> The 1-terabyte Deskstar is declared to have a typical Power-on to Ready time of 20 seconds whereas the Caviar GP takes only 13 seconds to accelerate its platters.
@allquixotic what is the software feature? is it directX? and why is there a single framebuffer? i always thought that each output "port" used a framebuffer? I am Never learning this either.
@Psycogeek no, there is only one framebuffer per graphics card... for instance if you have two 1920x1080 monitors, your framebuffer is 3840x2160 in size
under the hood what the OS does is it knows where each monitor is in the framebuffer, so it can use that information to decide where to push pixels to
and no, it's not DirectX, because on some cards it works while on others it doesn't, so it's clearly not built into Windows
what Nvidia Optimus does is, on a laptop, the Embedded DisplayPort connector that connects to the laptop's flat panel is connected to the Intel iGPU on the CPU or motherboard... then, when you start an application that uses 3D rendering, if you configure it to run in high performance mode, it dynamically links the application to the Nvidia 3D graphics drivers, which create an off-screen surface to do the rendering in, then that is copied to the Intel iGPU's framebuffer to be displayed on the eDP
if Optimus weren't in place, then putting an Nvidia graphics card in a system that boots to the Intel iGPU as its primary framebuffer would be a total waste, because the Nvidia card would never get used at all
@allquixotic how then does a single framebuffer which is usually 2 memory areas being swapped and filled work at 2 dirrfernet refresh rates for different monitor refreshes ? I am going to test and see if i can even get that going here, because mine have one basic res and refresh.
@Psycogeek conceptually, it doesn't really matter whether it's two separate areas in memory or one, because they are treated as a linear continuous space at the driver level
it would be perfectly possible to update one part of the framebuffer at a different time than another, if they have different refresh rates
it's basically a 2D array as far as the driver is concerned; if the MMU on the GPU wants to split it up into fragments, it can choose to do that
all the driver needs to know is: when the refresh interrupt comes in, push the latest data to the framebuffer
@Psycogeek I believe those types of configurations are treated as a square framebuffer where the total width and height are treated as the 2D linear framebuffer, and the space "beneath" the first monitor is treated as invalid area
@Psycogeek the stream of data from the framebuffer to the DisplayPort/HDMI/DVI interface might be "line by line", but the framebuffer itself isn't written out line by line necessarily; most modern drivers actually use 2D tiling, which means, split up the screen into much smaller squares and compute each square in parallel
I don't know how many tiles they use but 64 or 128 might be a reasonable number
@allquixotic that sounds like a good method to block it, but then how do you get the binary stream of the block through? is there comressed going on here where an advantage would exist to blocking it out and compressing and streaming compressed?
@Psycogeek if you're doing software rendering, the CPU would compute the framebuffer pixel values and send them to the GPU framebuffer to be eventually written out to the display device; the details of the transfer from the framebuffer to the display device are not really interesting to me
if you're doing hardware rendering, the GPU itself will write to its own framebuffer once it's done doing that OpenGL / Direct3D stuff that involves messing around with pixels in VRAM
unless you have Nvidia Optimus, in which case, the finished messed-around-with pixels get copied back into RAM then transferred to the Intel iGPU's framebuffer
there's probably not any compression going on at that level because it would degrade image quality
@allquixotic right i am going off the topic of the main aspect of the question, because i didnt get past the first part, where mentally i cannot imagine how that is done already. (2 framebuffers)
@Psycogeek typically what happens is, you have a front buffer and a back buffer... you do your drawing calls to the back buffer while the user is viewing what's on the front buffer. then, when you get an interrupt from the hardware saying it's time to refresh the screen, you "swap" the front and back buffers, by DMA transferring the back buffer's contents from RAM, into the front buffer on the GPU
the DMA transfer is very fast and basically involves a very high bandwidth data transfer from system memory to the GPU over PCIe
as soon as you make the call to swap buffers, you can start drawing on the back buffer again until the next frame is needed
@allquixotic new ideas for framebuffer transfers of info are cool, but i am still back trying to figure out the one they are using , in this hardware here.
of course, if you get interrupts starting to stack up because you can't produce frames fast enough for the refresh rate of the monitor, then the driver will ignore consecutive interrupts which results in dropped frames
@allquixotic which for me sucks :-) Way back in time when the PCI was much slower the activity of the bus(es) would slow the other transfers through the same bus. Dumping frame buffers through my PCI lines has been thing i am still scared of, when it is unnessisary.
With multiple monitors present, each screen will have its own graphics buffer. One possible scenario for programming is to present to OpenGL or DirectX a continuous, virtual frame buffer in which the OS or graphics driver writes out to each individual buffer.
With some graphics cards, its possible to enable a mode called "horizontal span" which accomplishes this. The OpenGL/DirectX programmer then renders to a very large frame buffer for output. In practice, and with recent cards, this mode is being phased out because it does not make very good use of GPU parallelism, and does not support arbitrary arrangements of monitors (they must all be horizontal).
A more recent technique uses the wglShareLists feature of OpenGL to share data across multiple GPUs, and then render to each individual monitor's frame buffer
Try and get this out SLI is "Scan-Line interleave" where every other line of 3d stuff is rendered in a different GPU processor and sent to one GPU framebuffer.
