Your support is needed and is appreciated as Amigaworld.net is primarily dependent upon the support of its users.
|
|
|
|
Poster | Thread | KimmoK
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 19-Aug-2014 18:59:59
| | [ #101 ] |
| |
|
Elite Member |
Joined: 14-Mar-2003 Posts: 5211
From: Ylikiiminki, Finland | | |
|
| @cdimauro
"TFLOP in double precision. Do you know of any PowerPC CPU that can give similar results?"
Not read into that. But they tell me that GPUs push TFLOPS faster than any CPU ? ( UPDATE: 216-GFLOPS is said to be the performance of a singe t4240 but only in marketing... )
In coremark single T4240 was on bar with more power consuming Xeons.
">legacy support is the size of FPU on intel core."
Not going to dig that up again. I would see it superb achievement if legacy support is not bigger than one FPU unit per core. And if the consume no power when running x64, then I should ignore it.
"Thousands? "
Yes. One pizzabox in dome has 3000 (virtual) cores, IIRC. So a rack of those has tens of thousands.
"Xeon Phi"
When you tell about that, is to me like from another time dimension (or from parallel reality). (and I tried to look what intel has, I did not see that)
I will look harder.
Last edited by KimmoK on 19-Aug-2014 at 07:58 PM. Last edited by KimmoK on 19-Aug-2014 at 07:04 PM. Last edited by KimmoK on 19-Aug-2014 at 07:01 PM.
_________________ - KimmoK // For freedom, for honor, for AMIGA // // Thing that I should find more time for: CC64 - 64bit Community Computer? |
| Status: Offline |
| | cdimauro
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 19-Aug-2014 19:28:17
| | [ #102 ] |
| |
|
Elite Member |
Joined: 29-Oct-2012 Posts: 4068
From: Germany | | |
|
| Quote:
KimmoK wrote: @cdimauro
"TFLOP in double precision. Do you know of any PowerPC CPU that can give similar results?"
Not read into that. But they tell me that GPUs push TFLOPS faster than any CPU ? |
On paper, but it's difficult to take full advantage of the raw processing power.
Xeon Phi makes it much easier, and has comparable performance. Quote:
In coremark single T4240 was on bar with more power consuming Xeons. |
I already expressed my opinion about such benchmark.
Have you looked at the performance on a concrete, real-world, scenario? I provided a link about FFT calculation, which is a WIDELY used algorithm at the basis of many other: have you took a look at it? What's your opinion? Quote:
">legacy support is the size of FPU on intel core."
Not going to dig that up again. |
If you express some statements about an argument, it's better that you dig-up to see if they are true or false. Better if you do it BEFORE reporting something public. Quote:
I would see it superb achievement if legacy support is not bigger than one FPU unit per core. |
I already stated what's the real relationship between the "legacy" x86 stuff and the FPU (x87): the latter has a MINOR role in the total computation, both in terms of transistors and power drawn. Quote:
And if the consume no power when running x64, then I should ignore it. |
First, x64 ha promoted SSE2 has the basis for the floating-point computation. The old x87 FPU is deprecated; in fact, you can verify it by disassembling applications (of course, more modern ones).
Second, I already posted a link about how modern x86 processors work. It's pretty explanatory about how much the (biggest) legacy stuff impact at runtime, running concrete code. Quote:
"Thousands? "
Yes. One pizzabox in dome has 3000 (virtual) cores, IIRC. So a rack of those has tens of thousands. |
OK, now it's clear. Quote:
"Xeon Phi"
When you tell about that, is to me like from another time dimension (or from parallel reality). (and I tried to look what intel has, I did not see that)
I will look harder. |
Do it. The many-core paradigm is the next frontier for massive computation, and Intel has already presented products based on it (the Xeon Phis). More and better has to come, with the next Knights Landing-based Xeon Phis that'll bring a lot more performance and benefits. |
| Status: Offline |
| | KimmoK
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 19-Aug-2014 19:43:42
| | [ #103 ] |
| |
|
Elite Member |
Joined: 14-Mar-2003 Posts: 5211
From: Ylikiiminki, Finland | | |
|
| @cdimauro
"Have you looked at the performance on a concrete, real-world, scenario?"
