| Forum Index > Tech Support |
|
Cyro
United Kingdom20321 Posts
There might be a 720p video that needs ~2mbits to look good and another one that needs ~8mbits to look good. Youtube would just assign ~3mbit for example to both of them. Meanwhile, just because another video is 1080 height, it might get 3x the bitrate allocation without other consideration for what the video contains (so a 720p video uploaded as 1080p with huge black bars could be way higher quality than one just uploaded as normal 1280x720, for example) | ||
|
techjunkie
13 Posts
Power consumption for total system on the gtx 680 under load is around 362W AnandTech Power consumption for my current card, the 4890 seems to be around 287W. I use a 430W PSU.AnandTech 4890 I googled for the power consumption on a few higher end cards and the consensus seems to be that to power cards like the 280x you need 550-600W. Can someone explain why benchmarks for total system power (under load) indicate 360W in furmark or 290W in Metro: Last light, yet people recommend 550W? Why wouldn't 430W be enough? | ||
|
Craton
United States17266 Posts
I would be a little leery about a 430W in a system consuming 360W under heavy load, since I think you might run into some of those issues. 550W is overkill, though. | ||
|
Cyro
United Kingdom20321 Posts
*Edit: A lot of that stuff applies to GPU boost 2.0 introduced ~16 months ago.. I'm not sure if all of it does apply to 1.0 which was launched early 2012 Power consumption for total system on the gtx 680 under load is around 362W AnandTech 362w from the wall, from an 80+ silver PSU. Efficiency guaranteed 85%, maybe 90 - if it's 90% efficient, then 362w from the wall means ~325w from the PSU, which is the number that it's rated for. That's with an overclocked 32nm sandy bridge 6-core CPU system. Modern, good PSU's rated at 450w are designed to take it 24/7. With ambient temperatures or 40c, sometimes even 50c - if you have them in a room with fresh 20c air and ventilation, they are not stressed anywhere near as much. Power supply recommendations generally take into account peak rated PSU's (which are usually terrible and there's no reason to buy one today) and trash tier stuff because a huge portion of PC users don't know even the basics of power supplies or how much power their parts consume (understandable because there's very little good education on that unless you go look it up and magically know good sources) - at best they give you a vague idea of what to buy, at worst they are just a huge waste of money. Modern good quality PSU's also don't really have significant losses to their ability to output power with age, not even remotely close to comparable to some old/bad ones, anyway | ||
|
techjunkie
13 Posts
On November 18 2014 14:15 Cyro wrote: Furmark or similar readings should always be ignored these days, for the last 3 years or so both sides throttle GPU's with those types of loads for a few reasons, one of which being that power usage was way higher than they were designed for and they would either have to ship the GPU's with like 70% of the performance in games and be really overbuilt, or they would be at risk of blowing up under furmark-style loads. The better solution is just to dynamically adjust performance with power draw, as has been done recently - you can't make a 680 on stock consume more than 170w for the card, for example. It'll throttle itself to death before it happens. Nvidia has gone so far as to include the VRAM power AND THE POWER FOR THE FANS ON THE GRAPHICS CARD (i just find that a little silly) to those GPU power calculations and build in extremely good dynamic throttling/boosting for dealing with power usage, ceilings etc *Edit: A lot of that stuff applies to GPU boost 2.0 introduced ~16 months ago.. I'm not sure if all of it does apply to 1.0 which was launched early 2012 362w from the wall, from an 80+ silver PSU. Efficiency guaranteed 85%, maybe 90 - if it's 90% efficient, then 362w from the wall means ~325w from the PSU, which is the number that it's rated for. That's with an overclocked 32nm sandy bridge 6-core CPU system. Modern, good PSU's rated at 450w are designed to take it 24/7. With ambient temperatures or 40c, sometimes even 50c - if you have them in a room with fresh 20c air and ventilation, they are not stressed anywhere near as much. Power supply recommendations generally take into account peak rated PSU's (which are usually terrible and there's no reason to buy one today) and trash tier stuff because a huge portion of PC users don't know even the basics of power supplies or how much power their parts consume (understandable because there's very little good education on that unless you go look it up and magically know good sources) - at