SemiAccurate Forums  

 
Go Back   SemiAccurate Forums > Search Forums

Showing results 1 to 25 of 500
Search took 0.02 seconds.
Search: Posts Made By: aaronspink
Forum: Article discussion 08-26-2014, 06:26 AM
Replies: 574
Views: 45,927
Posted By aaronspink
That's some pretty ridiculous BS that you are...

That's some pretty ridiculous BS that you are spewing there. The issue with 1 particular publisher is because that publisher wants to set digital prices that are ABOVE the same prices they are...
Forum: Article discussion 08-21-2014, 06:54 PM
Replies: 81
Views: 6,411
Posted By aaronspink
Ah, good old Juanrga: Paper with actual data...

Ah, good old Juanrga:

Paper with actual data that is published and reviewed by multiple people: Obvious Intel Plant

Random PR slides: has to be 100% accurate!
Forum: Article discussion 08-21-2014, 05:48 AM
Replies: 60
Views: 7,179
Posted By aaronspink
The Relevance is 100%. All known designs and all...

The Relevance is 100%. All known designs and all known devices are going with an MCM implementation. The markets that both HBM and HMC are aimed add either require too much power or too much...
Forum: Article discussion 08-20-2014, 07:34 PM
Replies: 60
Views: 7,179
Posted By aaronspink
You can show anything in a slide. So far, the...

You can show anything in a slide. So far, the only things HBM and HMC have shown as going to release are fully integrated stacks for MCM integration with other dies.
Forum: CPUs 08-19-2014, 08:28 PM
Replies: 680
Views: 77,038
Posted By aaronspink
The analysis in the MPR article doesn't even...

The analysis in the MPR article doesn't even arise to compare to many spreadsheet estimates I've seen in the industry. There is simply too much data lacking to ascribe the MPR article anything but...
Forum: Article discussion 08-19-2014, 08:21 PM
Replies: 60
Views: 7,179
Posted By aaronspink
That's not the intention of either HBM or HMC. ...

That's not the intention of either HBM or HMC. The intention is that the fully stacked parts including the comms/test chip will be bonded at the initial manufacturing facility and then shipped to...
Forum: Article discussion 08-19-2014, 08:18 PM
Replies: 60
Views: 7,179
Posted By aaronspink
Um, neither HMC nor HBM are really meant for a...

Um, neither HMC nor HBM are really meant for a direct vertical stack with the primary logic die. They are both designed around package level integration instead. While there is potential to replace...
Forum: CPUs 08-19-2014, 08:13 AM
Replies: 680
Views: 77,038
Posted By aaronspink
It matters to some degree, but using SIMD vs lack...

It matters to some degree, but using SIMD vs lack of SIMD as an argument is pretty fundamentally flawed. What matters in that case is the existence of SIMD, not the ISA expression of that SIMD.
...
Forum: CPUs 08-19-2014, 06:53 AM
Replies: 680
Views: 77,038
Posted By aaronspink
You also might want to wait for chips to actually...

You also might want to wait for chips to actually sample before starting to make comparisons with them. There is naturally a lot of handwaving in something like the uP report article and Cavium...
Forum: CPUs 08-19-2014, 06:32 AM
Replies: 680
Views: 77,038
Posted By aaronspink
You might want to try not quoting the same...

You might want to try not quoting the same article that presents no actual data to support your position over and over again. One can search the web to find an opinion piece that supports just about...
Forum: Article discussion 08-18-2014, 06:07 PM
Replies: 73
Views: 5,080
Posted By aaronspink
All three are either using or moving to FinFET in...

All three are either using or moving to FinFET in their next process. Your attempts of obfuscation on this are really poor.

Any of the major manufacturers could of reasonably switched to FDSOI...
Forum: Article discussion 08-18-2014, 07:02 AM
Replies: 73
Views: 5,080
Posted By aaronspink
If it was actually true then Intel, Samsung, and...

If it was actually true then Intel, Samsung, and TSMC would all be using FDSOI. I'll take the actual practice of those three over what some study by someone who isn't actually producing volume says....
Forum: Article discussion 08-18-2014, 02:47 AM
Replies: 73
Views: 5,080
Posted By aaronspink
No, they don't. IBM is the biggest...

No, they don't.



