Question about multithreading in hardware mode
#11
(07-02-2015, 12:16 PM)Serial Hacker Wrote: I Have a GTX 750 Ti and it's showing 86.4 GBps of Bandwidth and i'm sure that these limitations are not related to either my PCI-E link or my Memory type

no the 86.4GB/sec is much faster than PCIe and your system ram (which do 32GB/sec and ~15GB/sec respectively) the problem comes from GPU<->CPU transfers which happen a LOT with the playstation 2, as you can imagine, that slows things down dramatically, the reason GSDX is as fast as it is, is because it throws everything on the GPU and does as much processing as it can on there in the blisteringly fast graphics memory, but on occasions when things have to be moved back to system ram to work on then uploaded again to the GPU, this brings speed down by huge amounts.
[Image: ref-sig-anim.gif]

Reply

Sponsored links

#12
Yes. As Ref said it depends on the RAM speed too.

I'm working on some PCIE IP, so I know the spec quite well.
There are lots of overhead. You have overhead of physical layer (128b/130b). You have a packet overhead (I need to double check it but it is around 4-5 DW so 16-20B). In order to reduce the latency real data payload of packet is around 128B. There is also some extra packet to communicate the room available for transfer (credit mechanism, ack/nack). So globally I think RAM is faster than PCiE. (Ram has some overhead too)

Ofc I'm too slow with my tel...
Reply
#13
(07-02-2015, 12:20 PM)refraction Wrote: no the 86.4GB/sec is much faster than PCIe and your system ram (which do 32GB/sec and ~15GB/sec respectively) the problem comes from GPU<->CPU transfers which happen a LOT with the playstation 2, as you can imagine, that slows things down dramatically, the reason GSDX is as fast as it is, is because it throws everything on the GPU and does as much processing as it can on there in the blisteringly fast graphics memory, but on occasions when things have to be moved back to system ram to work on then uploaded again to the GPU, this brings speed down by huge amounts.

So the reason why SW mode is slow is because of the heavily lower Bandwidth ? interesting...
Reply
#14
(07-02-2015, 12:08 PM)refraction Wrote: Hmm not sure, the theoretical bandwidth of a PCIe 3 graphics card is 32GB/sec, which is obviously faster than system ram. That is theoretical however, much like dual channel memory is not a 50% increase in speed.

as a general rule, PCIe 3.0 isn't any faster than PCIe 2.0 which is a theoretical maximum of 16GB/sec, but again that could be a system ram limitation.

isn't that memory bandwidth nullified if you write to the vram directly? pcie hotlinked with the memory controllers on die. or not?
Reply
#15
Actually a faster gpu might helo too. When you read back a target, the gpu must have finished to render it first. It is often worst, because GPU is often late (vs the cpu)
Reply
#16
Okay so GPU-Z was showing that my card was running on PCI-E 1.1 as sort of a power option only when the card is idle. so i kinda disliked that and went to the BIOS and tweaked it

apparently Raising the second PCI-E from x2 to x4 disabled PCI-E 1.1/1.2 Tongue (small note in the BIOS)
Reply
#17
Yeh some motherboards have freaky PCIe setups and it tries to juggle PCIe lanes n stuff, which is weird Tongue

The jump from 1.x to 2.0 was huge for PCIe and bandwidthwise the jump from 2 to 3 was huge too, but in practice, games were at most 2-3fps faster in this jump

Edit: Okay it wasn't "quite" as big in PC gaming as i thought, probably due to how optimised PC games are, but here's a graph of Shadow of Mordor and the PCIe lane speed and it's impact http://www.guru3d.com/articles_pages/pci...ew,13.html
[Image: ref-sig-anim.gif]

Reply
#18
Well game are optimized to use the correct bandwidth, if the link is saturated perf go down very quickly. Besides GPU mem increase a lots. So you can store texture for severals frames. And finally API slow evevery thing down.
Reply
#19
BTW shouldn't DX12 / Vulkan (looking at you gregory) grant even a lower access without the need for those pesky GPU-CPU Transfers
Reply
#20
sure new API allow to teleport the data directly without passing by the pcie link. It is much faster Tongue2

Quote:Edit: Okay it wasn't "quite" as big in PC gaming as i thought, probably due to how optimised PC games are, but here's a graph of Shadow of Mordor and the PCIe lane speed and it's impact http://www.guru3d.com/articles_pages/pci...ew,13.html
Hum very high resolution could be limited by the GPU shader computing.
Reply




Users browsing this thread: 1 Guest(s)