That is a good point. If APM only meant to refer to the 460 then why the x?

Which brings us full circle to your original argument and one basic question - why is an obviously dated graphic being included in new presentations?

And curiously enough, none of the mambas are vipers. Viper are generally slower and often less lethal. If they were attempting to offer some continuity with the name its only marginal.

Frankly Andreas, I've never wholly trusted this company.

> That is a good point. If APM only meant to refer to the 460 then why the x?

I just gave it another thought and came to the conclusion that if we take the mentions of "40X", "44X" and "46X" not as relating to the CPU core but to the denominations of the chips it actually makes sense. After all, we know that Applied Micro have changed their naming scheme for new chips. PPC40x changed to APM80xxx, PPC46x changed to APM82xxx and APM86xxx/PacketPro/Mamba respectively, and PPC44x chips are not even listed any more on their regular website. So if we read the graphic this way then "TBA FY2010" (the leftmost one, that is) can actually refer to the PPC465 core based APM86xxx/PacketPro/Mamba chips.

Quote:

That is a particularly lucid analysis. Since Gemini and Viper are naming scheme rather than processor designations that would clear up the ambiguity of that page.

[ Edited by Jim on 2011/2/13 19:06 ]

> Since Gemini and Viper are naming scheme rather than processor designations
> that would clear up the ambiguity of that page.

Still the question is what those other two "TBA FY2010" products at 40 nm are. After all, it's 2011 by now and I'm not aware of anything else than APM86x90/Mamba meeting this spec.

Quote:

So what do you think they might do?
What are those two unaccounted for products likely to feature?
So far their products have only advanced a short distance from their licensed IBM origins.
Do they have the engineers needed to move forward with more advanced technologies? They're using TMSC for fabrication, so they don't have to worry about process and foundry issues.
But can they move to 64 bit processors, multi-cored processors(>2), processor that run >2Ghz?

By the time Mamba is introduced, the 5010 and 5020 will be available.

> So what do you think they might do?

Probably announce those two "TBA FY2010" products in 2011, of which there's still seven eights left. If they do, they'll only be one year behind their own roadmap which they reissued in ...2011 ;-)

Btw, I found some recent statements from Applied Micro principals:

------------------------------
Paramesh Gopi - Applied Micro Circuits Corporation - President and CEO
[...] Year to date, Applied Micro has introduced 7 new 40-nanometer TSMC products, of which 4 are processors.
[...]
Robert Gargus - Applied Micro Circuits Corporation - SVP and CFO
[...] we had introduced 7 new products this year through the first 9 months. Four of those were in the Processor family, and all of them were 40-nanometer. And all of them were in TSMC.
------------------------------
http://www.sec.gov/Archives/edgar/data/711065/000119312511021986/dex992.htm (transcript of Applied Micro's third quarter 2011 earnings conference call in late January 2011)

Still, I can only count two: APM86190 and APM86290.

> What are those two unaccounted for products likely to feature?

For that to know I'd have to be a prophet ;-) But let's assume that as usual the height level indicates performance. That would mean their next 40 nm processor should be really low end, undercutting PPC440 performance level, and the one shortly after should be slightly faster than Mamba.
Furthermore, PacketPro was originally announced to sport USB 3.0 as well as 10GbE. So far, Mamba only offers USB 2.0 and 1GbE.

https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7001&forum=3&post_id=80171#80171

> So far their products have only advanced a short
> distance from their licensed IBM origins.

They advanced IBM's PPC440 core by adding an FPU and later L2 cache, yes. With IBM's PPC464, PPC464FP and PPC465 cores I'm not aware of any advancements done by Applied Micro at all.

> Do they have the engineers needed to move forward with more
> advanced technologies? [...] can they move to 64 bit processors,
> multi-cored processors(>2), processor that run >2Ghz?

I really don't know. But as I said, their roadmap from September 2010 showed 64-bit, multicore (which wouldn't necessarily have to mean "more than 2") and 2.5 GHz. But there was no timeframe attached.