I have not seen t4240 in everyday use yet.
" you dig-up to see if they are true or false. Better if you do it BEFORE reporting something public."
I did. Just do not have the link at hand.
Started to read about xeon phi ... "binary incompatible" hmm... thought it's a x86/x64 chip ... better reserve more time for the reading... "...it is going to evolve into a stand alone platform. It's nearly a general purpose many core "CPU" already and Intel has the means to move it forward to become a complete high performance computing system. ..." Ok... now to sleep. http://en.wikipedia.org/wiki/Xeon_Phi (looks like a co-processor so far? ) Last edited by KimmoK on 19-Aug-2014 at 07:53 PM. Last edited by KimmoK on 19-Aug-2014 at 07:47 PM.
_________________ - KimmoK // For freedom, for honor, for AMIGA // // Thing that I should find more time for: CC64 - 64bit Community Computer? |
| Status: Offline |
| | cdimauro
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 20-Aug-2014 5:19:25
| | [ #104 ] |
| |
|
Elite Member |
Joined: 29-Oct-2012 Posts: 4068
From: Germany | | |
|
| Quote:
KimmoK wrote: @cdimauro
"Have you looked at the performance on a concrete, real-world, scenario?"
I have not seen t4240 in everyday use yet. |
Of course: it's a long time that you hope to have such system, but it's clearly out of the post-Amiga niche.
However I provided you a useful link about AVX and Altivec, and you (and many others) haven't yet express any opinion about... Quote:
" you dig-up to see if they are true or false. Better if you do it BEFORE reporting something public."
I did. Just do not have the link at hand. |
You can take a look at sites like chip architect, for example, which show the processor die split into the functional units.
Anyway, I hope that you find the link, because I'm really curious to see what was invented. Quote:
Started to read about xeon phi ... "binary incompatible" hmm... thought it's a x86/x64 chip ... better reserve more time for the reading... |
Please, find a better reading: Xeon Phi IS binary compatible with x86/x64, albeit it doesn't cover the FULL ISA. Knights Landing will cover this gap. Quote:
"...it is going to evolve into a stand alone platform. It's nearly a general purpose many core "CPU" already and Intel has the means to move it forward to become a complete high performance computing system. ..." Ok... now to sleep. http://en.wikipedia.org/wiki/Xeon_Phi (looks like a co-processor so far? ) |
Yes, right now they are available as PCI-Express cards. With Knights Landing they will also work as standalone CPU. |
| Status: Offline |
| | KimmoK
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 20-Aug-2014 6:36:34
| | [ #105 ] |
| |
|
Elite Member |
Joined: 14-Mar-2003 Posts: 5211
From: Ylikiiminki, Finland | | |
|
| @cdimauro
"However I provided you a useful link about AVX and Altivec, and you (and many others) haven't yet express any opinion about... "
Went through it fast, read more later. (I know Altivec is "old" and slower vs what intel has.)
UPDATE: in short, AVX can be 2.5...13x faster with same clock rate than previous altivec generation (when given enough bandwidth). (to my understanding altivec has improved only a little for e6500, the memory bandwidth has "exploded" for e6500 chips vs. previous freescale offerings)
I think I have not yet seen analysis of e6500 floating point performance vs intel. (if it would be totally obsolete vs intel, no company would have taken t4240 in use?) Last edited by KimmoK on 20-Aug-2014 at 06:46 AM.
_________________ - KimmoK // For freedom, for honor, for AMIGA // // Thing that I should find more time for: CC64 - 64bit Community Computer? |
| Status: Offline |
| | AmigaBlitter
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 20-Aug-2014 15:56:35
| | [ #106 ] |
| |
|
Elite Member |
Joined: 26-Sep-2005 Posts: 3514
From: Unknown | | |
|
| @cdimauro
Power 8 is a monster, but what about Power 9?
