• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

TSMC and ARM Tape-Out First ARM Cortex-A57 Processor on 16 nm FinFET Technology

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
zsdersw said:
It's always easier to look like you're making a lot of progress when you're far behind than when you're in the lead.
Click to expand...
Then why is it hard for most here on AT to acknowledge that Intel is being surpassed, perhaps not in absolute terms but certainly relative performance gains, & that sometime in the future ARM/AMD can even take the top performer crown from them ?
 
Idontcare said:
So a no-name chinese design house managed to upstage AMD in getting 28nm functional silicon out of GloFo? 😵 Really?
Click to expand...

You've got admit that ARM is a significantly simpler design than any of the big core AMD cpu's. Plus that chinese tablet company most likely isn't as sensitive to supply disruption.
 
R0H1T said:
You're getting into semantics here, it should be clear enough what I was talking about, though power efficiency would be more apt right ?As someone said earlier this is a niche market so its not wrong to assume that with multiple ARM vendors coming in & certain other costs going down it'd be naive to think that ARM servers would remain prohibitively expensive as they are now !I'd like to see when/where is the clover trail comparable to Exynos Octa or Snapdragon 600 ?
Click to expand...
Well, if we grant context, the throughput would be synonymous with just performance. How fast the computer gets things done.

The thing about performance per watt is that you it can be large or small depending on either of the variables. Hence, a high-performance, high-wattage rig could very well be more efficient than a low-performance, low-wattage rig. It all depends on the values in the ratio. You don't know the actual performance until you cancel out the units and are left with just watts or just a performance number.
 
R0H1T said:
Then why is it hard for most here on AT to acknowledge that Intel is being surpassed, perhaps not in absolute terms but certainly relative performance gains, & that sometime in the future ARM/AMD can even take the top performer crown from them ?
Click to expand...


Your glowing optimism about ARM/AMD and simultaneous vibrant pessimism about Intel is dubious.
 
R0H1T said:
Then why is it hard for most here on AT to acknowledge that Intel is being surpassed, perhaps not in absolute terms but certainly relative performance gains, & that sometime in the future ARM/AMD can even take the top performer crown from them ?
Click to expand...

Just review the other threads on this. The answers to your question have already been given.
 
You guys are acting like the quad Cortex-A9 CPU is the only thing on that SoC. That's only going to be a fraction of the die area. A much smaller fraction than it'd be on an AMD CPU, or even APU. That doesn't make one more complex than the other but it still means you're comparing pretty different things.

Here's the real discrepancy: RK3188 needs a power tuned (low leakage) process, while anything AMD would make on GF 28nm needs a performance/frequency tuned process. They're not the same thing.
 
Phynaz said:
Are you including when they moved backwards in performance? Or is that conveniently left out of your calculation?
Click to expand...
Yes but that can be excused if you're switching your microarch from the tried & tested K10 to Bulldozer & the AMD top dog responsible for this is now gone "permanently" for AMD's sake 😛
 
Phynaz said:
Just review the other threads on this. The answers to your question have already been given.
Click to expand...
There is no other thread on ARM 64bit IIRC anyways I'll leave it at that & wait for the time we have definitive numbers cause it's all speculation till then !
 
Charles Kozierok said:
I don't dislike Intel, but there are entirely valid reasons why people want to see the company have some strong competition.
Click to expand...

True. They weren't always steered in the right direction, i.e Netburst. They probably wouldn't be emphasizing all this power consumption reduction if it weren't for them being massacred in the mobile space.
 
The arm ecosystem is just on a roll now. We are a month or two before quallcomm hits with the s800 with everything integrated. Wasnt IP2 made with arm11? S600 is allower with the s4. Man its going so fast. If it continues like it have done the past 4 years, a57 is going to be a blast. Intel just didnt catch that train.

I dont think its slowing down. Qualcomm is learning it and just have everything lined up for the next years. Tsmc capacity is skyrockening. There is tons of synergy here between lots of business. There is lost of customers and lots of money. I havnt predicted that explosion initially created by apple. Ofcourse its going to stop at some time, but what are the sings for that?
 
sontin said:
And it seems that TSMC is two months ahead of schedule for their 20nm production: http://focustaiwan.tw/news/atod/201304010042.aspx

So the cap between Intel's 22nm Atom and ARM's 20nm SoC will be very close.
Click to expand...

Intel is moving Atom to 14nm in 2014, so don't keep your hopes too high.

Also, TSMC could launch 20nm chips if they wanted, but yields would be bad, very bad. By running ahead of their schedule, the process can be maturing faster than predicted or they may just be accepting lower yields to allow an earlier launch.
 
