maleko48

Did you try changing that setting in ThrottleStop yet?  Did you try HWiNFO yet to monitor for this? 
 
Intel calls this cTDP or configurable TDP.  It allows individual manufacturers to easily lower the TDP power limit below the 45 Watt rated limit.  When pressed hard, the CPU will throttle back its speed so it does not exceed this lower TDP value.  This throttling is power related, not temperature related.  
 
Your screenshot also shows that Asus lowered the thermal throttling temperature from the Intel rated 100°C to 88°C so I will not be surprised if they also lowered the power limit.  As a consumer, sometimes problems like this are impossible to solve.  The only solution is to put the laptop back in the box and tell Asus that they can have it back.  Its performance does not live up to its appearance.  

In the Turbo Power Limits window, it shows that TDP Level Control is checked and this is set to 1.  Try setting that to 0.  I think HWiNFO can show what TDP Level the CPU is currently running in.  TDP Level 1 is the low power level so you definitely do not want to be in that.  This setting can be set in multiple locations within the CPU.  ThrottleStop does not have access to all of these locations so you might be screwed.  
 
When Speed Shift is enabled, there is no need to use the Set Multiplier or SpeedStep options.  Both of these are ignored when Speed Shift is enabled. 
 
There is also no need to check the Clock Modulation setting unless you see a number less than 100.0 in the Mod column or in the TS Log File.  I do not think this throttling method is being used anymore.  Intel has created new power based throttling methods that laptop manufacturers love to use.  An 8750H is a wonderful CPU in theory but when limited to only 25 Watts, it is kind of a joke.

So I was looking at the new roadmap for AMD that was released when Ryzen 2000 was announced.
 
Does anyone else find it very odd that AMD is actually jumping ahead of Intel next year, bypassing 10nm and going straight to 7nm? Especially giving the impending end of silicon as we know it?
 
It would just make a whole lot more sense to try 10nm, then 7nm, then 7nm+ to extend silicon life out another year, don't you think?
 
Its not like these are just rumors either. In January 2018, AMD announced that the 7nm-based Zen 2 chips were at the "design completed" phase. Pretty sure AMD wouldn't be saying this if it were not true.
 
More worryingly, it looks like mobile phone advancements will be hitting a brick wall even sooner, seeing as the new Qualcomm SnapDragon 845 SOC already uses a CPU that is based on the 10nm manufacturing process.

Depends if your drive is MBR or GPT as to the specifics, but I triple boot Win7/Win10/Ubuntu no problem from a single SSD. I would recommend creating a second EFI partition (assuming you're booting UEFI) for your Linux install to prevent it screwing up Windows EFI. 

You will almost certainly benefit as long as your cooling system isn't already thermally soaked and maxed out. It really depends how many RPMs of fan you are willing to listen to since cranking up your RPMs will reduce the heat soak a bit, but there's only so much heat that can be dissipated from a given mass and surface area with a set fan rpm. Your entire system will be more thermally responsive, for better or worse. (Generally better in the world of desktops.)

EASY AS - Many here on LTT have probably done a few already like I have.
i7 3770 4Cores8Threads @3.4Ghz Base, 3.9Ghz Turbo, 3.7Ghz Typical
16GB DDR3 Ram (1600Mhz - Chipset Limited)
4GB MSI 1050Ti Low Profile (70-75w Usage MAX - No 6Pin Pwr) +0Mhz Core +500Mhz Vram in some Testing,
Sata3 128GB SSD & 1TB HDD
Adding vent holes in side panel drops 3-5*c.
260w PSU - Est Usage 150-200w /Depending on how much you squeeze out of both CPU/GPU at the same time.
SFF <- Important - SMALL FORM fACTOR STYLE SIZING
SFF sized 1050Ti (expensive now, prices should come down soon hopefully) will fit in those, performs 2% slower,.... IF AT ALL, plus Turboboost is always running it higher and lower.
Playlist of Titles Running - https://www.youtube.com/playlist?list=PL5ofzFlQXfuFEXJINN914cgz93B4qquk-
 
Way ample for many to game on.
 

I'm sorry but this is really not an efficient way of doing it, or practical. You are bringing the voltage down to step it back up again to bring it down again. Each one of these changes will result in losses.
 
My suggestion would be ditch the inverter, and monitor if possible. Try and see if you can find a regulator that matches your laptops input voltage and feed right into that. This will be the most efficient way of transferring the power.

You just save the other image outside the regular jpeg structure. It's how HTC were able to do depth effects with the M8. It saved two images within the same jpeg file (using a proprietary extension of jpeg) and then used those two to do black magic.
 
Regular image viewers will just read the regular jpeg data, and then skip everything else. That's why the pictures taken on the M8 camera displayed just fine in regular image viewers, but when they were opened with HTC's viewer it could read the extra info and use that for editing purposes. Here is a short discussion about it.
 
You can do this with a lot of file formats.
 
 
Take this image as an example.
Download it, open it in a text editor and then scroll to the bottom.

The purpose of this topic is to discuss the implications of child porn being hidden in the bitcoin blockchain, not to debate the merits of cryptocurrencies themselves. Please stay on topic, otherwise your posts may be removed or the topic may be locked.