"Dual Graphics" is an AMD APU/GPU shared work method using crossfire as the method. (from what i read most people did not bother with this, it did not increase speed much, even the claimed increases will not "fix" a bad Frame rate completly)
USB3.0 has pretty high tolerances, so a cheap cable can lead to pretty bad noise
@Andrew well then :P
standards like USB and Ethernet actually specify things like how often to twist the cable, how much can be untwisted at the end, the thickness of the metal, etc..
most cheap manufacturers either fuck up the twisting, skimp on the metal, or both
If you turn Xfire/SLI on then you get no output from the second (or third, or fourth) card.
You only use the calculating power from the additional cards
@jimmy Hoffa: Scary. I recognise your starred sentence without opening my book of ignition! or by re-reading things I will not work with. ;-)
10 - 20 seconds sounds about right for a desktop harddisk. More if you have half a dozen and start them sequentially. e.g. spin up [SCSI] disk 1 at boot. spin up [SCSI] disk 2 at boot plus 5 seconds spin up [SCSI] disk 3 at boot plus 10 second spin up [SCSI] disk 4 at boot plus 15 second spin up [SCSI] disk 5 at boot plus 30 seconds (usually delayed by ID times a fixed value)
Or if you set the HDD controller (e.g. the RAID card) to delayed spinup.
I tend to avoid Xfire/SLI because I do not need the performance and a single mid/high level card is much easier and often has less issues than going dual or triple cards
I can't imagine a consumer drive taking even 5 secs to spin up. mine take two, maybe three (when hot plugging. have not bothered to actually test normally)
Boot, initialise, try AHCI on second set of SATA chips (about 10 seconds). 30-ish more seconds for the RAID card to boot and initalise the arrays. Then about a second minute from POST to the OS.
So about 2 minutes to boot into windows where I can log in.
(Modern i7 920/X58)
Well modern. 5 years since it was the newest. Not old.
I am also used to servers. Some of which can wait a full minute before generating their first output. Nerve wrecking if you encounter that for the first time.
@Hennes i wouldnt mind a 1min boot, for the regular stuff what drives me up the wall is when it is slow to boot, AND i am having to boot 30 times an hour, setting up. I added my raid card last, and then still needed to do 2-many reboots for a few other adjustments.
Depends on the cards. Usually SCSI is not build with boot time as an issue. Especially not when you add wide high voltage differential hostadaptors. For those reliability and speed after booting is more important
I usually boot twice per day. In the morning. (like now). Power on desktop, feed cat, brew coffee, return to a waiting and fully booted system
And similar in the evening. Home, pet cat, power on desktop, kick off boots. Store work related stuff. Unload food from shopping, .... ... , go to booted desktop
I do not often reboot servers. Last time was Monday when the entire building went down (maintenance on a 10Kv line)
Before that: uhm... 200-ish days?
Exception: the windows file server at work. Weekly reboots
Twice per week, because a single reboot would leave remote desktop in a failed state
I don't think mine goes that high if I'm right next to it (in fact, not even a tenth of the strength, remember logarithmic scale), and mine is pretty close to the AU limits
@Sathya ah, must be that underground NSA network :P
superuser.com/review/suggested-edits#suggested-edits/… I dont usually like when people try and edit into the middle of some answer. But this one? why would gwavity miss this? I have never seen the USB port 5v power itself turn off, and that could be a disaster in the middle of a long flush plus the flushing of large disk caches?
""In Windows XP and Windows Server 2003, when a USB device is marked as Removed, the USB hub port to which it is connected is Disabled. When the port is Disabled, no further USB traffic is sent to the device."" that might be how it should read, but i dont know , mabey laptops or something would actually power off the USB port itself.
@Psycogeek Well, the monitor is 30", so compared to my old 19" 1280x1024, it was a nice transition. I wouldn't want 2560x1600 on a smaller than 30" monitor, that's for sure. All that'd give you is a headache from trying to focus on everything being tiny. :P
@William'MindWorX'Mariager i was wondering about that. But with all the res it does interpolate to a lower res well (for games specific)
@William'MindWorX'Mariager I would love the "space" as in having more room for more windows. right now i use 2nd monitor and even after using it constant, and getting used to it, it is still a 3rd leg, a side show, just not the same as having it all in around me, instead of ---> over there.
The (blind) landlord has the 27" at the same res (1920x1080 type) , and it is Huge, but without some Resolution , i see Squares , all the cells of the pixels.
Yeah. I frequently debug and dabble in reverse engineering, so it's not uncommon for me to have several instances of IDAPro and Cheat Engine running, so it clutters fast. The new monitor really alleviated the problem.
Speaking of three, I'd really love a way to split windows in three equal sizes horizontally. Similar to how you can shove a window to the left half or right half.
Meta Stack Overflow/(any room): red tags
Send a message consisting solely of [status-SOME_PLAINTEXT_STRING] and it will appear in chat like a moderator tag. Doing this as a reply to another message is okay, but other text or formatting will not work.
Gaming/The Bridge: minigame Eggs.Asteroids...
Meta Stack Overflow/(any room): red tags
Send a message consisting solely of [status-SOME_PLAINTEXT_STRING] and it will appear in chat like a moderator tag. Doing this as a reply to another message is okay, but other text or formatting will not work.
Gaming/The Bridge: minigame Eggs.Asteroids...