best they give you a vague idea of what to buy, at worst they are just a huge waste of money. Modern good quality PSU's also don't really have significant losses to their ability to output power with age, not even remotely close to comparable to some old/bad ones, anyway While I appreciate the history lessons and the large amount of details in your responses, I don't feel like I'm tech savvy enough to pull a clear conclusion from this. I would appreciate it if you would make it clearer whether or not you think 450W is enough to run high end cards? For instance when you say that 362w is being pulled from the wall so 325w from the PSU, do you mean the power supply is pulling a certain amount but a lower amount is actually being used by the system? Is the PSU rated for 325w? If so, where does that number come from? | ||
|
Cyro
United Kingdom20321 Posts
 Simple questions simple answers sdtksghjs;hyjsighsThe basic job of the PSU is to take in mains power (which is ~240v here) and turn it into 12v, 5v and 3.3v power for the system. Some power is lost in that process because it's not 100% efficient (efficiency ratings on PSU's are a pretty big thing) so the amount of power going into the PSU from the wall and the amount of power being output by the PSU are two different things. If a PSU is 90% efficient and the parts require 400w, that means it requires ~1.11x more power than 400w in order to output 400w (so it would draw ~444w from the wall) - that extra 44w is lost as heat. the PSU power rating (450w for example) is used to rate the output power number, which is after the efficiency losses on the power that's drawn from the wall, so it's a lower number because of that, if you want to max a 450w, 90% efficient PSU, power draw from the wall has to be 500w - because 500w goes in, but only 90% (450w) of that comes out of the PSU as 12v/5v/3.3v - so a number like 360w from the wall isn't very much at all. In that case it was even with a very high power draw CPU on the test rig where the numbers were taken, so they are inflated compared to most systems A modern system like i5 and a 970, power consumption would be somewhere around 250w at stock. For something like an fx9590 and 290, because of the vastly higher power consuming and factory overclocked CPU, then 450w isn't very wise, but it would probably work ok if you didn't OC the GPU etc. | ||
|
techjunkie
13 Posts
On November 18 2014 14:45 Cyro wrote: Sorry that's kinda my fault Simple questions simple answers sdtksghjs;hyjsighsIf a PSU is 90% efficient and the parts require 400w, that means it requires ~1.11x more power than 400w in order to output 400w (so it would draw ~444w from the wall) - that extra 44w is lost as heat because it's only 90% efficient, not 100% the PSU power rating (450w for example) cares about the output number (400w, what the parts are drawing from the psu) and not the input (444w, what the PSU requires from the wall in order to supply 400w to the parts) because of that, if you want to max a 450w, 90% efficient PSU, power draw from the wall has to be 500w - because 500w goes in, but only 90% (450w) of that comes out of the PSU as 12v/5v/3.3v - so a number like 360w from the wall isn't very much at all. In that case it was even with a very high power draw CPU on the test rig where the numbers were taken, so they are inflated compared to most systems A modern system like i5 and a 970, power consumption would be somewhere around 250w at stock. For something like an fx9590 and 290, because of the vastly higher power consuming and factory overclocked CPU, then 450w isn't very wise, but it would probably work ok if you didn't OC the GPU etc. Ok, if I understand correctly, my ~80% efficient Corsair CX430M is drawing at most 516W from the wall. So if I look at benchmarks and I see a number like 362W, the power supply is using even less then that because that is what is coming from the wall and not what my parts are drawing from the power supply itself? So if I see 362W on benchmarks, the power being pulled from my PSU to power my machine is actually something like 290W? That would mean for single cards and parts at stock 430W would be plenty! | ||
|
Cyro
United Kingdom20321 Posts
