Forum Replies Created

Page 4 of 9
  • It was difficult to find a direct comparison of PCIE 1.0 to PCIE 2.0 since the tech update was in 2008 and everything else has moved on.

    So the test was done with the best GPUs of 2008.

    https://www.tomshardware.com/reviews/…-2.0,1915.html

    “From this standpoint, we clearly recommend PCIe 2.0 to anyone, because it has no disadvantages.

    But is PCIe 2.0 really necessary yet? As long as a graphics solution can operate with data that is stored within its local video frame buffer memory, both the reasonably mainstream Radeon HD 3850 and the hardcore GeForce 9800 GX2 will operate close to their maximum performance, even if the PCI Express link width is limited to x8 or x4. Once larger textures need to be accessed, as is the case in Crysis or Microsoft’s Flight Simulator X, interface bandwidth becomes a crucial element. Any link width below x16 will noticeably limit these games’ playability.”

    PCIE 2.0 8 lanes = PCIE 1.0 16 lanes

    So, only if you are doing high bandwidth transfers or loading lots of complicated textures will you see a large difference. I would imagine that this would include GPU CUDA stuff, but YMMV.

    The above tests only found a big difference when used with Crysis and Microsoft Flight Simulator.

    They were testing with ATI 3850 and Nvidia 9800GX2, I would imagine that progress has been made with GTX570 and other modern cards over the bandwidth needs of cards in 2008. But perhaps not.

    When using a display, most people like having a card that makes the screen pop up the second you hit the power button. If you don’t mind waiting, any PC GTX570 will work. You should just keep an EFI card somewhere handy, should you ever need to install system software, or troubleshoot OS issues, a non-EFI card will NEVER fire the screen up.

    For Resolve the need to see boot screens is reduced, but I can’t imagine a situation where limiting the bus speed could be seen as a “good” thing.

    SO feel free to buy your 570 from Newegg or TigerDirect or whoever. If you find that your needs include a boot screen and/or high bandwidth transfers, get in touch.

  • David Pirinelli

    May 1, 2012 at 7:41 am in reply to: Adobe CS6, MacPro, and Nvidia Quadro 6000

    Hi, I run MacVidCards.

    The Quadro 6000 has been supported in the drivers since somewhere back in 10.6.

    That said, only 4 GB shows up for CUDA, 2 GB in System Profiler, and OpenCL so far is only place all 6 GB show up.

    I have heard that all 6 GBs will eventually show up everywhere but…..

    We have GTX470s and GTX480s that work in 10.6.8 and we just added GTX570 and GTX580 for 10.7.3.

    The GTX570 2.5 GB card is turning out to be a real powerhouse for Adobe apps. In the Resolve forum and at MacRumours forums there have been posts about how well that specific card works for CS6. Apparently they are supported from the get go, no hacks needed. One guy compared performance to his Quadro 4000. Didn’t go well for the Quadro.

    Our cards offer full boot screen support from first press of power button and can drive an Apple LED as well as a 30″ via DVI. The 570 with 2.5 GB is going to be “the” card to have for the near future.

    We have the Quadro 5000 and 6000 for Mac, and they are currently DisplayPort and DVI as well, though I could probably set them up to run 2 @ Apple Displays if there was interest.

    But for most people, the GTX570 with 2.5GB is hard to beat.

  • Hi, I represent MacVidCards.

    We offer a nice range of Nvidia cards for Mac 2008 and later. These are fully supported in the OS as they use same drivers as Quadro 4000.

    If you check up in the Resolve forum, our cards have been making their CUDA app work much better on Macs.

    We have GTX470, GTX480, Quadro 5000, Quadro 6000 and Tesla C2050 for Mac. These cards offer boot screen right from the second you hit the power button. The Quadro 6000 is limited to 4 GB of VRAM by OSX.

    You need to remove the AGPM kext and OpenCl support requires a minor hack, but OpenGl and CUDA are solid and powerful.

    And for those on 10.7.3 we just added GTX570 and GTX580. (THe gTX480 and GTX580 both require more power than is available in a Mac Pro so require some fiddling) There is a GTX570 with 2.5 GB of GDDR5, my feeling is that it will be the “sweet spot” for many Mac Nvidia GPU needs.

    We are based in Hollywood CA and offer free delivery in the Hollywood area. (Venice is NOT in the Hollywood area)

  • David Pirinelli

    April 27, 2012 at 4:26 am in reply to: GPU Ram for 4K projects, how much is enough?

    These cards I mod are being added to, nothing is really being lost.

    We ADD a Mac EFI, but the PC BIOS remains.

    In fact, the cards can reboot into Windows whenever you want.

    Even better, if you use your Mac to boot into Windows via Bootcamp, the PCIE fix goes there too. Currently if you boot into Windows with a PC only BIOS, you will have the same PCIE 1.0 2.5 GT/s issue that exists in OSX. FWIW, flashed ATI/AMD cards have same issue. Apple must have left this little trap to keep flashed cards “inferior” to official Apple ones. The PCIE 2.0 switch is a separate part of ROM that has been fiddled with by Apple for this purpose.

    So our cards GAIN Mac EFI and retain PC BIOS. It is this “Dual ROM” that requires a new EEPROM chip as the space necessary is never available on retail cards.

    So the cards are truly cross platform once the mod is done.

    The ONLY exception is with the multi outlet cards. These are BIOS modded to enable only 2 outputs. For the Dual DVI cards, both DVI ports will work, though HDMI will likely NOT work.