"In 2004 we had POWER5 followed by POWER5+. In 2007 we had POWER6, which led to POWER6+. In 2010 we had POWER7 and the most current, POWER7+. In 2014 we have POWER8, and there are already charts that show POWER9 is being planned for the future. IBM has consistently delivered on its roadmaps."
http://www.ibmsystemsmag.com/linuxonpower/Trends/IBM-Announcements/power8_first_look/
_________________ retired |
| Status: Offline |
| | KimmoK
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 20-Aug-2014 17:04:07
| | [ #107 ] |
| |
|
Elite Member |
Joined: 14-Mar-2003 Posts: 5211
From: Ylikiiminki, Finland | | |
|
| @AmigaBlitter
Also the first Xeon Phi's look like power consuming monsters.
I find it interesting that there suddenly is competition in supercomputer area, instead high IBM dominance.
power9... http://dancingdinosaur.wordpress.com/tag/power9/ and wiki says: "IBM has been designing the future POWER9 processor for quite a while according to William Starke, a systems architect for the POWER8 processor.[4] No other information was available in August 2013."
Last edited by KimmoK on 20-Aug-2014 at 05:08 PM.
_________________ - KimmoK // For freedom, for honor, for AMIGA // // Thing that I should find more time for: CC64 - 64bit Community Computer? |
| Status: Offline |
| | cdimauro
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 20-Aug-2014 19:26:06
| | [ #108 ] |
| |
|
Elite Member |
Joined: 29-Oct-2012 Posts: 4068
From: Germany | | |
|
| Quote:
KimmoK wrote: @cdimauro
"TFLOP in double precision. Do you know of any PowerPC CPU that can give similar results?"
Not read into that. But they tell me that GPUs push TFLOPS faster than any CPU ? ( UPDATE: 216-GFLOPS is said to be the performance of a singe t4240 but only in marketing... ) |
Yes, marketing. Anyway, it's 1/10 of the performance of a Xeon Phi.
Quote:
KimmoK wrote: @cdimauro
"However I provided you a useful link about AVX and Altivec, and you (and many others) haven't yet express any opinion about... "
Went through it fast, read more later. (I know Altivec is "old" and slower vs what intel has.)
UPDATE: in short, AVX can be 2.5...13x faster with same clock rate than previous altivec generation (when given enough bandwidth). (to my understanding altivec has improved only a little for e6500, the memory bandwidth has "exploded" for e6500 chips vs. previous freescale offerings) |
It's not only a question of memory bandwidth. ISA matters, and a lot. In fact, you can see that AVX is quite superior even to SSE on the same machine. Quote:
I think I have not yet seen analysis of e6500 floating point performance vs intel. |
Post them when you'll get. Also "integer" performance is important for stuff like emulators. Quote:
(if it would be totally obsolete vs intel, no company would have taken t4240 in use?) |
And no company would have taken ARMs in use, since they perform much slowly compared to Intel products, right?
There can be different factors which comes when choosing a processor to build a product. Performance is not always the key one. Take a look at the mobile market: Intel has interesting products with very good performance, and power consumption comparable to ARMs, but he was late and the market is dominated by ARM SoCs...
Quote:
KimmoK wrote: @AmigaBlitter
Also the first Xeon Phi's look like power consuming monsters. |
Taking the produced performance into account, it does very well. Quote:
I find it interesting that there suddenly is competition in supercomputer area, instead high IBM dominance. |
It's also interesting go see that IBM has provided x86-based supercomputer. Quote:
power9... http://dancingdinosaur.wordpress.com/tag/power9/ and wiki says: "IBM has been designing the future POWER9 processor for quite a while according to William Starke, a systems architect for the POWER8 processor.[4] No other information was available in August 2013." |
I don't understand why you're so interested in POWER processors, when it's well known that they'll not come to the post-Amiga PowerPC systems.
Is it only to signal that the PowerPCs aren't dead? It's too little...
