![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
What can we do to effect an actual change in Intel's corporate policy?
- Thread starter Really
- Start date
hainspuerterican
n00b
- Joined
- May 3, 2012
- Messages
- 7
Okay Really this is what you do. 1) Get an education, trust me buddy you need it. 2) Get a good paying job and invest whatever you can in INTC. 3) If you ever are presented the opportunity to become a VERY large shareholder in Intel go for it. Then they will listen, still wont change though. Also why are you pushing for hardware advancements? It's been known for a while now that software is the limiting factor in many aspects of computing.
Tsumi
[H]F Junkie
- Joined
- Mar 18, 2010
- Messages
- 13,755
Well, for normal computing, software has gotten very efficient and hardware has gotten more than powerful enough. You don't want a music player that fully loads one or two cores do you?
Software being the limiting factor is only applicable to games.
Hi guys,
I don't play games. I've nearly only played 1 game in the past ten years. And my wife never forgave me for that. I work 16 hours a day. That's why my Logitech rubber breaks.
I will probably die before ever getting a chance to play video games. Maybe in another life.
I'm sorry that I have to censor the words here,
For the XXXXX production I do, one single YYYYYY can use nearly a whole core. So we can only run a couple of these YYYYYYs at once, even though we need to run 100 of them to make proper ZZZZs.
The field of QQQ for XXXXX production is completely out of synch with Intel's technology. The Intel chips are an anachronism relative to the software requirements. The chips are just way way way behind by a factor of 100, so that you can barely even get your work done.
The slowness of the Intel chips at this time is just not compatible with emulating QQQQQQ JJJJJs. If you use any modern QQQQQQ JJJJJJJJJJJ or anything that has a high quality RRRRR, you are out of CPU instantaneously. You can only run a few instances of them. It's really bad.
It's very difficult to get work done, and everything you do has to be AAAAAA'd down. Most stuff can never be realtime. It's like working in the 1970's with JJJJ machines, we simply can't compute more than a couple of quality YYYYYYs.
Running something like the NNNNNN YYYYYY will use a half a core. You run out of CPU immediately.
There will be a time eventually when Intel's basic level of computation power allows most YYYYYY's to work fine. That time is not here.
lol
I don't play games. I've nearly only played 1 game in the past ten years. And my wife never forgave me for that. I work 16 hours a day. That's why my Logitech rubber breaks.
I will probably die before ever getting a chance to play video games. Maybe in another life.
I'm sorry that I have to censor the words here,
For the XXXXX production I do, one single YYYYYY can use nearly a whole core. So we can only run a couple of these YYYYYYs at once, even though we need to run 100 of them to make proper ZZZZs.
The field of QQQ for XXXXX production is completely out of synch with Intel's technology. The Intel chips are an anachronism relative to the software requirements. The chips are just way way way behind by a factor of 100, so that you can barely even get your work done.
The slowness of the Intel chips at this time is just not compatible with emulating QQQQQQ JJJJJs. If you use any modern QQQQQQ JJJJJJJJJJJ or anything that has a high quality RRRRR, you are out of CPU instantaneously. You can only run a few instances of them. It's really bad.
It's very difficult to get work done, and everything you do has to be AAAAAA'd down. Most stuff can never be realtime. It's like working in the 1970's with JJJJ machines, we simply can't compute more than a couple of quality YYYYYYs.
Running something like the NNNNNN YYYYYY will use a half a core. You run out of CPU immediately.
There will be a time eventually when Intel's basic level of computation power allows most YYYYYY's to work fine. That time is not here.
lol
rudy
[H]F Junkie
- Joined
- Apr 4, 2004
- Messages
- 8,704
IF the music player can do more with it that's fine. I think its a fallacy to think that software has hit a plateau. Take voice recognition, still not there, could still use more power. People will not deliver it because they are too concerned with battery life etc, people with weak systems etc... Large excel spread sheets still cripple a system.
Well that was convincing.
Tsumi
[H]F Junkie
- Joined
- Mar 18, 2010
- Messages
- 13,755
If you need more power, you pay for extra computing power in the form of Xeons. That is Intel's strategy, and it works. People that need the power pay the premium for the extra power. If you're not willing to pay for it, you don't really need it in Intel's eyes. Either way, you still end up buying their product, whether it's a higher tier one or lower tier.
