flat assembler
Message board for the users of flat assembler.

Index > Main > New interesting x86 CPU from VIA

Author
Thread Post new topic Reply to topic
Kuemmel



Joined: 30 Jan 2006
Posts: 200
Location: Stuttgart, Germany
Kuemmel 04 May 2008, 09:20
Hi people,

there's a new x86 cpu from VIA coming up: VIA Isaiah...the white paper is available on:
http://www.via.com.tw/en/downloads/whitepapers/processors/WP080124Isaiah-architecture-brief.pdf
Web page:
http://www.via.com.tw/en/products/processors/isaiah-arch/

It's more targeted on small notebooks, and features only single core and up to 2 GHz...noneoftheless it has as many as 7 instruction ports, the shitty old VIA FPU is gone, 128bit SSE bandwidth, etc...there are not many benchmark results out there, but it could be one of the most efficient cpu's up to now even compared to Core 2 Duo and the white paper is really interesting to read. Hope Agner Fog will cover it some day when it's out regarding coding issues Wink May be somebody even can see already now how it overcomes some bottlenecks of current cpu's ?
Post 04 May 2008, 09:20
View user's profile Send private message Visit poster's website Reply with quote
Madis731



Joined: 25 Sep 2003
Posts: 2139
Location: Estonia
Madis731 04 May 2008, 16:24
Wow! World-record 2 clocks and 7 ex. ports Very Happy
Post 04 May 2008, 16:24
View user's profile Send private message Visit poster's website Yahoo Messenger MSN Messenger Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 20414
Location: In your JS exploiting you and your system
revolution 04 May 2008, 16:57
Not enough L2 cache, only 1MB. Still has the limit of 16 bytes/3 instructions per cycle. The main advantage I would hope for this CPU is battery life, that is always a major problem for laptops.
Post 04 May 2008, 16:57
View user's profile Send private message Visit poster's website Reply with quote
Madis731



Joined: 25 Sep 2003
Posts: 2139
Location: Estonia
Madis731 04 May 2008, 18:29
What IF (just a thought) Intel has 3MB per core (my T9300 example) L2 and 2x32KB L1 and is able to cache all of it while processing the millions of requests...hmh...
Now - imagine - VIA (that I secretly didn't like before) is able to cache all the requests on time (or even earlier) and when running out of space, can intelligently lose the old entries. This means that it doesn't need more. Where can we get 1MB+ random data that we need urgently, anyway? ^o)
Copying huge datasets goes past L2 and...

..ok, lets not jump to conclusions - waiting for 'em to prove their design!
Post 04 May 2008, 18:29
View user's profile Send private message Visit poster's website Yahoo Messenger MSN Messenger Reply with quote
bitRAKE



Joined: 21 Jul 2003
Posts: 4060
Location: vpcmpistri
bitRAKE 04 May 2008, 19:54
The Intel Atom seems to be of similar performance, but wouldn't a dual Intel Atom system use less power than a single processor Isaiah? Some interesting design choices, but not in the same power class as the Atom, imho.

_________________
¯\(°_o)/¯ “languages are not safe - uses can be” Bjarne Stroustrup
Post 04 May 2008, 19:54
View user's profile Send private message Visit poster's website Reply with quote
Kuemmel



Joined: 30 Jan 2006
Posts: 200
Location: Stuttgart, Germany
Kuemmel 04 May 2008, 20:05
bitRAKE wrote:
The Intel Atom seems to be of similar performance, but wouldn't a dual Intel Atom system use less power than a single processor Isaiah? Some interesting design choices, but not in the same power class as the Atom, imho.
Hm, I'm not so sure, some benches were out, though I would only trust the ones I coded myself Wink, but here's one link to them:
http://forum.eeeuser.com/viewtopic.php?pid=224565
I'm really not aware if this CrystalMark is decent code or represents something real world like...but anyway Atom doesn't look too good...I just like the fact some engineers started from scratch and seem to stay on one level with Intel...like back in time when this small Intel Israel division (as far as I know) came up with the Pentium-M design.
Post 04 May 2008, 20:05
View user's profile Send private message Visit poster's website Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 20414
Location: In your JS exploiting you and your system
revolution 04 May 2008, 20:12
But performance is not the issue. Right? Power usage is the key here so all of the above benchmarks are useless.

Still, I doubt it can beat the ARM xScale at <1mW idle and <1W full tilt.
Post 04 May 2008, 20:12
View user's profile Send private message Visit poster's website Reply with quote
bitRAKE



Joined: 21 Jul 2003
Posts: 4060
Location: vpcmpistri
bitRAKE 04 May 2008, 20:24
8W TDP for a dual core processor verses 25W for a single core Isaiah. Confused
They have a long way to go at VIA.

_________________
¯\(°_o)/¯ “languages are not safe - uses can be” Bjarne Stroustrup
Post 04 May 2008, 20:24
View user's profile Send private message Visit poster's website Reply with quote
Kuemmel



Joined: 30 Jan 2006
Posts: 200
Location: Stuttgart, Germany
Kuemmel 04 May 2008, 20:30
revolution wrote:
But performance is not the issue. Right? Power usage is the key here so all of the above benchmarks are useless.
Still, I doubt it can beat the ARM xScale at <1mW idle and <1W full tilt.
Yeah, I wouldn't think eiter...as Intel also had the ARM with xScale I think it was a strategy decision to go for x86...even knowing may be that they will never reach the efficiency of any ARM... Wink I remember back in time the story when somebody wanted to measure power consumption of an ARM2 cpu and they forgot to attach power to the cpu and still measured it while running. The power supply of the measurement connectors was enough to make it Wink ...getting nostalgic about a world without x86 and nice ARM's with some nice FPU/SSE2 unit Wink
Post 04 May 2008, 20:30
View user's profile Send private message Visit poster's website Reply with quote
Madis731



Joined: 25 Sep 2003
Posts: 2139
Location: Estonia
Madis731 05 May 2008, 09:46
ARM xScale? But what about Atom typical wattage of 0.6-2.5W and idle 0.01W. They are comparable I think. I bet ARM take a lot more time to software render something now that Atom has SSE support Smile You're not using your CPU to idle, but to do constructive work. What if you compare the same algorithm optimized for both architectures. Which finishes first and which consumed more power while doing that?
Post 05 May 2008, 09:46
View user's profile Send private message Visit poster's website Yahoo Messenger MSN Messenger Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 20414
Location: In your JS exploiting you and your system
revolution 05 May 2008, 10:13
MIPS/mW? Forget it, ARM will win. It is just incredibly efficient. If the Atom doesn't finish first by a wide margin then something is very wrong with it, they are just wasting power.
Post 05 May 2008, 10:13
View user's profile Send private message Visit poster's website Reply with quote
Kuemmel



Joined: 30 Jan 2006
Posts: 200
Location: Stuttgart, Germany
Kuemmel 29 May 2008, 19:54
More information about energy consumption and some performance figures out:
http://www.pcper.com/article.php?aid=565
http://www.via.com.tw/en/downloads/whitepapers/processors/WP080529VIA_Nano.pdf
It seems it's on the same level like a Core2Duo Design (Single Core) but needs only half the energy...not too bad I would say...still it's laking at the moment of the second core, but for sure will make a nice notebook with long lasting power I guess.
Post 29 May 2008, 19:54
View user's profile Send private message Visit poster's website Reply with quote
Display posts from previous:
Post new topic Reply to topic

Jump to:  


< Last Thread | Next Thread >
Forum Rules:
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum


Copyright © 1999-2024, Tomasz Grysztar. Also on GitHub, YouTube.

Website powered by rwasa.