It's better to focus on some SoC which has a chance to be integrated into an AmigaOne. |
| Status: Offline |
| | RonaldGadget
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 21-Aug-2014 17:14:59
| | [ #109 ] |
| |
|
Member |
Joined: 12-Jul-2014 Posts: 17
From: Zurich | | |
|
| @all
We will be distributing our first P5040 based single node uServers soon to the partners within the DOME User Platform program. The system will look like this:
The 'baseboard' contains all the power supplies, an mSata RootFS, GbE, 12 Volt supply input, and you're ready to run F20 on the 4 core, 2.2GHz machine.
Ronald |
| Status: Offline |
| | cdimauro
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 21-Aug-2014 18:57:40
| | [ #110 ] |
| |
|
Elite Member |
Joined: 29-Oct-2012 Posts: 4068
From: Germany | | |
|
| @RonaldGadget: is it available to the public? How much it will cost? |
| Status: Offline |
| | RonaldGadget
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 22-Aug-2014 6:02:53
| | [ #111 ] |
| |
|
Member |
Joined: 12-Jul-2014 Posts: 17
From: Zurich | | |
|
| @cdimauro
Quote:
cdimauro wrote: @RonaldGadget: is it available to the public? How much it will cost? |
As mentioned, it is only made available to partners within the DOME user platform: click here for info
Within the DOME user platform, which is for companies who wish to contribute to the DOME program, this is made available free of charge (limited availability). For the public, we are talking to companies who are interesting in licencing the technology to bring this to market. At this point, no pricing has been set.
BTW, next week I will power-on our own built T4240 based part, I spent the last 3 weeks debugging the power supply (on a board that does not have the T4240 soldered onto it). |
| Status: Offline |
| | KimmoK
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 22-Aug-2014 6:19:33
| | [ #112 ] |
| |
|
Elite Member |
Joined: 14-Mar-2003 Posts: 5211
From: Ylikiiminki, Finland | | |
|
| @RonaldGadget
"We will be distributing our first P5040 based single node uServers"
Nice, I think that's one of the first P5040 based boards "ready" outside Freescale.
Feel free to post e5500 core performance numbers @ hardinfo & blender thread.
_________________ - KimmoK // For freedom, for honor, for AMIGA // // Thing that I should find more time for: CC64 - 64bit Community Computer? |
| Status: Offline |
| | KimmoK
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 22-Aug-2014 6:22:04
| | [ #113 ] |
| |
|
Elite Member |
Joined: 14-Mar-2003 Posts: 5211
From: Ylikiiminki, Finland | | |
|
| @cdimauro
"It's also interesting go see that IBM has provided x86-based supercomputer."
It's also telling that IBM is part in t4240 & P50x0 based projects (like DOME).
"I don't understand why you're so interested in POWER processors, when it's well known that they'll not come to the post-Amiga PowerPC systems."
I'm not absolutely sure about that. I would consider it more propable than xeon phi ever being on my miggy.
(some POWER systems are not more expensive than x1000, as an example. but freescale chips remain more cost efficient so far)
reading more about xeon phi... (image of x86 legacy info) Why do they drag x86 legacy with the chip when it anyway is binary incompatible.... perhaps they see some use for it in future stand alone versions of the chip. Hmmm... getting back to legacy size of the chip... x86 is about 2% of the core+L2 size. What is the size of a single FPUnit then... Also now sotted the power usage of a chip... "Knights Landing's TDP will range from 160 to 215 W". (some info about available xeon phi cards) (more...)
Roughly ... getting 1...2tflops would require 300W with xeon phi and perhaps 600W with t4240? (then it would be nice to see some compareable integer performance numbers ... as xeon phi is based on atom (originally 2.4MIPS/MHZ) ... it should be weaker than i7 (up to 8MIPS/mhz), but I'm sure there's improvements ...) (trying to quess ... xeon phi might do 580MIPS/Mhz via 244 thread? while t4240 should do 72MIPS/Mhz?)
