VIA QuadCore Preview @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,634
VIA QuadCore Preview - VIA Technologies and Centaur Technologies are back with what it is calling the world's most efficient x86 multicore processor on the planet. Seeing so much done with so little is impressive, but what is even more impressive is seeing four cores falling into a power envelope of less than 28 watts TDP.
 
Very cool, I'd love to use one of these in a low power consumption server.
 
While I will probably never have the occasion to use one of their processors, I'm glad to see both companies thriving.
 
I wonder why VIA chooses to be at least a generation behind on node process. Why go .40nm instead of .32?
 
It's funny how they compare every performance chart to Brazos but then when you get to the power consumption chart they compare themselves to Llano.

VIA has been way behind for a long time. They waste an x86 license that other companies would absolutely kill for. I hope this is at least some indication of an attempt to make a competitive product. Now, does VIA have an answer to Intel's trigate?
 
So, on average, the Nano QC isn't even 50% faster than AMD E-350 when it comes to general computing... All for 50% higher power consumption... I'm sorry, but while it's nice that VIA has finally made a quad core Nano, it is seriously too late.

I would be thoroughly impressed if it was 18w @1.4/1.5GHz, because a quad core Bobcat derivative with integrated graphics would still use less power, and more than likely kill VIA's QuadCore in those benchmarks (CPU Mark 99, seriously?)

I wonder how QuadCore compares to a a i3-2100 in that AES benchmark.
 
Stoly said:
I wonder why VIA chooses to be at least a generation behind on node process. Why go .40nm instead of .32?
Click to expand...
probably resources. they have to go to buy their fab time from 3rd parties and don't have a sweetheart deal like AMD does with GLOBALFOUNDRIES. 32nm wafers are expensive and 40nm is probably cheaper.
niconx said:
It's funny how they compare every performance chart to Brazos but then when you get to the power consumption chart they compare themselves to Llano.
Click to expand...
Yea, its a tough spot to be in.. not low enough power to compete with bobcat/atom for minimum power consumption type applications, not powerful to compete with llano or i3 for applications that have a bigger power envelope

niconx said:
VIA has been way behind for a long time. They waste an x86 license that other companies would absolutely kill for. I hope this is at least some indication of an attempt to make a competitive product. Now, does VIA have an answer to Intel's trigate?
Click to expand...
You can't waste a license.. its not like theres a fixed number of them out there. If nVidia wanted one they'd have to pay for it, and if they bought VIA it would almost certainly void the license. Does VIA need an answer to Intel's tri-gate? Aside from atom and the CULV platform, Intel's technology doesn't play in the same arena as VIA's.. If they can win on cost they might not need some fancy 3d transistors
.
 
I cannot stand VIA. However, one good use for this and the only use I would use for is a pfSense router.
 
Very interesting article, and thanks for the information.

I like Centaur a lot. My one concern is the one I've always had --driver support for the Chrome GPU. The CPUs work fine; I bought an inexpensive C7-based laptop for a relative that saw good use and got decent battery life. However, like most VIA solutions, the Chrome graphics struggled, and driver support was mediocre.

If a good solid effort could be made in ensuring good driver development here, I'd have no problem using a VIA QuadCore for an HTPC rig or other small form-factor systems.
 
Activate: AMD said:
You can't waste a license.. its not like theres a fixed number of them out there. If nVidia wanted one they'd have to pay for it, and if they bought VIA it would almost certainly void the license. Does VIA need an answer to Intel's tri-gate? Aside from atom and the CULV platform, Intel's technology doesn't play in the same arena as VIA's.. If they can win on cost they might not need some fancy 3d transistors
.
Click to expand...

You can't waste a license? It doesn't seem Intel will allow new licenses or even old ones to continue. Sounds like x86 licenses are impossible to get.


If Intel lowered prices for the CPU's they would only get dragged back into court by AMD for dumping product on the market. At any rate 3D transistors are a significant development and no other x86 process can compete with it. It took Intel 10 years to develop that tech to the point of being able to use it in mass manufacturing. Other vendors can't lower their prices forever. At a certain point their products become so uncompetitive that a low price isn't enough. The whole situation seems to indicate Intel will put everyone out of business or at least make everyone else irrelevant and x86 will be everywhere., probably even replacing ARM once 16nm chips roll around.
 
One of the pics shows a Windows Experience Index 4.3, thats not good but likely hindered by memory and video card.
I would love to see one of these in a NAS.
 
Stoly said:
I wonder why VIA chooses to be at least a generation behind on node process. Why go 40nm instead of 32?
Click to expand...
From the previous interviews (elsewhere?), the Centaur team behind the VIA Nano is run on a shoestring budget and it takes much longer to design and release new processors than its bigger competitors.

VIA is really not that far behind AMD if the competition is Brazos. Brazos is only months old itself and also on 40nm.

I'd guess price also has a lot to do with the 40nm choice too. It would be a cheaper process than 32nm depending on whatever factors like price and volume it expects from 40nm Isaiah products.
 
