debian hardinfo benchmarks

[KimmoK]

My opinnion of AOS4 HWs state
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=35140&forum=33#661434
In short:
Amiga HW technology gap (against wintel):
1986: ahead of all others
1993: 2 years behind? (ok (only) for instrumentation, video and games)
A1 2002: 5,8 years behind (ok CPU performance, otherwise...)
SAM440 2008: 8,6 years behind
SAM460 2010: 6 years behind
A1x1000 2011: 2,6 years behind (CPU is only in netbook level, otherwise ok)

So, current top of the line AOS4 HW is priced similarly to cheapest expandable Mac (PowerMac), it has y2006 caliber CPU, otherwise it has modern specs & expandability (when compared to mainstream shops).
What is not modern is the SW support, and untill that is fixed, it's not point to compare the modernity...

******
btw... Anyone tried partition to partition copy speeds with observation on system responsiveness?
(I'm surpriced how much better SAM667Mhz is when compared to 3800+ AMD system with linux. I'm eager to see how x1000 handles the situation (initially I've read about 70MB/s copy speeds, 4x faster than my best x86 from y2008.).)

[KimmoK]

@fishy_fiz


>Im curious as to how you come up with these highly amusing results. 70MBps 4x best from 2008?

Tested on freshly formatted 7200RPM Seagate Barracuda SATA + ext4 + ubuntu 11.x on a 2Ghz dualcore Athlon: 17MB/sec with total loss of system responsiveness.

(my core2Duo laptop at work is even much much worse)

(I bet with some tricks things can be improved on both SAM & x86.)

> I was getting those sorts of speeds close to a decade ago using an ide drive. (edit: just noticed you write "your" best pc.

It's a partition to partition copy (with bigger amount of data than cache can hold).
Not any simple read/write speed test.

> Let me guess, it's either ide, or sata1 vs a sata2/3 in the x1000 you speak of, in which case the machine being used isnt far from irrelevant.

I think it's SATA2 on both machines.

>X1000 only 2.6 years below x86? Again, its closer to a decade. Even my oldest, most budget core2duo from about 5 years ago will demolish it.

As I described on the other thread (I had it linked). 2.6 years comes from various things.
x1000 vs mainstream that is commonly available:
- CPU is 10 years behind x86 Mhz.
- CPU FSB speed is only 2 years behind x86 mainstream
- expansionbus (PCIe) is only 1 year behind
- USB is on the level of mainstream (or maximum 1 year behind)
- x1000 can have the same GPUs (HW wise) that the mainstream uses
After I sum it up, it seems x1000 is/was about 2.6 years behind the mainstream, while the previous HW was even more behind of the mainstream of it's time.

Clear now?

UPDATE/
In my timeline, x1000 caliber technologies:
-CPUs went beyond 2Ghz in 2000-2001
-FSB/memory access of 4GB/s happened around 2007 (but DDR2 1067 is recommended even today)
-PCIe got v2.0 in 2006, but about all modern cards work well on PCIe x16 v1.0
-Serial bus: USB2.0 came in y 2000, but it's still in mainstream + compatible also with USB3.0 of y2008/2009. (+PCIe enables USB3 cards)
-GPU: r700 is from year 2008, but PCIex16 enables the use of latest GPUs & SP accelerations. Modern PA6T bandwidths will not stand in the way (much).

And that's on the higher end, as far as I see it.
/UPDATE

>I can only imagine youve compared a best case scenario for os4 h/w vs worst case you found for x86.

I compared the best available AOS4 HW vs mainstream (not the low end of mainstream, neither highest)

(Even today, mainstream computers are sold with 1...1.4 etc x86 Chips as it's low end, I did not compare to those)

>Dont get me wrong, if people are happy with their amiga hardware Im happy for them, but when a person writes this sort of,.. umm,.. let's say "biased", or "unbalanced" comments on a public forum they have to be prepared to be corrected.

I think I was not (too) biased. What others think?

(I have AmigaMulticoloured glassess, as other fans do as well, but anyway)

>Feel free to offer benchmarks and I'll give my results from both an 8 years old athlon64 and a 4.5ish year old budget core2duo.