(I think we see xeon phi in coremarks when stand alone system appear. imagine it will be about 900k coremarks from 244 threads. So far: Xeon E5 2687W 3,4Ghz 32 threads 400k, Power7 3.55Ghz 64threads 366k, T4240 1,8Ghz 24threads 187k i7-3930k 12 threads 3,2Ghz 151k And so far t4240 is superior when considering power consumption... gives 50% of Power7 coremark result with 10% of the power (roughly)) (coremark & power usage) And yes CoreMark is only meant to show core's internal efficiency for multicore embedded use. So, it does not give good estimate for desktop or supercomputing use. (some supercomputing read) (ARM supercomputing study from y2012) UPDATE one wrong link removed. Last edited by KimmoK on 22-Aug-2014 at 09:05 AM. Last edited by KimmoK on 22-Aug-2014 at 08:05 AM. Last edited by KimmoK on 22-Aug-2014 at 08:03 AM. Last edited by KimmoK on 22-Aug-2014 at 08:01 AM. Last edited by KimmoK on 22-Aug-2014 at 07:48 AM. Last edited by KimmoK on 22-Aug-2014 at 07:33 AM. Last edited by KimmoK on 22-Aug-2014 at 07:31 AM. Last edited by KimmoK on 22-Aug-2014 at 07:07 AM. Last edited by KimmoK on 22-Aug-2014 at 07:01 AM. Last edited by KimmoK on 22-Aug-2014 at 06:56 AM. Last edited by KimmoK on 22-Aug-2014 at 06:54 AM. Last edited by KimmoK on 22-Aug-2014 at 06:49 AM. Last edited by KimmoK on 22-Aug-2014 at 06:46 AM. Last edited by KimmoK on 22-Aug-2014 at 06:39 AM. Last edited by KimmoK on 22-Aug-2014 at 06:34 AM.
_________________ - KimmoK // For freedom, for honor, for AMIGA // // Thing that I should find more time for: CC64 - 64bit Community Computer? |
| Status: Offline |
| | cdimauro
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 22-Aug-2014 19:44:24
| | [ #114 ] |
| |
|
Elite Member |
Joined: 29-Oct-2012 Posts: 4068
From: Germany | | |
|
| @RonaldGadget: OK, thanks.
Quote:
KimmoK wrote: @cdimauro
"It's also interesting go see that IBM has provided x86-based supercomputer."
It's also telling that IBM is part in t4240 & P50x0 based projects (like DOME). |
I was kidding. You know, IBM is the father of POWER & PowerPC, and he as build supercomputers solely using his own processors. Now, looking at him doing the same with other processor is... funny. At least. Quote:
"I don't understand why you're so interested in POWER processors, when it's well known that they'll not come to the post-Amiga PowerPC systems."
I'm not absolutely sure about that. I would consider it more propable than xeon phi ever being on my miggy.
(some POWER systems are not more expensive than x1000, as an example. but freescale chips remain more cost efficient so far) |
But be pragmatic: do you think that there's a remote possibility that a POWER processor can be used for a future AmigaOne? Quote:
reading more about xeon phi... (image of x86 legacy info) Why do they drag x86 legacy with the chip when it anyway is binary incompatible.... |
As I stated before Xeon Phi IS binary compatible with x86, but only a subset of the ISA is implemented. To be more clear, just A FEW instructions are missing, and there's no old SIMD units integrated (no MMX, SSE, not even AVX; albeit the encoding of the latter is used for some scalar instructions for the masks). Going further in detail, the few missing instructions are the ones that are mostly used by an o.s. and/or driver. Since the Xeon Phi was born as a coprocessor running on a PCI-Express card, they were not needed; so they were dropped. Quote:
perhaps they see some use for it in future stand alone versions of the chip. |
Exactly. That's why they will be integrated back on Knights Landing. Quote:
Hmmm... getting back to legacy size of the chip... x86 is about 2% of the core+L2 size. |
Which is very low, right? Quote:
What is the size of a single FPUnit then... |
Sorry, I cannot tell you. You have to take a look at sites like chip architect, where you can see the total die space occupied by all the legacy x86 stuff, and compare it taking only the x87 part. Quote:
I don't think that a T4240 can be comparable, because you have to integrate more of these to reach a similar performance. It means several boards / chipsets / memory connected by an high-speed network. All this draws power. So you cannot take the power consumption of a single T4240 and multiply it buy the number of these processors needed to reach the target performance.