> By the time Mamba is introduced, the 5010 and 5020 will be available.

According to Applied Micro, production quantities of Mamba are expected in Q2/2011. You told me some weeks ago that Freescale anticipated Q2/2011 for P5010 production as well (which was originally set at H2/2011). So let's see who'll be first :-)

Update:

>> So what do you think they might do?

> Probably announce those two "TBA FY2010" products in 2011

I found something that may be considered a clue:

"Many of our customers worldwide currently use Abatron’s probes with PowerPC 405, 440, 460, and 836xx processor families"
http://www.abatron.ch/news/support-for-appliedmicro-s-apm86290.html

(APM)836xx? What's that? I couldn't find anything about it. And how would it relate to APM83290/Gemini?

Quote:

Odd, either they transposed the 3 and 6 or that designation could be interpreted as a Titan successor.
What do you think it is?
I sent Abatron an inquiry directly, but I have my doubts that they'll clarify this.

> either they transposed the 3 and 6

If I understand right this would mean it should have been '863xx', wouldn't it? But that would be odd as well as the sentence mentions the "APM86xxx family" separately.

> or that designation could be interpreted as a Titan successor.

Or rather a Gemini successor with either a Titan core or a successor to that core.

> What do you think it is?

It could be a typo of some sort. The other two Applied Micro processor families that are supported by Abatron but are missing in that press release's list are 821xx and 832xx/Gemini. So "836xx" could mean one of those. But I don't know really.

> I sent Abatron an inquiry directly

I'm curious :-)

Update:

> It could be a typo of some sort. The other two Applied Micro
> processor families that are supported by Abatron but are missing
> in that press release's list are 821xx and 832xx/Gemini. So
> "836xx" could mean one of those.

Seems I was right. They corrected "836xx" to "83xxx". And I see now that they made the same mistake 10 months ago already, obviously going unnoticed so far:

http://www.abatron.ch/news/support-for-appliedmicro-s-apm821xx.html

Exactly. Thanks to your attention to detail and my e-mail - they changed it.

Abatron:
Hi

This is a error, it should read 83xxx. We will fix this.

Thanks,
Abatron Support

Of course this means their product supports Titan, which in itself is interesting.

[ Edited by Jim on 2011/2/24 17:23 ]

> this means their product supports Titan, which in
> itself is interesting.

Yes, this is a known fact. I linked to their January 2010's press release in December 2010 here in this very thread:

https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7001&forum=3&post_id=80181#80181

Furthermore, I documented Abatron's support for Titan/Gemini in another thread:

https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=6268&forum=11&post_id=79905#79905

It's a pity that they had to drop that project. It seems more capable than Mamba.
So many companies announced support for Titan that there would appear to have been a large number of evaluation systems in the hands of developers.
At 1.5Ghz, the Mamba doesn't appear to be that competitive with Freescale's P20XX.

> It seems more capable than Mamba.

I'm not sure if Mamba is really less capable than Gemini. For instance, Mamba has a DDR3 memory controller whereas Gemini only has a DDR2 memory controller. Is there even one single technical spec where Gemini outdoes Mamba?

> At 1.5Ghz, the Mamba doesn't appear to be that
> competitive with Freescale's P20XX.

I think that depends on which QorIQ P20xx you compare Mamba to specifically. Mamba is clearly superior to e500v2 based P2010/2020, but e500mc based P2040 could be about on par with Mamba (not taking into account the difference in CPU core amounts, that is).

Quote:

I'd say your comparison of the 2040 to Mamba is fair. But what we really ought to be comparing are the 5500 cored products.
Unless APM releases something similar to the 64bit products you've mentioned before, then they don't have anything close to that line.

> I'd say your comparison of the 2040 to Mamba is fair.

And my comparison of P2010/2020 to Mamba isn't? ;-)

> what we really ought to be comparing are the
> 5500 cored products.