All data should be available on the net already.

UPDATE:
To sum my opinnion up:
-   PA6T, G4, P1022 and e6500 based CPUs are good enough for 95% of user needs.
-   USB2.0 is today still good enough
-   PCIex16 is still good enough
-   etc…
-   There is no-one building simplified PPC motherboards, that’s why they remain too expensive to be sane outside (existing) AOS4 crowd
-   We need drivers + SW anyway, before we can be fully modern. Untill then, the rest is futile.

[KimmoK]

I want to see my droid cellphone bench marked against these.

Its obviously faster than the sam boards, but I'd be very interested in seeing how it matches against the mac and x1000.

This can give some info:
http://www.roylongbottom.org.uk/linpack%20results.htm
ARM 1.5Ghz -> 170MFlops
Atom 1.6Ghz ->   183MFlops when optimized, 89 MFlops without optimizations
etc...

Latest mobile ARM chips perform like notebook x86 chips. Roughly said.
(PA6T is in same gategory in some parts, x86 top of the range seem to get 2GFlops, 2.2Ghz 970(MP I assume) seems to have achieved 1,6GFlops, Power6 did 5GFlops(IIRC), e6500 should not be far behind the top of x86)

[KimmoK]

@fishy_fiz
>1). Youve compared an outdated x86 netbook cpu vs the strongest arm cpu

wtf?
I just took the GFlops values from the list of results.
(that outdated x86 chip is being sold with most of the below 500eur netbooks)

>2). X1000 is a desktop machine. X86 desktop cpus are orders of magnitude faster than thier piddly little netbook cousins

So?
Did I say otherwise?
I have said PA6T is netbook caliber chip (with better I/O).

>3). You GFlops number are off by a factor of about 50-70. An i7-2700k for example has a rating of roughly 130-150 GFlops.

See the list of results. It depends on if you look at the peak values, optimized results, etc.

@takemehomegrandma

Care to elaborate?
To me it seem you have not looked at e6500 specs?

(not far behind is same as a lot less than decade behind, higher performing than low end x86 dekstop today)
UPDATE/
FYI: some e6500 bits
- instructions per second per MHz is almost on the level of i7 core (6 vs .
- Flops ... not sure, but unless Altivec has dropped behind the mainstream, it should be compareable
- e6500 chips can have a lot of hyperthreading cores @ 1.8Ghz or only a few @ 2+Ghz
- it can have 2.3Ghz DDR3 on three 64bit memory bus (memory controller is built in)
- can support PCIex16 v3.0 (IIRC)
/UPDATE

[KimmoK]

In mainstream they sell this kind of low end for Home use (cheapest available on local shop):
-Smart TV 121eur
-Acer Aspire X1430/AMD Dual Core E-300/2 GB/320 380eur
-HP Compaq CQ2710EO Celeron G530T/4 GB/500 GB 400eur
etc
Apple HW:
- MacMini 2.3Ghz dualcore for 599eur
etc..


@fishy_fiz
>This whole comparitive things started with you suggesting X1000 is 2.6 years behind.

IMHO it still is. HW wise it's even up to date in some aspects.

> If a person wants to compare an x1000 vs an x86 pc its inaccurate to just compare it to x86 gear thats in a similar ballpark.

I compared to what is commonly available at the mainstream (more like high end than low end).

> Just because that's as far as the X1000 goes doesnt mean a person should disregard pretty much all modern x86 hardware (the original atoms are weaker than gear from a decade ago for example).

x86 is not AOS4 compatible. So, all this is very futile.
For grazy people theres also 300eur PC sold with 3000eur "sticker" from C USA, go & buy it, but it does not even give you AOS4 fun.

I have SAM for now. I do not plan to buy x1000 for it's current price, even though it would be ok HW. I planned to get 3Ghz i3 system for 300eur, but I think I donate it to AOS4 SW projects instead. etc... Money where my mouth is, etc.

>Comparing a $3000 modern pc to an X1000 is like comparing a commodore64 to an a4k+top of the line csppc Completely different league.

