>computers in 2015 are barely faster than computers from 2007
what the fuck happened? computers now should be like 100 times faster man
>>51658842
Consoles happened.
>falling for the Moore's law meme
Bullshit they are barely faster. You're probably just comparing cpu clock speeds and ignoring things like architecture changes.
http://www.tomshardware.com/reviews/500-gaming-machine-2007-edition,1681-14.html
>$500 machines running Quake 4 for at a shitty resolution
>only cracks 50 fps by a little
http://www.computerworld.com/article/2538159/mobile-wireless/top-10-cell-phone-favorites-for-2007.html
>their #1 phone has a 400 MHz CPU and 256 MB of space
But you know. They're only faster by a little bit
>>51658842
>what the fuck happened?
All sorts of idiotic system shit by Mickeysoft. For "user friendliness."
And not only that, but coders not capable of writing anything other than javascript for "apps."
Shit's going to hell in a handbasket.
Samsung is releasing a 128GB ram stick "real soon now." After about a year and price drops so that everyone gets one, Microsoft Office will occupy 90 percent of that.
Yes, I'm cynical.
>>51658939
>a blog as knowledgable at making computers as my dick is knowledgable about vaginas
>a clickbait site about phones being able to run games at 1/3 the framerate if all settings are put to minimum and additional details are removed via ad-hoc hacks
Great sources champ, you totally showed everyone!
Wirth's Law
>>51658978
Almost as good as you baseless claim, right?
>They're only marginally faster because I said so!
>>51658842
Laws of physics. There are limits to everything and we are quite near them.
>>51658995
If you don't know shit about tech, you're free to visit >>>/trash/
>>51659015
Wow you sure did bring in the evidence with that post. You can probably get it published in a journal.
Hardware gets more powerful by us engineers.
Then you developer faggots use that hardware to put out bloated, horribly optimized shit, because you can kind of get away with it,
Because we've gone down the Moar Cores path for the most part with most improvements being in the mobile device field. Low-power CPUs for laptops and obviously ARM stuff in phones has moved along at a tremendous rate since 2007.
Desktop PCs are the ones stagnating (and in some cases going backwards) and there have been barely any breakthroughs or major changes since 2009 or so.
>>51658842
Please try running 8k 120fps and Farcry 4 on a 2007 PC u faggot
Because """""""""""""""""""""""""""""""programmers""""""""""""""""""""""""""""""" look at hardware improvement and see a free pass to write shittier code.
>>51659601
I don't see this as a problem. More like an opportunity.
Serious question.. but what is even the endgame of computers for relative "normie" people. Seems like a lot of features now are just gimmicky flashy things that look cool but aren't really all that useful. At what point is a computer fast enough.. because higher end laptops/phones/computers in general seem to be pretty decent nowadays.
>>51659658
The question was why things aren't running faster, and that's the reason.
Also, generally, improvements have been adding more cores. Most people aren't good enough to write multithreaded programs that actually run faster than a singlethreaded equivalent, so the power ends up being used solely to run more programs at once.
>>51659601
>having to explain variable packing to a "CS graduate"
>>51659688
But anon, that's not the reason why things aren't running faster. The reason is physics.
What you rant and flame about is software. What makes you so butthurt about modern devs anyway?
>>51659748
Computers right now are way faster than computers in 2007. See >>51658891.
I'm saying that it's still hard to tell, because of a general decline in software quality.
>>51659767
They have been getting marginally slower because of diminishing returns inherent to CPU design and manufacturing
>>51659767
speed != throughput
>>51658842
Plateauing due to the limitations of silicon transistors?
>>51659789
That's true, but it's not what the OP stated.
>>51659799
Yes, that's what I said.
>>51659816
You said and I quote:
>Computers right now are way faster
And that's not true, because:
>>51659799
Get your terminology straight lad.
>>51659789
>They have been getting marginally slower
No, they're getting marginally FASTER, the issue is the word "marginally". You already correctly pointed out the diminishing returns in CPU manufacturing and design.
