. Another "advantage" of P1022 is that it has integrated sound card into the SoC. Otherwise you have to add PCIe slots on the board or integrate sound chip on the motherboard which will result with more complexity and development cost of the board. If you add PCIe slots then you cannot have mini-ITX format of the motherboard any more.

All above can increase cost of the board and then it will not be "low cost" any more. There are for sure other factors that will influence the cost that we do not know about. If I remember correctly the mentioned price for motherboard was 400 EUR.

Only the price change for CPU from P1022 to P3041 will cost more than 100 EUR more for the motherboard itself without anything else.

P1022 = 1000 @ US$54.29 each

P3041 = 1000 @ US$181.70 each

And all of this because of problem that we still do not know if will affect performance in a real world applications. In fact P1022 have double precision FPU integrated. In this kind of "low cost" product everything counts! Maybe next mid level cost product will have more powerful CPU.

Unfortunately "T" series of CPUs was not available for delivery at the time when design was started.

-Dooz

Thanks that does explain things a bit better, I didn't know about the integrated audio codec.  The P2040 at 1.2GHz is about double the price around $125 plus they would have to do something about audio, pushing up the final price up at least another 100.

We'll see how the performance and compatibility are when the NDA is lifted.

[nicholas]

So will a Tabor outperform a 1.67GHz PowerBook G4 that costs a quarter of the price?

[Iggy]

So will a Tabor outperform a 1.67GHz PowerBook G4 that costs a quarter of the price?

Don't talk sense to them, Nik.
After all, they're busy making foolish comments about e500mc cored products while ignoring the fact that there are cheaper dual and quad e5500 core cpus that could be used.

Again, Aeon has a LOT of processors that they've already purchased that they HAVE to put on something.

Someone has to pay for the decision to invest in the development of Tabor and the purchase of those processors. Might as well be OS4 users (because the Linux users appear to want to build their own 64 bit PPC laptop). :hammer:

[SACC-guy]

So will a Tabor outperform a 1.67GHz PowerBook G4 that costs a quarter of the price?

If you can find one that still works (as laptops, they got lots of abuse) and the batteries aren't dead. Not to mention the caps are ready to leak!

buy new, and miss all that 20year old stuff!

[dooz]

Thanks that does explain things a bit better, I didn't know about the integrated audio codec.  The P2040 at 1.2GHz is about double the price around $125 plus they would have to do something about audio, pushing up the final price up at least another 100.

We'll see how the performance and compatibility are when the NDA is lifted.

I investigated how much will *really* cost to buy those CPUs in high volume:

P1022 @ 1.2 GHz ($88.30)
https://www.arrow.com/en/products/p1022nse2mhb/nxp-semiconductors

P2041 @ 1.5 GHz ($214.06)
https://www.arrow.com/en/products/p2041nsn7pnc/nxp-semiconductors

P2041 @ 1.2 GHz ($182.22)
https://www.arrow.com/en/products/p2041nsn7mmc/nxp-semiconductors

P3041 @ 1.5 GHz ($256.03)
https://www.arrow.com/en/products/p3041nse7pnc/nxp-semiconductors

Yes, obviously it is not possible to have $182 CPU in 400 EUR A1222. Maybe a little bit cheaper if you order 1000 units but those are approx. prices. As a user I understand that some people would like something like P3041 @ 1.5 GHz but that is not possible for A1222 type of price range in a given moment when A1222 was designed.

-Dooz

[amiadudeorwat]

I investigated how much will *really* cost to buy those CPUs in high volume:

P1022 @ 1.2 GHz ($88.30)
https://www.arrow.com/en/products/p1022nse2mhb/nxp-semiconductors

P2041 @ 1.5 GHz ($214.06)
https://www.arrow.com/en/products/p2041nsn7pnc/nxp-semiconductors

P2041 @ 1.2 GHz ($182.22)
https://www.arrow.com/en/products/p2041nsn7mmc/nxp-semiconductors

P3041 @ 1.5 GHz ($256.03)
https://www.arrow.com/en/products/p3041nse7pnc/nxp-semiconductors

Yes, obviously it is not possible to have $182 CPU in 400 EUR A1222. Maybe a little bit cheaper if you order 1000 units but those are approx. prices. As a user I understand that some people would like something like P3041 @ 1.5 GHz but that is not possible for A1222 type of price range in a given moment when A1222 was designed.

-Dooz

I'm not disagreeing with you but I was referring to this P2040 here at $126 for 60.  It could be the wrong package type or any other thing wrong plus the lack of audio as mentioned.

https://www.digikey.com/product-detail/en/nxp-usa-inc/P2040NSE7HLC/P2040NSE7HLC-ND/5155398

[dooz]

I'm not disagreeing with you but I was referring to this P2040 here at $126 for 60.  It could be the wrong package type or any other thing wrong plus the lack of audio as mentioned.

https://www.digikey.com/product-detail/en/nxp-usa-inc/P2040NSE7HLC/P2040NSE7HLC-ND/5155398

Yes of course, there are many suppliers and price can vary. But you cannot use cheaper P2040 because it does not have any network (NIC) integrated into the SoC. Also it does not have L2 cache. You must use 2041 and it is more expensive then because you are getting then bunch of NIC integrated ;-) That is why some CPU QorIQ models are sometimes expensive.

Complete Pxxx series is problematic for desktop and only few CPU models can somehow fit in our desktop use.....and not without compromises.

Interesting is also that in fact P1022 is the only representative in complete "P" series that have sound and gfx integrated into the SoC. And all that for the low price tag and a very good performance which is comparable to G3 at 1.3 GHz raw CPU power. The compromise for P1022 was non-standard integrated FPU for which we will see if that is a problem in the future. If that is not a problem then this P1022 is perfect choice since "T" series was not ready at the time when A1222 was designed.

-Dooz

[dooz]

So will a Tabor outperform a 1.67GHz PowerBook G4 that costs a quarter of the price?

No - it will not. At least not in pure CPU crunching. I published dnetc OGR-NG benchmark on this link:

http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=41954&forum=33#799349

SAM460EX 1.1 GHz: 11,708,929 nodes/sec
A1222 1.2 GHz: 13,837,837 nodes/sec
Mac mini G4 1.5 GHz: 14,881,017 nodes/sec

Those are OGR-NG numbers that do not use Altivec present in G4. Disapointing is how despite bigger L1/L2 caches and 300 MHz faster clock G4 can score only about 1000 nodes/sec more against A1222. A1222 is a real winner here and SAM460 is not very far.

In real world applications A1222 is probably faster then the fastest Mac mini G4 at 1.5 GHz. Probably faster FSB and memory is doing some good to A1222 performance. Not to mention what GFX performance will show in a real world. Also P1022 CPU is dual core which is still not exploited on OS4/MorphOS.

Now I must find some FPU benchmark under Linux on A1222 ;-) Any suggestions from anyone what I can use?

-Dooz

[nicholas]

If you can find one that still works (as laptops, they got lots of abuse) and the batteries aren't dead. Not to mention the caps are ready to leak!

buy new, and miss all that 20year old stuff!

Thankfully Hyperion don't share your typically American consumerist view.