IBM is the biggest consumer of SOI. Outside of IBM, use of SOI is pretty miniscule in the industry. There are multiple reasons for that and it isn't because SOI is a brain...
Forum: Article discussion 08-18-2014, 02:38 AM
Replies: 73
Views: 5,080
Posted By aaronspink
This is pretty much the same as it ever was. ...

This is pretty much the same as it ever was. Every couple years, TSMC et al, appear to be getting close. You should go back memory lane and read some of the hype around how TSMC 28nm was so close...
Forum: CPUs 08-13-2014, 05:01 PM
Replies: 680
Views: 77,038
Posted By aaronspink
AFAIK, only 1 company has ever gotten anything...

AFAIK, only 1 company has ever gotten anything approaching a true ultimate license. And that was DEC. DEC as basically allowed to actually extend the ISA to fit their needs. The DEC SA1500 design...
Forum: CPUs 08-12-2014, 03:35 PM
Replies: 680
Views: 77,038
Posted By aaronspink
I assure you that even in the world of TPC,...

I assure you that even in the world of TPC, single thread performance still matters. There are a lot of compound effects of running lots of low single thread performance cores.




Most GPUs...
Forum: CPUs 08-11-2014, 06:40 PM
Replies: 119
Views: 7,601
Posted By aaronspink
Move only is cleaner than all of them, doesn't...

Move only is cleaner than all of them, doesn't mean it wouldn't suck. Clean != good.

And I see you don't understand what volume actually means. ARM could sell 10 trillion toasters, it has jack...
Forum: CPUs 08-11-2014, 06:22 PM
Replies: 680
Views: 77,038
Posted By aaronspink
No its because lots of small cores doesn't work...

No its because lots of small cores doesn't work for lots of workloads cause of small minor issues like Amdahl's law, interconnection network complexity/size, coherency, conflict, et al.


...
Forum: CPUs 08-11-2014, 08:51 AM
Replies: 680
Views: 77,038
Posted By aaronspink
Most memory is directly attached to chips these...

Most memory is directly attached to chips these days...

I assume you mean packaged on chip and connected via TSV. Yeah, that's not going to happen for pretty much any chip that is going to use...
Forum: Article discussion 08-11-2014, 05:54 AM
Replies: 81
Views: 6,411
Posted By aaronspink
Migration to x86-64 actually took quite some...

Migration to x86-64 actually took quite some time.



You should be wary of using embedded advertising as your source material...
Forum: Article discussion 08-11-2014, 05:51 AM
Replies: 81
Views: 6,411
Posted By aaronspink
All depends on what peripherals are used in the...

All depends on what peripherals are used in the servers and what the state of the drivers are. Servers tend to be anything but plug and play generally. And the current industry trend is greatly...
Forum: CPUs 08-11-2014, 05:45 AM
Replies: 119
Views: 7,601
Posted By aaronspink
Who claimed the code couldn't run? **crickets**....

Who claimed the code couldn't run? **crickets**. I think you are missing the argument.



No, they aren't.




And not it isn't from "ISA Aging: A x86 case study". Its from...
Forum: CPUs 08-11-2014, 05:30 AM
Replies: 119
Views: 7,601
Posted By aaronspink
Mont-blanc I presume. Here's a phenomenal idea. ...

Mont-blanc I presume. Here's a phenomenal idea. Lets replace the easy to get high performance part with a low performance part, but instead of using the already strained complex hard to design...
Forum: CPUs 08-11-2014, 05:14 AM
Replies: 119
Views: 7,601
Posted By aaronspink
The ARM ISA is essentially the same ARM ISA as...

The ARM ISA is essentially the same ARM ISA as was used in the 80s. The only tweaks were basically thumb.

And ARM64 is basically a copy of Alpha with more cruft. Its really nothing new under...
Forum: CPUs 08-11-2014, 05:13 AM
Replies: 119
Views: 7,601
Posted By aaronspink
You do realize that there is nothing as clean as...

You do realize that there is nothing as clean as a single instruction ISA. AKA MOV only ISA. Yeah, they are still in fantasy land for a good reason. Clean is simply clean. Clean != good.


...
Showing results 1 to 25 of 500

 
Forum Jump

All times are GMT -5. The time now is 08:57 AM.


Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2014, vBulletin Solutions, Inc.
SemiAccurate is a division of Stone Arch Networking Services, Inc. Copyright © 2009 Stone Arch Networking Services, Inc, all rights reserved.