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: CPUs Today, 07:59 AM
Replies: 291
Views: 38,889
Posted By aaronspink
No you don't. Neither of those is measuring SoC...

No you don't. Neither of those is measuring SoC power. Both at measuring wall and include the whole entire laptop. Both laptops also have significant difference aside from just the processors. ...
Forum: CPUs Today, 07:46 AM
Replies: 291
Views: 38,889
Posted By aaronspink
No, I'm suggesting the only actual measurement of...

No, I'm suggesting the only actual measurement of SoC power in the whole review shows the SoC at <=15W.
Forum: CPUs Today, 07:23 AM
Replies: 875
Views: 115,508
Posted By aaronspink
Latency gets worse with bank conflicts. When...

Latency gets worse with bank conflicts. When running a continuous CAS stream, latency is basically unchanged.

Latency takes a nose dive anytime you have to re-ras a bank that you are already...
Forum: CPUs Today, 07:04 AM
Replies: 875
Views: 115,508
Posted By aaronspink
Latency isn't 1/2. tFAW doesn't really have much...

Latency isn't 1/2. tFAW doesn't really have much at all to do with latency. In fact, if you are running into tFAW, your latency is going to BLOW. Cause it basically means you are doing a full...
Forum: CPUs Yesterday, 10:09 PM
Replies: 875
Views: 115,508
Posted By aaronspink
No it isn't. The time of flight latency is...

No it isn't. The time of flight latency is rather minimal. For example, removing all time of flight latency and using only SoC edge round trip timing (the most favorable to your argument!) would...
Forum: CPUs Yesterday, 09:48 PM
Replies: 875
Views: 115,508
Posted By aaronspink
HBM presents the bog standard PRE/RAS/CAS...

HBM presents the bog standard PRE/RAS/CAS interface that dates back to at least SDRAM. HBM decidedly does not move the memory controller onto the stack. For that, you have to go to HMC. The logic...
Forum: CPUs Yesterday, 09:46 PM
Replies: 875
Views: 115,508
Posted By aaronspink
Actually their slides say EXACTLY NOT WHAT YOU...

Actually their slides say EXACTLY NOT WHAT YOU ARE STATING. Their slides have CAS/RAS/PRE at exactly the same latencies.



Again, latency will basically be unchanged from any other modern...
Forum: CPUs Yesterday, 09:28 PM
Replies: 291
Views: 38,889
Posted By aaronspink
Your supposed evidence shows a 15W part taking...

Your supposed evidence shows a 15W part taking 15W. You might need to find some actual evidence for your claims.
Forum: CPUs 02-25-2015, 01:23 PM
Replies: 875
Views: 115,508
Posted By aaronspink
There's just too much to overcome. HBM/HMC loses...

There's just too much to overcome. HBM/HMC loses roughly 27-40 off the top in density. Its going to lose likely another 10%+ in yield. That puts it 40-55% behind the cost curve. That's ~$40 in...
Forum: CPUs 02-24-2015, 04:15 PM
Replies: 875
Views: 115,508
Posted By aaronspink
Its not really opinion. It is pretty much fact. ...

Its not really opinion. It is pretty much fact. The additional steps for TSV in the wafer processing, the additional yield fallout, additional testing steps, higher cost more complex packaging. In...
Forum: CPUs 02-24-2015, 05:16 AM
Replies: 875
Views: 115,508
Posted By aaronspink
Yeah, I just don't see any only HBM designs...

Yeah, I just don't see any only HBM designs anytime soon. I wouldn't be surprised if intel went with either HBM or HMC instead of their custom RAM for GT3/4e. And I could see AMD having a higher...
Forum: CPUs 02-24-2015, 05:02 AM
Replies: 875
Views: 115,508
Posted By aaronspink
As long as the HBM interface is still...

As long as the HBM interface is still PRE-RAS-CAS, the memory controller will largely be exactly the same. If anything, HBM has more complexity than standard DDR3/4 interfaces. HMC is the only...
Forum: CPUs 02-24-2015, 04:55 AM
Replies: 875
Views: 115,508
Posted By aaronspink
1) HBM/HMC will always have higher costs than...

