Poster | Thread |
Rob
 |  |
Re: Some Power related news Posted on 24-Aug-2016 19:18:57
| | [ #121 ] |
|
|
 |
Elite Member  |
Joined: 20-Mar-2003 Posts: 6399
From: S.Wales | | |
|
| @AmigaBlitter
Quote:
OpenPOWER systems allow you to keep your valuable assets and proprietary engine code safe and secure through full owner control, while outperforming similarly priced systems running on the x86 and ARM architectures.
Is the latter half of that sentence a realistic expectation?Last edited by Rob on 24-Aug-2016 at 07:20 PM.
|
|
Status: Offline |
|
|
iggy
|  |
Re: Some Power related news Posted on 24-Aug-2016 19:44:49
| | [ #122 ] |
|
|
 |
Super Member  |
Joined: 20-Oct-2010 Posts: 1175
From: Bear, Delaware USA | | |
|
| @Rob
Neat demo, but "outperforming similarly priced systems running on the x86 and ARM architectures"?
Well, Power8 is not cheap, and comparably priced ARM and X64 systems would not usually be used for game play.
IF Talos ever makes this board or find a company to license it, I will consider one. |
|
Status: Offline |
|
|
AmigaBlitter
|  |
Re: Some Power related news Posted on 24-Aug-2016 20:29:19
| | [ #123 ] |
|
|
 |
Elite Member  |
Joined: 26-Sep-2005 Posts: 3514
From: Unknown | | |
|
| @iggy
Seems that P8 servers are now cheaper than x86 xeon servers or at least the price is very similar.
_________________ retired |
|
Status: Offline |
|
|
OneTimer1
|  |
Re: Some Power related news Posted on 24-Aug-2016 22:21:07
| | [ #124 ] |
|
|
 |
Super Member  |
Joined: 3-Aug-2015 Posts: 1160
From: Germany | | |
|
| |
Status: Offline |
|
|
Yssing
|  |
Re: Some Power related news Posted on 25-Aug-2016 11:30:23
| | [ #125 ] |
|
|
 |
Super Member  |
Joined: 24-Apr-2003 Posts: 1118
From: Unknown | | |
|
| @OneTimer1
895 dollars is not a lot of money. _________________
|
|
Status: Offline |
|
|
pavlor
|  |
Re: Some Power related news Posted on 25-Aug-2016 11:47:58
| | [ #126 ] |
|
|
 |
Elite Member  |
Joined: 10-Jul-2005 Posts: 9682
From: Unknown | | |
|
| @Yssing
Add up to $1,845.43 for shipping (outside US)...
4.2 GHz Power6 scores 16.9 to 19.2 SpecInt2006 (Comparable to Core 2 E6600 2.4 GHz) = 2x faster than P5020 2 GHz in the X5000.
Edit: Shipping to my country (Czech Republic) seems to be far more expensive than to war-torn Yemen!  Last edited by pavlor on 25-Aug-2016 at 11:52 AM. Last edited by pavlor on 25-Aug-2016 at 11:49 AM.
|
|
Status: Offline |
|
|
pavlor
|  |
Re: Some Power related news Posted on 25-Aug-2016 12:00:10
| | [ #127 ] |
|
|
 |
Elite Member  |
Joined: 10-Jul-2005 Posts: 9682
From: Unknown | | |
|
| |
Status: Offline |
|
|
Yssing
|  |
Re: Some Power related news Posted on 25-Aug-2016 21:28:24
| | [ #128 ] |
|
|
 |
Super Member  |
Joined: 24-Apr-2003 Posts: 1118
From: Unknown | | |
|
| @pavlor
I agree that the shipping is ridiculous, I was only commenting on the price for the hardware _________________
|
|
Status: Offline |
|
|
iggy
|  |
Re: Some Power related news Posted on 26-Aug-2016 16:05:39
| | [ #129 ] |
|
|
 |
Super Member  |
Joined: 20-Oct-2010 Posts: 1175
From: Bear, Delaware USA | | |
|
| @pavlor
Quote:
Comparable to Core 2 E6600 2.4 GHz |
So...I have a dual Xeon X5460 3.16GHz Quad core system with 32GB of memory and 4 1TB drives that I built for under $500.
Similar processors (although no hyperthreading) for four threads per cpu or eight threads total at 3.16 GHz vs. four threads (with hyperthreading) for the e6600 at 2.4 GHz.
Obviously Intel hardware is cheaper and faster.
But I really wouldn't want to ship something like this to the Czech Republic as it is big and heavy.
And, even though it is illogical, I'd still like a Talos board. |
|
Status: Offline |
|
|
Rose
|  |
Re: Some Power related news Posted on 26-Aug-2016 18:55:31
| | [ #130 ] |
|
|
 |
Cult Member  |
Joined: 5-Nov-2009 Posts: 982
From: Unknown | | |
|
| @iggy
Quote:
Obviously Intel hardware is cheaper and faster. |
Nope. G5 is faster than i7 and if you disagree with "benchmarks" you are "dick". |
|
Status: Offline |
|
|
kolla
|  |
Re: Some Power related news Posted on 26-Aug-2016 19:49:15
| | [ #131 ] |
|
|
 |
Elite Member  |
Joined: 20-Aug-2003 Posts: 3418
From: Trondheim, Norway | | |
|
| How many i7 mac minis can comfortably fit and run inside a G5 tower? :) _________________ B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC |
|
Status: Offline |
|
|
Rob
 |  |
Re: Some Power related news Posted on 27-Aug-2016 21:46:30
| | [ #132 ] |
|
|
 |
Elite Member  |
Joined: 20-Mar-2003 Posts: 6399
From: S.Wales | | |
|
| @kolla
Depends if you crush them first or not. |
|
Status: Offline |
|
|
fishy_fis
|  |
Re: Some Power related news Posted on 28-Aug-2016 4:37:50
| | [ #133 ] |
|
|
 |
Elite Member  |
Joined: 29-Mar-2004 Posts: 2168
From: Australia | | |
|
| Was interesting to see Power based chips are now starting to favor little endian.