On November 18 2014 15:04 techjunkie wrote: Ok, if I understand correctly, my ~80% efficient Corsair CX430M is drawing at most 516W from the wall. So if I look at benchmarks and I see a number like 362W, the power supply is using even less then that because that is what is coming from the wall and not what my parts are drawing from the power supply itself? So if I see 362W on benchmarks, the power being pulled from my PSU to power my machine is actually something like 290W? That would mean for single cards and parts at stock 430W would be plenty! Yea, if the same PSU is used. A wall power reading of 300w means 270w used on a 90% efficient PSU, but 240w used on an 80% efficient one. You have to look up efficiency curves of the PSU's used to really know, but the 80+, 80+ bronze, silver, gold, platinum, titanium(?) are good basic guidelines for that number if you know what they mean For example 80+ gold is at least 88% efficient at 20% and 100% load and 92% efficient at 50% load You need to take the approximate output power of their PSU (wall power multiplied by efficiency) and then compare that against the rated power on your unit, so there they have ~325w draw from the PSU even with high powered overclocked CPU etc and a 450w unit would be fine | ||
|
rebuffering
Canada2436 Posts
Here's a reddit thread for the issue http://www.reddit.com/r/dragonage/comments/2mjmwd/inquisitions_drm_denuvo_antitamper_harms_your/ and heres the original article translated from Russian https://translate.google.it/translate?hl=it&sl=ru&tl=en&u=http://gamemag.ru/news/drm-zashchita-dragon-ageinquisition-denuvo-antitamper-ubivaet-ssdhdd-101753 "Want to know how many times the average of the code LoF.exe being transferred from HDD to memory and back in one hour? 150000 retry copying records. This is 10,000 times greater than normal. DRM continuously decodes the game code into memory and encrypts obratno.Eto one of the most idiotic use of encryption techniques, which I have ever seen. Despite the fact that the small pieces of code (maximum steam kilobytes per entry), are stored in one memory block. Playing the game for 4-8 hours (depending on the quality SSD) means that the unit can bye c forever. " Thoughts? as i said in the DA:I thread, i will be asking for a refund if this is actually true. | ||
|
Ropid
Germany3557 Posts
Just click around in "resmon" in the "Disk" part of the window and see what it can do. I think there's also columns that are by default hidden that might show interesting stuff like counters that show a sum of all reads and writes and not just the current bytes-per-second. Check that out by right-clicking on the top of the table. I bet that "perfmon" tool would be able to record everything you need for proof, like what exactly is happening to that .exe file on the disk that's mentioned, but no idea how it works. You should also check this out as it might show what you want somewhere (it's a great Task Manager replacement): "Process Exploer": http://technet.microsoft.com/en-us/sysinternals/bb896653 Other tools you might want to look at: "ProcessActivityView": http://www.nirsoft.net/utils/process_activity_view.html "Process Monitor": http://technet.microsoft.com/en-us/sysinternals/bb896645 | ||
|
rebuffering
Canada2436 Posts
On November 19 2014 06:06 Ropid wrote: There's tools in Windows that should help you confirm this for yourself. There's a "Resource Monitor" or "resmon" which might already show everything you need, and there's a complicated tool "perfmon" that can record all kinds of stuff. I can't really help right now about where to click exactly in the Resource Monitor as I don't have access to Windows at the moment. Just click around in "resmon" in the "Disk" part of the window and see what it can do. I think there's also columns that are by default hidden that might show interesting stuff like counters that show a sum of all reads and writes and not just the current bytes-per-second. Check that out by right-clicking on the top of the table. I bet that "perfmon" tool would be able to record everything you need for proof, like what exactly is happening to that .exe file on the disk that's mentioned, but no idea how it works. You should also check this out as it might show what you want somewhere (it's a great Task Manager replacement): "Process Exploer": http://technet.microsoft.com/en-us/sysinternals/bb896653 Other tools you might want to look at: "ProcessActivityView": http://www.nirsoft.net/utils/process_activity_view.html "Process Monitor": http://technet.microsoft.com/en-us/sysinternals/bb896645 Thanks for the tips! As someone who is too lazy to do tests im following several of the reddit threads that are talking about the potential problem. There's a few people who are reporting no abnormalities in terms of HDD read/write, so there's no "definitive" proof one way or the other, since to me, the DRM itself could possibly be hiding some numbers. Obviously, im an idiot so what i just said might be totally bullshit, but fucking Securom is the devil and i wouldn't put it passed them if they were skewing the read/write numbers. The closest thing ive found to an answer is this from a reddit user: + Show Spoiler + I made some tests and as far as I know this is utter and complete bullshit (Not