You and me know very well that regarding capability and performance Mamba stands no chance compared to QorIQ P5. So that's a no-brainer, really.

> Unless APM releases something similar to the 64bit
> products you've mentioned before, then they don't have
> anything close to that line.

Absolutely.

> when I discussed PPC based systems with the owner of Varisys
> (the designers of the X1000 motherboard) I got this response:
> "In terms of PPC we are thinking QorIQ here for higher end designs."

Apparently, they've started to follow suit:

"XMC/PMC [...] VM400 (XMC) QorIQ Pxxxx"
"COM Express [...] VPX1 QorIQ P1022"
"PC104 [...] MC104PRQ QorIQ P2022"
http://www.varisys.co.uk/products.html

It's a pity we don't have the resources need to play this game.
Varisys could probably design a killer P5010 board for our market.

> Varisys could probably design a killer P5010 board for our market.

That's kind of what A-Eon's plan seems to be:

"Also check out the new multi-core CPUs being developed by Freescale, in particular the 64-bit dual core, 2.2 GHz P5020 and the 8-core, 1.5GHz P4080."
http://www.a-eon.com/news.html (December 22nd 2010)

"Ben Hermans encourages people to look at Freescale P4 P5 architectures for future of AmigaOne range."
http://www.amigakit.com/catalog/amiwest.php

"Ben Hermans also referred to the P4 and P5 processors from Freescale for future AmigaOne computer."
http://obligement.free.fr/articles_traduction/amiwest2010_en.php

So if A-Eon are pleased with Varisys's performance regarding development of the Nemo board I don't see a reason why they wouldn't want to expand their business relation to an X1000's successor (if there really will be any, that is), based on QorIQ.

Interesting! At least we have an idea where they're headed after the PA6T.

"Also check out the new multi-core CPUs being developed by Freescale, in particular the 64-bit dual core, 2.2 GHz P5020 and the 8-core, 1.5GHz P4080"

Maybe they ought to focus on getting the first product out the door before planning on its successor.

[ Edited by Jim on 2011/3/4 19:25 ]

> Maybe they ought to focus on getting the first product out the door
> before planning on its successor.

I guess that was just an answer to all the many people who counter A-Eon's plans by claiming that there's no suitable Power Architecture CPU after (and probably improving over) the PA6T, rendering an ISA change to x86 or ARM inevitable anyway, so that it makes no sense to go for the PA6T in the first place. I doubt there's been any real planning going on beyond X1000 yet.

Addendum:

> In real life performance (i.e. not Dhrystone) I expect the PA6T to
> have significant higher performance/wattage figure than the PPC970
> by providing only slightly less performance/clock.

On amigaworld.net someone linked an interesting PDF presentation containing comparisons between PA6T, 2xPPC970FX and MPC8641D:

http://amigaworld.net/modules/news/article.php?storyid=5844&start=20#78629

Addendum:

>> I found an interesting PDF presentation file dating August 2010 over there.
>> The most interesting pages are #16, which shows a product roadmap mentioning
>> "Future Viper SoCs", and #18, which reads "Viper Technology -> Breakthrough
>> Performance at Disruptive Cost Basis". To my surprise, the roadmap on page 16
>> still contains the Titan core based "Gemini" (APM83290), which had been
>> semi-officially declared dead one month before. Strange.

> Still all the same in this most recent presentation (February 2011).

New presentation (March 2011), but content same old, same old.

Update:

Quote:

Sam460ex at 1.15 GHz (production model):

Code:
Source: http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=32815&forum=14#605621

So that means a drop with the production model in all categories, specifically:

L1 cache read: -4.5%
L1 cache write: -5.7%
L2 cache read: -2.5 %
L2 cache write: -11.3%
RAM read: -8.8%
RAM write: -11.5% ('Tricky': -31%)
Video bus read: -8.3%
Video bus write: -40%

Quote:

Further RAGEMEM benchmark results of production model:

Code:
Source: http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=32815&forum=14#606210