And another thing. Xeon Phi reaches 2TFLOPS in single precision, but also a very big 1TFLOPS in double precision. You cannot take advantage of Altivec for double precision, so you've to resort to the much slower FPU for the T4240. So, it means that you have to put much more of these processors to reach similar performance.
Don't underestimate the double precision support: it's fundamental in many HPC areas. And that's why Xeon Phi is actually in pole position on the TOP500 (LINKPACK uses double precision). Quote:
(then it would be nice to see some compareable integer performance numbers ... as xeon phi is based on atom (originally 2.4MIPS/MHZ) ... |
An Atom core/thread can execute up to 2 instructions per cycle, so it's limited to 2MIPS/MHZ. Quote:
it should be weaker than i7 (up to 8MIPS/mhz), but I'm sure there's improvements ...) |
An i7 core/thread can execute up to 4 instructions per cycle, so it's limited to 4MIPS/MHZ.
Atom is weaker than the i7, of course. It was not build for extreme performance, but for a good performance/watt ration. The new Atom (BayTrail) has changed the architecture from in-order to out-of-order, so it perform MUCH better, but the i7 remains much better. Quote:
(trying to quess ... xeon phi might do 580MIPS/Mhz via 244 thread? while t4240 should do 72MIPS/Mhz?) |
Every Xeon Phi core (not thread) can execute 2 instructions per cycle. So, the theoretical performance is 122 MIPS/MHZ.
Anyway, you cannot use the MIPS a unit measure: they are totally different from an architecture to another, and x64 and PowerPC are VERY different. Quote:
(I think we see xeon phi in coremarks when stand alone system appear. imagine it will be about 900k coremarks from 244 threads. |
I don't think that Coremark will be able to take advantage of the Xeon Phi's big SIMD units and of so many cores. Quote:
So far: Xeon E5 2687W 3,4Ghz 32 threads 400k, Power7 3.55Ghz 64threads 366k, T4240 1,8Ghz 24threads 187k i7-3930k 12 threads 3,2Ghz 151k And so far t4240 is superior when considering power consumption... gives 50% of Power7 coremark result with 10% of the power (roughly)) (coremark & power usage) |
It's superior on ONE benchmark, which is pretty limited itself. See below. Quote:
And yes CoreMark is only meant to show core's internal efficiency for multicore embedded use. So, it does not give good estimate for desktop or supercomputing use. |
That's why it's better to take real-world applications for benchmark. Do you use Coremark to surf, emulate platforms, play games, watch videos, play music, compile projects, make query on databases, ecc. ecc.? I don't think so. Quote:
I'll read the PDFs when I've time.
One request to you: please, don't edit the existing comments. If you fix some typo, it's right. But if you add new sentences, or a whole new message like you did now, it's better to create a new comment. |
| Status: Offline |
| | virgolamobile
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 15-Oct-2014 17:34:23
| | [ #115 ] |
| |
|
Regular Member |
Joined: 23-Feb-2004 Posts: 192
From: Somewhere in Northern Italy | | |
|
| | Status: Offline |
| | KimmoK
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 16-Oct-2014 8:27:31
| | [ #116 ] |
| |
|
Elite Member |
Joined: 14-Mar-2003 Posts: 5211
From: Ylikiiminki, Finland | | |
|
| @cdimauro
"But be pragmatic: do you think that there's a remote possibility that a POWER processor can be used for a future AmigaOne?"
On a new Amiga niche specific product? No. Not with current POWER offerings.
Via rebranding some TYAN board, perhaps, as there seems to be people that want to buy the most powerfull HW for AOS4 at any cost.