that it is any proof of knowledge, but I have been a software developer for 10 years now and dabbled in software protection). I verified that the file is same before and after an 1h session. This means the executable was not modified. If as the source claims encrypted pages were written back the hash would differ (that can be checked by anybody using a hash computation software) The executable is digitally signed by Bioware. The file cannot be modified without invalidating the certificate. You can check that by looking at the file properties in Windows Someone checked using ProcessExplorer (a very well know utility) how much data the game wrote to the disk. After 2h the amount was 20MB (probably save games). The screenshot is here http://postimg.org/image/46v2ej7kx/ I did not check this myself but the utility used (ProcessExplorer) is free so anyone can check that Moreover, if the executable changed constantly it would wreak havok with antivirus (false positive). So far I haven't heard anything of the sort. | ||
SwizzY
United States1549 Posts
My computer recently failed to boot (but I heard the CD drive and saw some activity from my SSD from the LEDs I have on my case) but it did not light up my mouse/keyboard and did not show anything on my monitor. I opened it up and barely touched anything (just checking to see if anything came loose, which it didnt) and it booted up as normal. The only real change I made was moving my monitor plug from the power strip which powers my computer to the wall socket. Any opinions on what might've been causing it in the first place? The monitor is new (3 days ago) and has been in use so insufficient current through the power strip or some bs like that was my theory, though I know it's probably weak. | ||
|
SoleSteeler
Canada5449 Posts
I got the WoW expansion and I was disappointed that it too strains my computer to ~95 C which seems to throttle the CPU, lowering performance. Tonight when I played, I was getting 40-50 FPS while the CPU was cooler (80 and lower), then the FPS dropped to 15-25 as it climbed higher and higher. I'm not positive it was doing this, but I definitely had higher FPS when I first was playing. So anyway... I looked up one of those laptop fan/stand things. Do they work well? I was looking at this one. Since it only costs $18. this one costs $40, and this and this are ~$20. I can either buy one of those (doubt I'd go with the $40) or just tough it out for 3 more months until I move back to Canada and can build a desktop. But if any of these will actually keep my laptop cooler by 15+ degrees I will buy one... I'm not really sure what I'm asking. But I guess just: do these laptop stands tend to work decently or are they a waste? I'd need at least a 15+ degree drop to get a performance increase, I believe. | ||
|
y0su
Finland7871 Posts
| ||
|
Cyro
United Kingdom20321 Posts
For WoW you could try lowering graphics settings as much as possible (your laptop might share a heatsink between cpu and cpu) while also simultaneously limiting your FPS to a number that's low, but that you're more happy with, like 30 for example. You could also get hwinfo (www.hwinfo.com) and use that (click sensors tab when you open it, and then you can double click any value there to open a real time chart of CPU clock speed for example) to monitor performance info. CPU, GPU temperatures and clock speeds | ||
|
Myrmidon
United States9452 Posts
It is also possible tearing apart the laptop and repasting the CPU and GPU could help (assuming the heatsinks are already perfectly clean), but that could be more effort than it's worth. Fundamentally, the components do not use more power than they used to, and digital logic circuits do not flip bits slower with age (in this context), so something's changed either about the programs you're running including background processes like potential malware, or something's changed with the cooling mechanism like heatsinks needing cleaning or a fan dying. Regardless, you should definitely investigate the "why" and what kinds of loads and conditions are producing the results you're seeing, as Cyro's instructed. | ||
|
SoleSteeler
Canada5449 Posts