>>51658891
>>51659767
>>51659799
>>51659789
They aren't barely faster, it's something like 2-5 times, but compare a computer from 1999 to a computer from 2007, and then compare that computer from 2007 to a modern day one.
It's slowing down.
>>51659856
He meant that they were getting faster slower.
>>51659467
>>51659688
>>51659767
>>51659789
>>51659799
>>51659814
>>51659856
http://www.anandtech.com/bench/product/49?vs=1500
>>51659870
GPUs matter a fuckload more than CPUs.
>>51658891
>>51658995
>>51659767
Eight years separate the i7 4970k from the Core 2 Quad.
Eight years separate the Core 2 Quad from the Pentium II.
Are those really similar jumps in performance?
>>51659856
Oh yeah I had a major brainfart, I meannt to say that computers have been getting only marginally faster in recent years
>>51659870
And what is your fucking point?
http://www.anandtech.com/bench/product/288?vs=1544
Do you not understand what the words "marginally" and "diminishing returns" mean?
>>51659885
How big was the jump from p2 to c2q? 1000x?
>>51659601
But it's more like then, it was worth making the code harder to understand in exchange for a performance boost, now everyone is the top middle dev.
>>51659880
http://www.anandtech.com/bench/product/521?vs=555
http://www.anandtech.com/bench/product/1494?vs=1447
>>51659902
I tried look that up, and found this.
http://cpuboss.com/cpus/Intel-Pentium-II-300PE-vs-Intel-Core2-Quad-Q6600
Now I can't stop laughing.
Completely useless.
>>51659902
Oh I get it, you're just salty that clockspeeds haven't been increasing and improvements are reliant on architectural improvements.
>>51659931
hue
bet you if you clocked a C2Q to 300mhz and limited it to one core, it'd be a similar number as C2Q to Skylake. Remember Pentium 4 1st gen was actually a decrease in IPC (by alot) from Pentium 3; and the second gen was even worse.
>>51659902
Quickest, most general benchmark I can think of:
Pentium II 450mhz: 0.45 GFLOPS
Core 2 Quad 3ghz: 48 GFLOPS
i7-4790K 4ghz: 107 GFLOPS
>performance
>benchamark
>other things that are not clockspeed
FFS anons, get your shit together and try to stay on topic.
Pentium 2 / 233 MHz
Core 2 Quad / 2.4 GHz / speed improved 10x
i7 / 4GHz / speed improved less than 2x
Hence OPs deep concern.
>>51659992
Shut the fuck up and stop cluttering the thread with your weak bait. We all know you could push a Pentium 4 near 5ghz and it'd still suck dick by modern standards.
big companies are purposefully holding back technology in order to make more money
>>51659917
So around 2-3x slower for GPUs, 2x slower for CPUs. Whew!
>>51660006
You suck dicks by all the standard dipshit.
>>51659980
This means both perspectives are true because total processing power increased by ~50GFLOPS in the first eight years and ~50GLOPS in the following eight years, but the first jump was a 100-fold increase whereas the second jump barely doubled it. And if it only goes up to 150GFLOPS in 2023 that'll be embarrassingly depressing.
>>51659006
this anon gets it
it's all downhill from here
>>51660027
I've been thinking this is what's going to happen. Computers were only getting exponentially more powerful because it was exponentially easier to improve them.
>>51658842
Are you retarded?
>>51660063
We aren't anywhere near the physical limit of processing power. What's holding us back is that we are near the limit of human resources to develop better processors.
>>51660086
Don't know if he is, but you just proved you are.
>>51659006
>>51660063
BULLSHIT
FUCK OFF
I BET THERE'S BEEN BITTER FUCKS LIKE YOURSELVES ALL THROUGHOUT HISTORY
>>51660063
Thanks. I'm still eagerly waiting for someone to post about how the nano-world and thermodynamics ... oh shi
>>51660097
Class A debil.
>>51660099
Sick m8. Did your intuition struggle to find out that this is the development of top supercomputers, and now you are throwing that sick banter because of your lack of response?