But untill we have multicore support and better suitable POWER variants, they are pretty much out of reach. _________________ - KimmoK // For freedom, for honor, for AMIGA // // Thing that I should find more time for: CC64 - 64bit Community Computer? |
| Status: Offline |
| | olegil
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 16-Oct-2014 10:52:50
| | [ #117 ] |
| |
|
Elite Member |
Joined: 22-Aug-2003 Posts: 5895
From: Work | | |
|
| @KimmoK
POWER is useless without multicore support, so until we actually HAVE that, the point is moot. For any of the alternative offerings (I mean ARM/x86/etc).
The ONLY way to get significantly more CPU performance without multicore is to go for an Intel CPU, which ALSO supports multicore.
So personally I'm on the fence about the WHOLE debate until the confusion around OS4.2 and multicore support has been cleared up.
But the available long-term options for PowerPC/POWER architecture chips IS coming to a trickle, if not stopping entirely. This IS sad, because if EVERYONE uses ARM then where will the innovation come from? Innovation comes from competition, not monopoly. Interesting side-story that the ultimate goal of capitalists is a market monopoly, the very antithesis to capitalism in the first place Last edited by olegil on 16-Oct-2014 at 10:53 AM.
_________________ 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. |
| Status: Offline |
| | WolfToTheMoon
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 16-Oct-2014 12:35:34
| | [ #118 ] |
| |
|
Super Member |
Joined: 2-Sep-2010 Posts: 1405
From: CRO | | |
|
| @cdimauro
Quote:
The new Atom (BayTrail) has changed the architecture from in-order to out-of-order, so it perform MUCH better, but the i7 remains much better. |
IIRC, Bay Trail is partially OoO, SIMD/FPU are mostly done in order._________________
|
| Status: Offline |
| | AmigaBlitter
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 16-Oct-2014 17:06:29
| | [ #119 ] |
| |
|
Elite Member |
Joined: 26-Sep-2005 Posts: 3514
From: Unknown | | |
|
| | Status: Offline |
| | cdimauro
| |
Re: DOME Microserver with Freescale QorIQ P5040 and T4240 Posted on 17-Oct-2014 20:41:05
| | [ #120 ] |
| |
|
Elite Member |
Joined: 29-Oct-2012 Posts: 4068
From: Germany | | |
|
| @KimmoK
Quote:
KimmoK wrote: @cdimauro
"But be pragmatic: do you think that there's a remote possibility that a POWER processor can be used for a future AmigaOne?"
On a new Amiga niche specific product? No. Not with current POWER offerings.
Via rebranding some TYAN board, perhaps, as there seems to be people that want to buy the most powerfull HW for AOS4 at any cost.
But untill we have multicore support and better suitable POWER variants, they are pretty much out of reach. |
And a port is needed, to enable OS4 running on such new system. So, basically there's no future.
@olegil
Quote:
olegil wrote:
So personally I'm on the fence about the WHOLE debate until the confusion around OS4.2 and multicore support has been cleared up. |
It was already cleared: there'll be no SMP support. So it'll be AMP or a variant of it. Quote:
But the available long-term options for PowerPC/POWER architecture chips IS coming to a trickle, if not stopping entirely. This IS sad, because if EVERYONE uses ARM then where will the innovation come from? Innovation comes from competition, not monopoly. |
What innovation(s) came with the PowerPCs? I mean: compare them with the already existing CPUs, and let me know if there were significant innovations that this processor family brought to the market.
@WolfToTheMoon
Quote:
WolfToTheMoon wrote: @cdimauro
Quote:
The new Atom (BayTrail) has changed the architecture from in-order to out-of-order, so it perform MUCH better, but the i7 remains much better. |
IIRC, Bay Trail is partially OoO, SIMD/FPU are mostly done in order. |
I don't think so. It's the first time that I read it, but IMO doesn't make sense to spend A LOT of transistors to introduce OoO for the "integer" part, and don't use them for the FPU and/or the SIMD unit.
Take a look at this: http://www.anandtech.com/show/6936/intels-silvermont-architecture-revealed-getting-serious-about-mobile/2 |
| Status: Offline |
| |
|
|
|
[ home ][ about us ][ privacy ]
[ forums ][ classifieds ]
[ links ][ news archive ]
[ link to us ][ user account ]
|