I've had issues for about 1.5 years now, and I've only had the laptop for 2 years. The first time it overheated and shut off was the first day playing the HotS (SC2) campaign. I was so excited to play it and then it just died. I was so disappointed! I thought my computer was completely fried. Fortunately I figured out it was overheating. I had issues until I learned to blast the fan with compressed air. That literally dropped the temp by 15-20 degrees at the time. Since then I've done it once every month or so and it rarely shuts off due to overheating. It did a few times in Heroes of the storm when I first started playing it, but it hasn't for many months. After doing some more reading, I put my laptop on a big book (5th book of Game of Thrones, hardcover, minus the book jacket). I just played WoW for about 25 minutes and my highest temp was 80 C, FPS never dropped below 40. If this was all I've needed to do for 1.5 years now is this, I'll be slightly annoyed at myself, but also happy that the issue is fixed. I made sure the fan wasn't blocked, obviously. The first 10 minutes I didn't even hit 70. It hit 75-80 when I was in an area with a ton of mobs fighting each other (all NPCs). Anyway it looks promising. I was running hwinfo but I didn't click the sensors tab so I didn't get to see the load. (I've been running a different CPU/GPU temp monitor). I'll try it more after work, hopefully for an hour+ and see how it goes... Thanks again! edit: played for a few hours, never went over 85 C (usually stayed around 75-80 C) and experienced no FPS drops which is awesome! I want to try SC2 and see if I can micro better instead of having my FPS drop significantly 5 minutes into the game -_-; | ||
|
Cyro
United Kingdom20321 Posts
+ Show Spoiler + ![[image loading]](http://i.imgur.com/T2RZ73k.png) ![[image loading]](http://i.imgur.com/JkGUkDI.png) any help/tips appreciated. That's with windows power plan on max performance, but c6/c7 still enabled - not sure if that's ruining my results so i'l check edit: Turned off c-states, EIST etc.. + Show Spoiler + ![[image loading]](http://i.imgur.com/hlmGP2C.png) should i just leave them off and use full vcore 24/7? I was having actually noticeable issues due to that, maybe i should turn c-states etc on/off one by one. As it is now, probably can't get much better than 44 after 1 minute of testing because i can't disable HPET on this board (i will be noting that for my next purchase..) but running everything 100% speed and voltage all the time makes me kinda unwilling to run an aggressive overclock | ||
|
Ropid
Germany3557 Posts
Did something get updated recently, like did it only start with the new NVIDIA drivers for example? Windows Update could also have secretly installed a new driver for a device. It could also be an update to Windows internals like kernel and the graphics stuff as I think Microsoft regularly ports things from Windows 8 back to 7. You could try to see if people post in forums about something like that by having Google only report search result from the past week or past month. About strange reasons, I wonder if something like the C: drive having issues could cause this. | ||
|
Cyro
United Kingdom20321 Posts
On November 27 2014 01:16 Ropid wrote: I only know about going through all devices and their drivers and seeing if people report issues online for one of them. Did something get updated recently, like did it only start with the new NVIDIA drivers for example? Windows Update could also have secretly installed a new driver for a device. It could also be an update to Windows internals like kernel and the graphics stuff as I think Microsoft regularly ports things from Windows 8 back to 7. You could try to see if people post in forums about something like that by having Google only report search result from the past week or past month. About strange reasons, I wonder if something like the C: drive having issues could cause this. well it dropped from 5k to 25 (dropped by 99.5%..) as soon as i disabled those c-states etc. I'm not sure what could have changed because i didn't check it highly regularly, but i'm using a new mouse and a new display driver - that's the ONLY thing plugged into USB on my whole system (ps/2 keyboard, nothing else) and those two were very high on the list on the second pic i posted on first spoiler while i'm posting here: My vg248qe got another messed up pixel, stuck pixel is i think the right term. It appears pink when you put white behind it, changes a bit with different colors and is not really visible (or not visible at all) when it's supposed to display black. There's two of them now (both were not there when i first got monitor) within about a 1/5'th slice of the monitor near the right side, is that anything to worry about? One was a very minor annoyance to me, but i don't want another 5 to show up at this rate considering i didn't have the monitor for very long and i kinda wanted to keep it for something like 5+ years (if not for my usage, maybe someone elses close to me like my dad or cousin for example) | ||
| ||
|
|
Monday Night Weeklies
Replay Cast
ChoboTeamLeague
WardiTV Korean Royale
BSL: GosuLeague
The PondCast
Replay Cast
RSL Revival
herO vs Zoun
Classic vs Reynor
Maru vs SHIN
MaxPax vs TriGGeR
BSL: GosuLeague
RSL Revival
[ Show More ] |
|
|
EPT