>>51659980
0.45
48 / (4 cores * 300/45) = 1.8
107 / (4 cores * 1.2~ (HT) * 400/45) = 2.5
also thats float; integer is a better metric because less sneaky SIMD tricksies
>>51659931
>>51659902
Ok, could only find a some p3s on here, but they seem to get a score of 300.
Core2Quads seem to go between 3-4k.
So over 10 times increase.
The 5775 is 11k.
3-4 times boost.
While I get what you're saying, the p4 is an anomaly, clock per clock, single threaded performance has always steadily gone up.
It still is though, and I wonder if that rate has changed.
However, the more important thing, is that regardless of how that was achieved, with more cores, or with better single threaded performance, If you want from a computer with a P3, to a Core2Quad, it was a 10 times leap, now it's a 3-4 for a longer time span.
Even the later P4s did get faster with raw clock speed. And that's really all that matters, if a processor can have 4 times the clock, but half the IPC, it's probably going to perform a lot better.
The moar cores argument is valid though, as it won't always help, but will be really strong in benchmarks.
Still, it's part of the boost in power we go, and 2 cores, at the very least, is a huge boost.
4 starts getting into diminishing returns/not enough paralellism.
You're right though, most of the reason things seem to be slowing down is we aren't getting the more cores boost anymore.
It'd be interesting if someone actually compiled single threaded performance increase over time, and checked if that has actually been changing, since it actually looks pretty constant.
>>51660125
>>51660122
I don't think it's fair to try to equalize for clock speed, the clock speed going up is part of the equation here.
Multicore sure, those cores are often not used, but the clock won't ever clock that far down.
I'm trying to look at this from a perspective of a user, what raw power increase am I getting, + what raw power increase will I notice.
>>51658842
Go to the daily programming thread and see who is writing your software.
Software is usually running about 1/10th the speed it can at least.
>>51660027
This pretty much explains why a 8-year-old computer can continue to be perfectly adequate nowadays, whereas in 2007 a computer from 1999 would be more or less useless for any modern applications.
In a way this isn't so bad. All the diehards with their "muh last of the Thinkpads" T60 units can still shitpost in modern society.
>>51660103
it's ok, we we're all in denial at first
maybe lecture about exponential growth will cheer you up
https://www.youtube.com/watch?v=DZCm2QQZVYk
>>51660157
Well frequencies aren't going to increase by and large, so it's more realistic to look at architectural improvements going forwards.
hence why tracking if the performance is from the architecture or the process
They are WAY more power efficient, which is good
Why do people always neglect that fact?
>>51660200
Because the average person just wants more power.
>Fuck the whales, I have lag
>>51660192
But performance isn't a direct function of clock speed at all. It's highly architecture dependent but also depends on CPU-specific optimizations and other factors.
>>51659902
>1000x
probably
>>51660200
power efficiency isn't going to make my physics simulations faster
>>51660192
Where the performance comes from is irrelevant though, I'm trying to look at single threaded performance over time, which clock speeds are very much a part of.
The multicore was a one time boost, and doesn't effect many things, so that's fine to factor out of the increase in power, but clock speed has historically been a major factor, and does effect everything. It might have hit a wall, but even in modern processors, it can vary from 1-3 ghz, which is a fairly significant factor.
The overall metric I'm using here is how fast it can complete a single threaded program, which needs to factor in clock speed.
>>51660230
Exactly
hence equalization for clocks
>>51660236
already posted multiple times in the thread, it was 10x
1000x increase would be insane, it would be over 100x current processors
>>51660086
what part of the concept of exponential growth don't you get?
>>51660267
That graph is log scale.
>>51660250
I don't think it's irrelevant. you could do parallel processing with multiple P4s but that doesn't mean they were faster individually. shoving another processor onto the same die is the same thing.
![Dell_XPS_Gen4[2].jpg](https://i.imgur.com/GpGZjWz.jpg "Dell_XPS_Gen4[2].jpg")
>>51658842
>what the fuck happened?
Phones and tablets happened. The big focus the last ten years has been the attempt to put as much power as a desktop had in 2005 into a phone in 2015. And they succeeded.