That's why I do not do it. Neither should you.
Comparing x1000 to non AOS4 HW is also pretty silly.
Comparing x1000 to AROS HW (when single core in use) might be a little bit more sane.

Other than that:
x1000 vs 1000eur x86: 1/10 CPU, 1:1 RAM speed, 1:1 expansions (about), 1:1 USB, SATA2 vs SATA3, etc.
x1000 vs 3000eur x86: 1/10 CPU, 1:2 RAM speed, expandability about one year behind, USB about one year behind,  SATA2 vs SATA3, etc. etc.
G4 Mac/AmigaOne vs 1000eur x86: 1/10 CPU, 1/10 RAm speed, PCI/AGP vs PCIe, USB1 or USB2, PATA100 vs SATA3, etc.  ((10 years+ 10 years+15 years+10 years +10 years ... almost 10 years behind in HW??))
my SAM440 vs 1000eur x86: 1/30 CPU, 1/10 RAM speed, PCI vs PCIe,....

>Not that it means much at the end of the day, its all down to what a person enjoys, but you keep trying to mould these things to your liking near as I can tell, and that really doesnt give an accurat overview.

I've done my study. I can be wrong, again. But I recommend people do some reality checking as well. x1000 is big leap forward in technology for AOS4 fans. Even though it's CPU is very far behind the mainstream (good thing is that you can do a lot with netbook/notebook caliber CPU, anyway).

For a AOS hobbyist I find it fun (and funny) that my SAM440 can do some things nicer than my y2008 x86 linux system, like the filecopy.

[KimmoK]

@fishy_fiz

>RAM speed is much less than 1:2. Quad channel ddr3@2133 is about 10x that of x1000 ram bandwidth. Worst case scenario is closer to 1:2, but its not a great comparison to compare bottom of the barell to premium.

>1000euro x86 pc = dual/tripple channel ddr3@1866 (about 5:1 vs x1000), 2x pci express 3.0 vs single pci express 2 (about 5:1),etc.

Ok.
But every x86 that I have seen sold below 1000eur has been with single or dual channel RAM, same for Apple.
And the RAM has been DDR3 1333 (that seems slower in practice than DDR2 1067)
And none of them have had more than one PCIex16 or multiple GPUs.
etc…
I need to look harder next time, perhaps…?

And at the same time …
One can put the same GPU in x1000, unlike ever before.
One can put USB3 adapter in, unlike ever before.
One can use SATA2 (+RAID) unlike ever before.
…

And again. We need SW to use multiple cores or GPUs. Again. x1000 is ok/very good, except the price.


>Heck, even my 4.5 year old core2duo (which cost me about $500 4.5 years ago) blitzes the x1000 (dual channel ddr2@1066 (3:2)

Why is that? x1000 has dual channel DDR2 1067.
Are you saying DDR2 runs slower when there is Amiga sticker somewhere?

>Also, dont forget than memory bandwidth doesnt equate to efficient memory usage. Athlon64's using ddr1 used to beat p4's using dual channel ddr2.

Like Athlon did on x86, PA6T is the pioneer of putting memory controller onboard -> best possible bandwidth.

>I absolutely agree that its a big step forward for OS4 hardware

Agreed!!!!!!!!!! ;-)

>(although no closer than the original a1's where vs pcs at the time)

I disagree.  To my math A1 GAP in 2002 vs x86 was 5,8 years.   (gaps: CPU 4y, FSB 6y, Expansions 10y, USB 4y, GPU 5y, when it is 10y, 2y, 1y,1y,0y for x1000)
And I will not go in more details with you. ;-)

>The whole point to my responses is that youve pretty heavily misrepresented where x1000 stands vs. x86.

To my math, x1000 was about 2.6 (..3) years behind the medium-high end in y2011, and it’s just my math. Everyone can do their own math.
The gap was growing until SAM460 and x1000 were released.
Today it’s possible to build more modern PPC systems (with latest peripherals) than some time ago + I do not think PPC catch up x86 ever again, though.

>Anyway, I dont really want to argue. If people are happy then Im happy for them. If youre happy convincing yourself that youre not shaping things so they fit what you want to be true then Im also happy for you

I think my glasses are not that red as you think + your glasses seems pretty black, btw, but anyway.
I should remember that when I try to be non-biased, I’m not.  Perhaps you should too.