You [H]Staffers need some photography classes LOL. :D But semi-kidding aside, awesome article!
 
niconx said:
You can't waste a license? It doesn't seem Intel will allow new licenses or even old ones to continue. Sounds like x86 licenses are impossible to get.

Click to expand...

Are you trying to disagree with me? You basically just said what I said and and added some links to prove it. The company has the license, what they do with it is what they choose to do with it. If its not a fixed commodity it can't be wasted. A "wasted license" would be if it WAS transferrable. In that case, a company that could/would do more with it would find it useful. as it is, no company can make use of VIA's license but VIA, so them having it doesn't deprive anybody else a license. Yes, nvidia wants/wanted one, but no matter how much they drool over VIA's license, they can't have it unless Intel says OK, which is exactly the same as Intel giving NV the ok to license x86 directly.

niconx said:
If Intel lowered prices for the CPU's they would only get dragged back into court by AMD for dumping product on the market. At any rate 3D transistors are a significant development and no other x86 process can compete with it. It took Intel 10 years to develop that tech to the point of being able to use it in mass manufacturing. Other vendors can't lower their prices forever. At a certain point their products become so uncompetitive that a low price isn't enough. The whole situation seems to indicate Intel will put everyone out of business or at least make everyone else irrelevant and x86 will be everywhere., probably even replacing ARM once 16nm chips roll around.
Click to expand...
First off, 3D transistors are not an x86 process, but a physical process. If other manufacturers can be price and application competitive without them, having them is completely irrelevant. TSMC, GF, and the other foundries don't have them yet and make enough wafers at the processes people want to still make money.

Its also doubtful that Intel would be allowed to put everyone out of business. You're not taking into account that Intel doesn't have to lower its price if the performance is good enough. They're a public company and want high margins. Other competitors are willing to cut their margins in order to fall into the space that Intel doesn't want because its not profitable enough.
 
Trackr said:
Wait.. 27.5 watts?!

A4 takes under 2!
Click to expand...
The A4 is an ARM architecture and Via is x86. Completely different. Via's chips are in direct competition with AMD's bobcat and Intel's Atom architectures. Now go play with your iPhone while the grownups continue the discussion.:p
 
Wait, did they ever fix that issue with data corruption through the 686B south bridge? My HPT370 is on its last legs. :(
 
zoso947 said:
Wait, did they ever fix that issue with data corruption through the 686B south bridge? My HPT370 is on its last legs. :(
Click to expand...

Please don's start me on that one! :(

Good to see the guys over at Centaur still so open, hope it works well for them. I look forward to getting my hands on it.
 
"Lowest Power x86 quadcore architecture"

This is just not true. Check this out:

http://www.notebookcheck.net/AMD-Phenom-II-X4-P960-Notebook-Processor.38642.0.html

Four cores, 25w TDP, 1.8 GHz. Not only is that 4 cores plus an integrated memory controller at a lower power point, but it will destroy the Via PressRelease(TM) by %50 clock-for-clock (compare AMD V Series V105 and VIA Nano U2500, which are both 1.2 GHz).

This makes the AMD CPU nearly %100 faster (when you take the clock speed advantage into account).

So, can we stop with the Via lovefest? They produce nothing but shiny powerpoint presentations - why the hell do major sites like the [H] even bother?
 
defaultluser said:
"Lowest Power x86 quadcore architecture"
Click to expand...
Oh no, marketing made untrue statements! :p I wouldn't get too excited about a 25W quad core when a slightly more expensive 20W dual core processor outperforms it. And the 20W processor isn't even a mobile chip. We all have bones to pick with something. ;)
 
pxc said:
Oh no, marketing made untrue statements! :p I wouldn't get too excited about a 25W quad core when a slightly more expensive 20W dual core processor outperforms it. And the 20W processor isn't even a mobile chip. We all have bones to pick with something. ;)
Click to expand...

But that's obvious: there are TONS of lower-wattage dual-cores that would wipe the floor with this POS. I was just pointing out that the carefully-worded marketing literature (essential to make this crap seem relevant) is complete bullshit.

Also, while I'm posting, I'd like to highlight another critical failure of this design: the lazy shared bus topology. They pour on the cores to save power versus higher frequency/voltages, BUT the poor shared bus used by these processors means multithreaded performance scaling sucks. Just take a look at those benchmarks I linked above.

h.264 performance results:

Nano Dual core 1.8: 467
Nano Single 1.2: 193

Looks impressive, right? Nope. Scale all the performance results linearly to 1.2 GHz, and you get the following:

Nano Dual core 1.8: 304
Nano Single 1.2: 193

That's roughly %50 scaling when you add the second core. Intel managed a whole helluva lot better using the same shared bus topology, so this is just sad. I can't imagine how slow it would be with 4 cores sucking-up bandwidth.
 
any word on releasing these small monsters ? i dont remember seeing VIA in action from a long time!
 
You must log in or register to reply here.
Back
Top