The advancement of the desktop has been irrelevant. A high-spec computer from 2005 will still browse the internet, stream netflix, and play light video games just as well as a low-spec computer from 2015. The needs of desktop computing are no longer a driving force for advancement.
>>51660290
it also has no source and is contradictory to everything else in this thread, which actually have sources
>>51660304
I think it's also that we hit a performance level where most things that people would want to do can be done. Maybe once VR becomes mainstream we will see some real pushes for more power, not just as a selling point for having a bigger number than your competitor.
>>51660309
Holy shit! Grasping for straws much?
http://www.hpcwire.com/2014/06/23/breaking-detailed-results-top-500-fastest-supercomputers-list/
>>51660295
Yes, but as a consumer, it was impractical to do that.
I'm just trying to look at what you get when you buy a high end processor, in terms of power, over time.
Processor speed has always normally increased from improving clock speed and raising IPC, we've hit the limit for one more or less, so if single thread performance is going to keep increasing, the other has to pick up the slack.
Multicore also has the issue of being hard to use.
Single thread helps everything, and it's both clock speed, and IPC that got us processors as powerful as they are today.
What we're talking about in this thread, is the power you had when you built a new desktop PC, so the IPC should be considered with clock speed, because without it, it'd seem like processors only improved a couple times over, which is far from true.
>>51660304
>The needs of desktop computing are no longer a driving force for advancement.
Put another way, there's only so much the average desktop user needs to get by doing what the average user does without noticing considerable loss of productivity...
I know this is just regurgitating what you said, just thought it might be clearer to other anons ;)
>>51660338
>supercomputers
thanks for invalidating yourself.
>>51660336
Right. Although I believe 4k video will also be contributing. You can't get away with it on a machine a decade old.
However, I wonder: can you still stream HD netflix on a Pentium 4? I was doing it without issue up until 2013. Nothing should have changed, but it often does with this sort of thing...
>>51660369
How are supercomputers invalid?
Believe it or not, competition actually slows down the progression of technology. It causes people to separate resources and produce forks of similar technology just to compete with someone doing the same thing. There is no better example of this than GPUs and CPUs, as there are only 2 at the top of each and detrimental business practices are more evident.
Yes, cooperation also has its downsides. Why would companies charge a fair price for what their product is worth rather than increase the prices because nobody will undercut them? Well, that actually brings us back to where we began with people trying to maximize profits rather than create great products for us and future generations.
The answer? Don't be a menace.
![picture[1].jpg](https://i.imgur.com/VLGxRpg.jpg "picture[1].jpg")
>what the fuck happened?
Lack of interest. The high tide mark for desktop computers was reached with the AMD Athlon 64. That was the peak of the generation of young men who knew their PCs from top to bottom AND had a significant percentage of the population.
Find any dude around 30 years old and ask him what each piece of the inside of a computer does. He'll be able to tell every single one with out hesitation. He'll talk about how he used to overclock.
Ask the same thing of a 20 year old...
>>51660396
literally just MOAR CORES
you can always make a supercomputer faster by buying more processors, also nobody here has a supercomputer. OP stated "computerS", which would imply mass produced computers. (I know you could interpret it differently, but it's obviously not...)
>>51660459
True, but that's in an ideal world, where even without competition, companies are motivated to improve their product.
Even if work is duplicated, competition motivates companies to do more research to stay ahead of your competitor.
>>51660459
Except it doesn't. All of the investments in semiconductors are R&D. Production margins are minimal.
The ability of Qualcomm to be better than Intel in the mobile industri is key to their survival, so they have to keep ahead in performance and performance per watt.
>>51660460
Le wrong generation!
>>51660497
>Le wrong generation!
I was there, bud.
>>51660468
Except that it's not how it works. You're just assuming that the development of supercomputers can be done in quantity.
>>51660103
>>51660097
>>51660107
There's a hard limit at around 10 nm with CMOS processing at which point quantum tunneling through the gate oxide dissipates too much power for continued scaling to be effective. We are fast approaching this point (currently 14 nm), but there is also interesting research into further areas beyond CMOS technology.
https://en.wikipedia.org/wiki/Beyond_CMOS