Anyway... I was never good in math anyway. One better do his own.

[KimmoK]

@takemehomegrandma

>Uh, not going to try to pretend I understood what you just wrote there, you lost me somewhere prior to the comma.

PA6T was 10 years behind desktop chips when x1000FC came to market. e6500 based chips are less behind. Does it now compile?

If Freescale manages to deliver AMP chips as they have planned, PPC motherboard builders should have ok material for faster than before PPC motherboards (+that can accept all modern peripherals off the self).

[KimmoK]

Hey, I'm a big PPC fan, but a 1.8GHz processor (even with 8 cores) isn't going to be a threat to an X86.

No does not.

But the 1.8Ghz 12 core (24cores visible via hyperthreading) should have roughly similar performance than multicore x86 chips. (for heavily multithreaded + SMP + SIMD using tasks)

(it just must not take another 10 years before we have e6500 based A1 + SMP capable OS + good SW)


@fishy
>I must admit to being surprised at just how many obsolete x86 products are still in circulation, even though thier significantly more powerful and advanced replacements are also in the marketplace for similar prices.

Mainstream has become sloppy in generating enough bloatsoftware and that's why there is less need for upgrade?

In real life I see Linux going for the bloatware crown. Even Mint LXDE seems slowish on 2500+ dumbster HW (that in amiga CPU land would compete with PA6T).

But at work the corporate IT service is the king of all. It's amazing how slow and crashy they manage set these core2 machines.
(already spent 15+ minutes today rebooting this kludge twice and restarting everything, I wonder where they even find this IE8 for these machines... my SAM440 almost beat this in javascript tests .... and it seems every office apps relies in bugged IE bits and pieces ...)

UPDATE, just in case G5 results were not previously mentioned:
http://amigaworld.net/modules/newbb/viewtopic.php?mode=viewtopic&topic_id=35671&forum=34&start=40&viewmode=flat&order=0#664098

[KimmoK]

@Iggy
>So you're specifically thinking about the T4280.

There will be 1...12 core variants. And e6500 is said to cope with up to 2.5Ghz clock rate.

>I can't think of a PC application that could efficiently use that many cores (outside of the communications applications this chip was designed for).

I have used some.
Mainly gcc, and dvd:ripp. But also most of rendering apps can use a lot of cores.
(they can also use cheap heterogenous clusters)
 
>And no Amigoid OS supports SMP.

Before we have second multicore based motherboard out, AOS4.2 with SMP will be out.

>So the T4280 would be no faster then a PA6T.

Even in single core + HT disabled those e6500 chips should be slightly faster than PA6T.
To my understanding they should perform like G4 per Mhz. And when only one core is used e6500 based chips should  go up to 2.5Ghz. (+2.3Ghz DDR3 + PCIe3.0 etc...)

But anyway... SMP is mandatory for all Amiga flavours. Otherwise I see no future hope for those niches growing.
(and all are going SMP ... it will take year(s), though)

[KimmoK]

T4 chips are clocked up to 1.8Ghz.
But they seem to have downgraded e6500 info to 2Ghz.
http://www.freescale.com/files/32bit/doc/brochure/PWRARBYNDBITSQIG.pdf

The old info:
http://media.freescale.com/phoenix.zhtml?c=196520&p=irol-newsArticle&ID=1576370

[KimmoK]

@Iggy
>they haven't announced any T1, T2, or T3 products yet.

"These high-performance QorIQ products will span 2-24 virtual cores"

They seem to have dropped also original single core option.
But the plan is to get new AMP version out per every quarter.
Hope they manage.

>I'm not sure I'm interested in AMP processors outside of the T4280.
>That processor offers the most SerDes lanes (other processors in this family have less).
>This will allow it to support better expansion options (like 16X PCIe cards).

I think we have to wait and see to know how many serdes lines they have on other versions.
+ it takes years before our SW can utilize PCIe-x4 fully, not to mention v2.0 or v3,0 or x16.

(to me it seems dual core version should have at least 16serdes lines, like the B4860 has)