R0H1T said:
Yes assuming its run on an Intel machine, who's to say that the same VM wouldn't deliver better throughput(as in power/perf) on ARM servers ? This all is hypothetical of course but there is no reason to believe that ARM can't deliver this level of performance based on their recent past(last ~5yrs or so) while maintaining similar levels of efficiency !
Click to expand...

Because the article states that you can't use it virtualized (yet) and it is not intended fo that use anyway:

Returning to our Boston Viridis server, the whole idea behind the server is not to virtualize but to give each server application its own physical node.
Click to expand...

It's not all about efficiency and performance/watt. If you have great performance/watt. but your performance is 1000 times to weak the product is still useless.

And what I said before. managing 24 virtual machines will be much more flexible because each machine can have a different load you only have to consider the total load the server. Also managing them is much, much easier and hence you save on the most expensive part of all: employees.
 
Just the fact that we are discussing serious competition for Intel is very interesting/exciting. I am a big fan of any technology that kicks ass. And the more competition the more ass kicking.

We have the Athlon (in large part) to thank for the Core architecture and we'll see more advancements as ARM and Intel start to cross paths. For the time being they have occupied different territory but now they are starting to take shots across one another's bow.
 
mrmt said:
Intel is moving Atom to 14nm in 2014, so don't keep your hopes too high.
Click to expand...
In the best case that will be the year of TSMCs 16nm Finfet process, too, if I count the months correctly:

TSMC aims to have chip design kits for its 16-nm process available in January with the first foundation IP blocks such as standard cells and SRAM blocks ready a month later. It will start limited so-called “risk” production of the 16-nm process in November 2013. Production chip tape outs will follow about four or five quarters later.
Click to expand...
http://www.eetimes.com/electronics-news/4398727/TSMC-taps-ARM-V8-in-road-to-16-nm-FinFET

Nov. 2013 + 4 Quarters = Nov. 2014.
 
SocketF said:
In the best case that will be the year of TSMCs 16nm Finfet process, too, if I count the months correctly:
Click to expand...

Actually the situations are a bit different between the two foundries. Intel is scheduled to have a 14nm process next year, and their 22nm process will have been online for two years, TSMC will field theirs in 2014.

So if TSMC is going for 16nm next year, this means that:

1) They will be swallowing a huge R&D bill by not monetizing their 20nm process for long enough.

2) They are going for GLF-like half-baked solutions, like their 14nm extreme lemon process.

Given TSMC track record, I can't see them doing neither 1 nor 2, meaning that 16nm will arrive in the 2015-2016 window.
 
Phynaz said:
You've got admit that ARM is a significantly simpler design than any of the big core AMD cpu's. Plus that chinese tablet company most likely isn't as sensitive to supply disruption.
Click to expand...

Big-core yes, but the 28nm bobcat shrink that had to be cancelled after AMD already invested 3yrs worth of R&D into it?
 
Do we have any projected benchmarks for the cortex a-57, for example dmips/mhz numbers? (yes dmips/mhz is a limited benchmark, but we are just now getting real world performance numbers for cortex a-15 and s600)
 
SocketF said:
In the best case that will be the year of TSMCs 16nm Finfet process, too, if I count the months correctly:

http://www.eetimes.com/electronics-news/4398727/TSMC-taps-ARM-V8-in-road-to-16-nm-FinFET

Nov. 2013 + 4 Quarters = Nov. 2014.
Click to expand...

In the best case yeah. In the more realistic case, aka a repeat of their 28nm timeline, it's about 6 quarters from risk production to actual products being released. (Risk production for 28nm HP was Q2 of 2010 - http://www.tsmc.com/tsmcdotcom/PRListingNewsAction.do?action=detail&newsid=4042&language=E ) As for 20nm and 16nm, yeah, they claim vague dates for risk production - http://www.tsmc.com/english/dedicatedFoundry/technology/future_rd.htm - of 2012 and 2014 respectively, so if we add in the realistic Q4 2012 for 20nm and Q4 2013 for 16nm we'd end up with actual products in Q2 of 2014 for 20nm and Q2 of 2015 for 16nm.

All that's assuming that these nodes go at least as well as 28nm did for them. Yeah, good luck with that.

Edit: Do note that the delay between risk production and product release depends upon both the manufacturing health and the design being fabbed. If the process health isn't adequate on the initial run and parts come back non-functional that pushes silicon design validation, which is a process that takes 2 quarters at a minimum and is more likely somewhere around a year. It ends up working reasonably well typically though since all that time can be used to tweak the new process to get yields up to production values.
 
Last edited:
You must log in or register to reply here.

TRENDING THREADS

Back
Top