Okay, there might be a few non-gaming software out there that can use more power and aren't written properly to take advantage of multiple cores. However, the vast majority of productivity programs that need the computing power are able to utilize most, if not all, the cores in a system.
Okay, there might be a few non-gaming software out there that can use more power and aren't written properly to take advantage of multiple cores. However, the vast majority of productivity programs that need the computing power are able to utilize most, if not all, the cores in a system.
defaultluser
[H]F Junkie
- Joined
- Jan 14, 2006
- Messages
- 14,398
You do know that you can buy computers with 2, 4 or even more sockets from various Intel OEMs? That would more than satisfy your demands for heavy computing, and Intel would get the large sums of money from YOU that should come with a top-performance workstation/server.
It does not sound to me like you are locked into some proprietary or outdated platform if you are using Intel CPUs, Given that, if you/your company cannot afford to purchase these monster machines, then obviously your time is not worth very much.
Quit bitching at Intel for charging top-dollar for top-performance. This is the way of business...after all, if money was no object to YOUR COMPANY, then you wouldn't have this performance issue, would you? Quit being cheap and whining about it.
EDIT: while we're on the subject, have you ever considered that your software might be unoptimized? Throwing hardware at a problem that could be helped via a new vendor or an internal software audit is just a fool's game.
That is what stuck out for me in that rambling post. It sounds like the software he's using is horribly optimized but he's blaming Intel for it.
There is no such thing as perfect optimization, unless you're using sculelos logic.
It's inherently parallelizable, could run on a supercomputer.
It's not perfect in every sense because latency increases by doing so.
Although that increased latency value is only a problem relative to networking technology.
EDIT:
FORGET THE SUPERCOMPUTER. I CANNOT AFFORD ONE.
I am doing regular work! This is not institutional stuff.
This is one person worth of work.
Just plain old normal stuff that we are in the dark ages for.
The Intel cpu's are an anachronism for my work, like I said.
It's not perfect in every sense because latency increases by doing so.
Although that increased latency value is only a problem relative to networking technology.
EDIT:
FORGET THE SUPERCOMPUTER. I CANNOT AFFORD ONE.
I am doing regular work! This is not institutional stuff.
This is one person worth of work.
Just plain old normal stuff that we are in the dark ages for.
The Intel cpu's are an anachronism for my work, like I said.
Last edited:
Tsumi
[H]F Junkie
- Joined
- Mar 18, 2010
- Messages
- 13,755
Again, your work apparently isn't worth getting 2P and 4P Xeons for, and that's your problem.
If you're confused about what Really wants to do, he just wants to do folding while he does CPU-intensive work:
http://hardforum.com/showpost.php?p=1039958978&postcount=4
http://hardforum.com/showpost.php?p=1039958978&postcount=4
Tsumi
[H]F Junkie
- Joined
- Mar 18, 2010
- Messages
- 13,755
Again, his problem that he isn't willing to pay for the performance...
defaultluser
[H]F Junkie
- Joined
- Jan 14, 2006
- Messages
- 14,398
Remember when I said earlier that software could still be your limiting factor? Let me give you a great example:
x264 (as well as it's predecessor, DivX)
Way back in the day the only way to get high quality in a small file size was to use two-pass encoding with a target bitrate, since the real-time optimizer was not up-to-snuff for single-pass.
But today you can use the fixed quality settings in x264 and see THE EXACT SAME QUALITY as two-pass, without the slowdown of the second pass. Also, from my experiments, I was able to cut the file size by half compared to optimized 2-pass encodes I made a few years ago! That's a HUGE improvement in throughput, and enough space savings to convince me to re-rip my collections!
So you'd think that they're done optimizing things, but you'd be wrong. I just upgraded my Handbrake to the latest version of x264 after a year, and saw a 20% performance increase!
The point is, just because you think the tool is "optimized" doesn't men it really is. This goes doubly true if you're not a software developer (which I get the feeling you are not), because you have no real understanding of how your tool actually works. If you want us to help with your performance issue, you're going to have to stop playing games and tell us what you're running.
wabbitseason
[H]ard|Gawd
- Joined
- Jun 16, 2010
- Messages
- 1,511
We, being a tiny fraction of Intel's customer base, can't do a thing. That's the sad reality. If you think you can influence their business decisions you're deluding yourself.
All you can do is look at what's out there and weigh a couple factors. How badly do I need Intel performance? Do I want to contribute to them shafting my demographic? Will I be their bitch and buy everything they sell every generation, knowing they're holding back on me?