    For the Dual DVI, HDMI and DIsplayPort cards, 1 @ DVI and the DIsplayport are the only ports working. TO enable all 4 again will require ref lashing with PC BIOS. THis is a result of how Mac OSX Nvidia drivers work. Note that on the PC Quadro 4800 and Quadro 4000 compared to the Mac ones, they REMOVE a port. The PCB still has the traces, they just don’t place the port in . This is why I am very curious to see how the GTX680 drivers will work.

    In any case, our new GTX5xx cards work quite well in Windows as well as OSX. If you ever need to use ALL of the ports in Windows, you can ref lash the card in Windows.

  • David Pirinelli

    April 26, 2012 at 9:42 pm in reply to: GPU Ram for 4K projects, how much is enough?

    Who wants to find out?

    So your Cubix shows up in System Profile under the “PCI Cards” section?

    I have heard from several Cubix folk that this section doesn’t work.

    If it DOES work, I would be eager to do some 580s for someone.

    In addition to selling the finished cards, I am also going to be doing a mod service for people’s pre-purchased GTX5xx cards. Got the first one one today and will be shipping it out today or tomorrow. If there is a post-production house in Hollywood with some 580s in a Cubix, get in touch. Will take me a couple hours per card, especially if it is a card I haven’t written an EFI for, they have to be custom tailored to each card.

    From this discussion, it appears to me that the “sweet spot” for people NOT wanting to Cubix is a GTX570 with the 2.5 GB of RAM. Works in a Mac Pro power envelope and fast as a 480 with less heat and power.

    A guy picked one up 2 days ago, I’ll see if I can get him to post about it. The fact that these can also run Apple LED displays is icing on the cake.

  • David Pirinelli

    April 26, 2012 at 6:15 am in reply to: GPU Ram for 4K projects, how much is enough?

    “So somewhere between 1 GB and 2 GB seems to be the current “Magic Number”.

    Looks like my guess & source at BMD were right.

    Good to know what the “Minimum” is.

  • David Pirinelli

    April 26, 2012 at 2:37 am in reply to: GPU Ram for 4K projects, how much is enough?

    Was a BMD employee who wrote that.

  • David Pirinelli

    April 25, 2012 at 5:54 pm in reply to: GTX 470, GTX 285, PPro, and Mac Pro 3,1

    This is good news since we just started selling Mac EFI GTX570

  • It is quite possible that for what most of you are doing, the difference would be small.

    It may require cards in a Cubix to make a big difference. But I am very hard pressed to imagine a situation where it would be BETTER to have the PCIE interface running at half speed.

    In any case I have now gotten GTX570 with DIsplayport and DVI working as well as GTX580 3 GB and GTX580 3GB Ultra Classified. All 3 cards run perfectly and at full PCIE 2.0 5.0 GT/s speed.

    They all can show boot screens on either port (including Apple 27″ MDP)

    They all sleep and wake perfectly. ANd they will show boot screens or work with System Install and Recovery Disks, something the PC cards can’t do. They also work in all released versions of 10.8, something the PC cards can’t do with out some hacking.

    In addition, as a result of the fact that they are integrated at EFI level, the “PCI Cards” section of System Profiler actually WORKS, allowing you to see all of your cards, how they are being identified, and what speeds they are running at. I imagine that might help when trying to troubleshoot issues.

    I don’t know how cards function in a Cubix or Cyclone because I don’t have one. It would be interesting to find out if the 2.5 bug from Non-EFI cards carries over to those. It would be interesting to know if the 580 classifieds fit into one of those and if there is enough power. (they require 2 @ 8 pin and 1 @ 6pin…EACH !!!) but they are quite a bit faster than standard GTX580. 3 of them in a Cubix would be capable of quite a lot. I think the Cubix by itself kills the “PCI Cards” section so to find out what speed the link is, we would need to use CUDA-Z or the lspci app I mentioned.

    Using PC cards OOTB the box works in the current OS, I am offering a better implementation.

  • “Under system profiler it is reported that it is at full speed”

    I am curious where you see this under System Profiler ?

    On cards using the “Magic Driver” the section of system profiler that shows PCI card speeds doesn’t work. So there should be no way to see what speed they are running at. The “Graphics & DIsplays” section doesn’t show link speed. People using Cubix also frequently report seeing the same “Error while gathering PCI info” or something to that effect.

    Another way to see whether your card is running in PCIE 2 at full speed or is stuck in PCIE 1.0 at half speed is to use lspci, this is a piece of the OS that Apple decided not to include but you can install it. RUn CUDA-Z in background to create a load.

    using “lspci -vvv” you will likely see this:

    nkCap: Port #0, Speed 2.5GT/s, Width x16, ASPM L0s L1, Latency L0 <1us, L1 <4us
    ClockPM+ Suprise- LLActRep- BwNot-
    LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk-
    ExtSynch- ClockPM+ AutWidDis- BWInt- AutBWInt-
    LnkSta: Speed 2.5GT/s, Width x16, TrErr- Train- SlotClk+ DLActive- BWMgmt- ABWMgmt-

    Our cards show this instead:

    LnkCap: Port #0, Speed 5GT/s, Width x16, ASPM L0s L1, Latency L0 <512ns, L1 <4us
    ClockPM+ Suprise- LLActRep- BwNot-
    LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk-
    ExtSynch- ClockPM+ AutWidDis- BWInt- AutBWInt-
    LnkSta: Speed 5GT/s, Width x16, TrErr- Train- SlotClk+ DLActive- BWMgmt- ABWMgmt-

    I will have a crack at 680 EFI soon, but currently there is little point since the drivers don’t have support.

    So if I write a good EFI for it it will show boot screen but crash at desktop, or show unaccelerated 2D.

Page 4 of 9

We use anonymous cookies to give you the best experience we can.
Our Privacy policy | GDPR Policy