As far as Power8 and 9 servers go, there has been a concerted attempt to make them comparable to x86 servers in terms of price/performance. It's worth remembering however that a servers performance metrics are measured differently to a desktops. Raw cpu/gpu grunt wise x86 is a completely different kettle of fish. Nothing else really comes close.
p.s. That Unreal4 Engine demo is quite cool. An AROS version is hopefully not too far away (and no, this isnt just blind hope, but rather based on work being done on it). Last edited by fishy_fis on 28-Aug-2016 at 08:43 AM.
|
|
Status: Offline |
|
|
Beans
|  |
Re: Some Power related news Posted on 28-Aug-2016 13:47:28
| | [ #134 ] |
|
|
 |
Regular Member  |
Joined: 26-Aug-2016 Posts: 447
From: Bear Delaware USA | | |
|
| @kolla
Quote:
How many i7 mac minis can comfortably fit and run inside a G5 tower? |
How many Mac Minis can you use at the same time? And can you use them if they are inside of a G5 PowerMac? ;)_________________ Amiga! "Our appeal has become more selective" |
|
Status: Offline |
|
|
Beans
|  |
Re: Some Power related news Posted on 28-Aug-2016 13:51:13
| | [ #135 ] |
|
|
 |
Regular Member  |
Joined: 26-Aug-2016 Posts: 447
From: Bear Delaware USA | | |
|
| @fishy_fis
Quote:
Raw cpu/gpu grunt wise x86 is a completely different kettle of fish |
Seriously, that turn of phrase is completely obtuse. In what fashion is the X64 ISA that much more suited to desktop use?_________________ Amiga! "Our appeal has become more selective" |
|
Status: Offline |
|
|
fishy_fis
|  |
Re: Some Power related news Posted on 28-Aug-2016 15:20:36
| | [ #136 ] |
|
|
 |
Elite Member  |
Joined: 29-Mar-2004 Posts: 2168
From: Australia | | |
|
| @Beans
Who said anything about the ISA? Simple fact is server performance isnt measured by the same metrics.
Different workloads. Until recently for example an Opteron compared favorably against a Xeon, despite the core architecture being somewhat weaker for typical desktop use.
As for desktop use, x86's performance advantage vs other architectures is pretty obvious. Last edited by fishy_fis on 28-Aug-2016 at 03:30 PM.
|
|
Status: Offline |
|
|
AmigaBlitter
|  |
Re: Some Power related news Posted on 29-Aug-2016 14:24:03
| | [ #137 ] |
|
|
 |
Elite Member  |
Joined: 26-Sep-2005 Posts: 3514
From: Unknown | | |
|
| |
Status: Offline |
|
|
Hypex
 |  |
Re: Some Power related news Posted on 29-Aug-2016 15:55:28
| | [ #138 ] |
|
|
 |
Elite Member  |
Joined: 6-May-2007 Posts: 11351
From: Greensborough, Australia | | |
|
| @fishy_fis
Quote:
Was interesting to see Power based chips are now starting to favor little endian. |
Yeah just looks like Power is losing its personality on that one.
I would say it is more being forced into the litle endian mould than to favor it. Since most of the worlds computers today run in little endian, code would be tested on little endian and I'd expect on nothing else. If code isn't really portable they won't see nor care since their market is covered.
We see evidence of this with JavaScript interpreters which now have endian problems on big endian. I thought compilers might actualy get a bit smarter and have a WYSIWYG data model, at least for text IDs, where looking for "THIS" will find "THIS" regardless of it looking like "SIHT" or "THIS" in memory. Or support an endian attribute. But no it looks not. Even in these high level times. Of course writing bi-endian code is bit more complicated.
I don't know why I dislike little endian so much now. I used to be entertained by it on the 6510 of a C64. Even if I had a 7501/8501 in my C16 at home. When dealing with lo byte, hi byte. Fun times. Perhaps it was when I loaded MS BMP images into my Amiga and looked at it in a hex dump thay my dislike began. No wonder I couldn't understand the damn header! And when porting cryptic little endian specific code to big endian a few years back must have hated it more.  Last edited by Hypex on 29-Aug-2016 at 03:57 PM.
|
|
Status: Offline |
|
|
AmigaBlitter
|  |
Re: Some Power related news Posted on 6-Sep-2016 18:41:19
| | [ #139 ] |
|
|
 |
Elite Member  |
Joined: 26-Sep-2005 Posts: 3514
From: Unknown | | |
|
| |
Status: Offline |
|
|
Beans
|  |
Re: Some Power related news Posted on 6-Sep-2016 19:36:49
| | [ #140 ] |
|
|
 |
Regular Member  |
Joined: 26-Aug-2016 Posts: 447
From: Bear Delaware USA | | |
|
| @fishy_fis
Quote:
x86's performance advantage vs other architectures is pretty obvious. |
Maybe to you, I just don't see it. On the desktop, X64 has multimedia instructions that clearly make it a better performer. In servers its all come down to performance per watt. And on that metric X64 loses.
@ AmigaBlitter
Interesting article. Good performance on the benches available. And the 8 core processor is at the low end of the Power8 range with 10 and 12 core cpus topping out at over 4GHz.
So...X64 is probably NOT "a better performer".
Last edited by Beans on 06-Sep-2016 at 07:57 PM.
_________________ Amiga! "Our appeal has become more selective" |
|
Status: Offline |
|
|