Personally I'm seriously considering going AMD next round. For now I see zero reason to upgrade from my 920.
All you can do is look at what's out there and weigh a couple factors. How badly do I need Intel performance? Do I want to contribute to them shafting my demographic? Will I be their bitch and buy everything they sell every generation, knowing they're holding back on me?
Personally I'm seriously considering going AMD next round. For now I see zero reason to upgrade from my 920.
Why? AMD has been phoning it in even worse than Intel. Though, to give them credit, at least they came out publicly and acknowledged that they have no interest in chasing the enthusiast market.
I wrote the program. It's opensource.
That's part of the reason why I'm so damn broke.
It's got linear scaling across cores.
What more do you want from me... exponential scaling?
This is like trying to argue with some guy in the 1980's..
that the 286 processor is not fast enough for 3D Gaming.
And he keeps saying:
THE 286 IS FAST ENOUGH!
YOUR SOFTWARE MUST BE UNOPTIMIZED.
CHECK WITH YOUR VENDOR.
YOUR SOFTWARE MUST BE UNOPTIMIZED.
CHECK WITH YOUR VENDOR.
defaultluser
[H]F Junkie
- Joined
- Jan 14, 2006
- Messages
- 14,398
I want BETTER than linear scaling.
If you're using Intel, see about optimizing for Hyperthreading. In certain workloads, it can yield improvements of 15-25%, and for most developers it's the LAST thing they bother optimizing for, so it's a good bet you haven't yet!
Also, since you won't provide me with your machine's processor model number, I have to assume from this thread:
http://hardforum.com/showpost.php?p=1039958978&postcount=4
That you're still using a 1366 chip that does not support AVX or AVX2. If your code is highly vectorizable, you're costing yourself performance by not upgrading to SB-E (or IB-E later this year).
I don't go out of my way to optimize for hyperthreading.
It's plain multithreaded code.
It benefits from HT without me having to do anything.
Somebody else optimizing something else may reasonably query for real core count vs. virtual core count in order to decide upon thread granularity.
But I can't even do that, because the yyyyyy's need a thread anyways.
So I let them use a thread.
It's plain multithreaded code.
It benefits from HT without me having to do anything.
Somebody else optimizing something else may reasonably query for real core count vs. virtual core count in order to decide upon thread granularity.
But I can't even do that, because the yyyyyy's need a thread anyways.
So I let them use a thread.
So basically in your OP you're bitching about Intel because their hardware is supposedly too slow for you. Someone then posts a suggestion on how you could possibly squeeze out more performance from your current hardware, but your reply is essentially "Nope, not going to do it, my code is fine"?
Last edited:
So basically in your OP you're bitching about Intel because their hardware is supposedly too slow for you. Someone then posts a suggestion on how you could possibly squeeze out more performance from your current hardware, but your reply is essentially "Nope, not going to do it, my code is fine"?
Sorry, but you have 0 information on the topic, so it's not intelligent for you to make so many assumptions in a post, when you could easily have no idea what's actually going on. And then rely on your assumptions which are formed upon 0 information to make aggressive statements and "so basicallys" and frownie faces. That's just silly. Sorry bud.
The yyyyyys are 3rd-party modules written by other people all around the world.
I can't modify their code to try to "make their algorithms utilize hyperthreading".
It's not realistic or even possible, it doesn't even make any sense.
They get a thread. And I get the benefits of HT automatically. End of story.
lol
Last edited:
Fenris_Ulf
[H]ard|Gawd
- Joined
- Nov 29, 2004
- Messages
- 1,907
Wow. You "work 16 hours a day" and your wife left you for a Bitcoin miner? And you need more CPU power for....Folding? I think you need to take two steps back and re-evaluate your life rather than ranting about Intel. Besides, Intel has a legal duty to its stockholders to maximize returns. If you want to change Intel, buy enough stock to become a majority shareholder.
Yeah, it's totally my fault that you said you wrote the program and didn't mention that part of it consisted of 3rd party modules. "lol" indeed.
livefastdieyoung
Limp Gawd
- Joined
- Sep 22, 2012
- Messages
- 307
Droogie128
Weaksauce
- Joined
- Oct 20, 2012
- Messages
- 119
Yeah. Every time I read through a post... it usually ends with "Wait..... what?"