Testing on a real PC wouldn't show if it's been flashed to show it's something it is not.
You'd have to run something like GPU-Z and compare the results to a known real one to see the difference.
There is a YouTube video that might help on this:
 

GODDAMNIT beat me to it 

Here's the kicker, though:
Article Link: https://arstechnica.com/gaming/2018/03/star-wars-demo-shows-off-just-how-great-real-time-raytracing-can-look/
 
Info on Nvidia DGX-1: http://www.nvidia.com/en-us/data-center/dgx-1/
 
Here's the direct link to the YouTube video of the graphics demo (watch in fullscreen at max res): 
 
IMO, this looks freaking amazing--almost looks as if it's an actual filmed scene from a Star Wars movie. We kind of got a taste of it, aesthetically, with the Star Wars Battlefront photogrammetry stuff, but this is on a different plane.
 
Update: Nvidia reached out to Ars Technica to correct a detail about the rendering engine:
 
 

I want to know how to remove small holes in the wall caused by drilling the wall.

Why is this on a tech forum?

What is it that you are trying to change about it?
Let me elaborate on my earlier snippet. I've removed the redundant double print statemement. Take a look at this example (code below as well): https://repl.it/repls/BleakGrizzledPolygon
#Show the values that will be iterated over. list(...) is needed since in Python3 range is a generator, like xrange in Python2. i = list(range(0, 5)) print(i) # Loop over them. for i in range(0, 5): print('Iteration ', i) if i == 3: # Change i to hold the value 5 temporarily. i = 5 print('The value of i will now be', i, 'for the rest of the loop.') print('The value of i is', i, '\n')  
The range function will, as its name implies, generate a range of numbers between a given start and end. By definition the value i will take the next value in the range at the start of each iteration. Once you assign a new value to it, it will keep that value for the rest of the loop, or until it is changed again. Once the next iteration starts however, i will be (re)set to its intended value: the next one in the range.
 
This may help us give more concrete help.

Threads are not the same as running on multiple cores. Your are confusing multi threading and multi processing. Both of which are relatively easy to do in python.
 
Multiprocessing in python (running on multiple cores) is as simple as having a list of work and using the multiprocessing module to make a pool of cores and then map the list to a function to process.

from multiprocessing import Pool   def f(x):     return x*x   mylist = [1,2,3,4,5] pool = Pool(2) pool.map(f, myList)
 
This will square the list across 2 cpu cores. If you need to access the result you can do a for on the map
 
For result in pool.map():
    print(result)
 
There is a downside and that's each core gets a copy of the list this with large lists can be very memory intensive.
 
You can somewhat get around this by using generators but that's another topic for another day.

Yes I understand how Quantum physics work I just didnt feel the need to get into details with people from this community since most are already aware:
 
(Basics video for those unaware)
 

Yeah haha I haven't forumed in ages, me did forgot how done the internet.

@straight_stewie
 
110% agree ^, you took the words straight out of my brain like one of those damn brain suckers lol.

I like where your head is at here. I'm not sure how to say what I'm thinking, but I think it goes hand in hand with your line of thinking so bear with me for a minute here, but....
 
A substantial challenge we face is actually finding a parallel computing hardware architecture / model that compliments parallel programming algorithms. As it stands right now we have far more computing resources than we make use of most of the time. Additionally, threaded multicore programming is still pretty young on the whole and somewhat tricky or difficult to maximize with current system architectures in terms of keeping things properly synced with consistent, minimal latency, bugs, etc.
 
I agree that the right hardware architecture could really unlock the ability to use our CPUs to their fullest, or possibly require a lesser CPU (in terms of power/heat offput) to easily run the same code as yesterday's processing systems. It's not all about raw specs and throughput, but finding models that compliment each other and handle processing with more finesse than their predecessors.

I like where your head is at here. I'm not sure how to say what I'm thinking, but I think it goes hand in hand with your line of thinking so bear with me for a minute here, but....
 
A substantial challenge we face is actually finding a parallel computing hardware architecture / model that compliments parallel programming algorithms. As it stands right now we have far more computing resources than we make use of most of the time. Additionally, threaded multicore programming is still pretty young on the whole and somewhat tricky or difficult to maximize with current system architectures in terms of keeping things properly synced with consistent, minimal latency, bugs, etc.
 
I agree that the right hardware architecture could really unlock the ability to use our CPUs to their fullest, or possibly require a lesser CPU (in terms of power/heat offput) to easily run the same code as yesterday's processing systems. It's not all about raw specs and throughput, but finding models that compliment each other and handle processing with more finesse than their predecessors.

Another very good idea but dude seriously next time snip out most of the post when its like 20 paragraphs long lol

Throwing transistors at a problem isn't the only solution. The organization of those transistors matters.

There is a new-ish technology called Network on Chip. I have already written and tested in FPGA a makeshift parallel processor that uses a single instruction stream and can pass instructions to one of many ALUs depending on the busyness of each individual ALU. This allows the programmer to think about a program as being executed sequentially, when it is in fact, being executed in parallel. This could easily be modified to take advantage of SIMD type instructions. This is a completely new way to think about vector processing units (which is what GPUs are).

The above example would provide a fundamental change to the way that GPUs are organized, programmed, and reasoned about. 

I agree. The other consideration is that general purpose processors will start adding specialized components to handle increasingly rarer tasks. 

But yes, certain components are not fully fleshed out yet. We aren't even done reasoning about how ISA's should work or what a good one is. We haven't solved any of the major DRAM memory issues, there is still much thought to be given to how to best make a GPU... I agree fully that there is still a lot of area for improvement in computer architecture, regardless of whether or not components can continue to get smaller.