1) HBM/HMC will always have higher costs than DDR3/DDR4 unless/until DDR3/4 doesn't exist as a viable product. The inherent costs of HBM/HMC are significantly higher.

2) Bandwidth will increase...
Forum: CPUs 02-24-2015, 04:41 AM
Replies: 875
Views: 115,508
Posted By aaronspink
DRAM latency is largely fixed. The time of...

DRAM latency is largely fixed. The time of travel delay on the wires is a minimal factor. DRAM bank latency improves rather slowly. Pretty much the best latency 10 years ago (~8-10ns CAS) is the...
Forum: CPUs 02-24-2015, 04:07 AM
Replies: 875
Views: 115,508
Posted By aaronspink
Lets put it this way, any chart that says that...

Lets put it this way, any chart that says that WideIOn is going to cost more than HBM simply cannot be trusted. HBM is basically multi-stack WideIOn. AKA HBM will always be more expensive than...
Forum: Article discussion 02-07-2015, 06:54 AM
Replies: 477
Views: 56,745
Posted By aaronspink
FMULS/FADDs same difference really. Still quite...

FMULS/FADDs same difference really. Still quite shocked that a major bug like that wasn't the lede for the story and instead was buried.
Forum: Article discussion 02-07-2015, 12:10 AM
Replies: 477
Views: 56,745
Posted By aaronspink
Don't know, since Denver has a massive functional...

Don't know, since Denver has a massive functional issue ATM, I'm not sure you can consider any Denver numbers valid.
Forum: Article discussion 02-07-2015, 12:09 AM
Replies: 477
Views: 56,745
Posted By aaronspink
They are probably still trying to get it not to...

They are probably still trying to get it not to do a reset when you do FADDs ;)
Forum: Article discussion 01-08-2015, 04:42 AM
Replies: 10
Views: 1,302
Posted By aaronspink
One of the major advantages of an impeller based...

One of the major advantages of an impeller based design is that it is mostly self cleaning. The centrifugal forces on any particle (ie dust/dander/etc) that enters the vortex and the impeller area...
Forum: Article discussion 01-07-2015, 07:59 PM
Replies: 10
Views: 1,302
Posted By aaronspink
the air bearing is dynamic and self calibrating...

the air bearing is dynamic and self calibrating (at such a small gap its the only way to do it, esp when a small bump could destroy it other wise). As the impeller spins up, the friction forces...
Forum: Article discussion 01-07-2015, 03:50 AM
Replies: 47
Views: 4,652
Posted By aaronspink
yep, there will also be all the -e and -ep parts...

yep, there will also be all the -e and -ep parts as well.
Forum: Article discussion 01-07-2015, 03:48 AM
Replies: 47
Views: 4,652
Posted By aaronspink
You can keep thinking that and keep being wrong...

You can keep thinking that and keep being wrong all you like. Unfortunately the reality is that even the vast majority of games can't really make use of more than a couple cores. There is basically...
Forum: GPUs 01-06-2015, 01:31 AM
Replies: 135
Views: 12,934
Posted By aaronspink
No, no it cannot. Don't let Nvidia's marketing...

No, no it cannot. Don't let Nvidia's marketing fool you. Counting FP16 as the same as FP32 and FP64 is complete marketing BS.
Forum: Article discussion 01-06-2015, 01:24 AM
Replies: 47
Views: 4,652
Posted By aaronspink
I'll bet you a fairly sizable amount of money...

I'll bet you a fairly sizable amount of money that Intel is going to ship parts with upwards of 150w TDPs on 14nm!
Forum: Article discussion 01-06-2015, 01:23 AM
Replies: 47
Views: 4,652
Posted By aaronspink
Come on lets not be absurd. The number of people...

Come on lets not be absurd. The number of people that really use 2 CPUs is fairly small. The number that can use 4 is smaller still. The number that can use 8 is practically non-existent.

The...
Showing results 1 to 25 of 500

 
Forum Jump

All times are GMT -5. The time now is 11:48 AM.


Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2015, vBulletin Solutions, Inc.
SemiAccurate is a division of Stone Arch Networking Services, Inc. Copyright 2009 Stone Arch Networking Services, Inc, all rights reserved.