Flashing R9 280X for boot screens/PCI System Information

Discussion in 'Mac Pro' started by Mr. Zarniwoop, Oct 24, 2014.

  1. MacVidCards Suspended

    Joined:
    Nov 17, 2008
    Location:
    Hollywood, CA
    #101
    I have seen boot screens.

    But keep in mind there is no frame buffer in 8000 controller other than the 6 mdl one.

    So no framebuffer to assign. So likely that rom will never behave the way we have come to expect.

    But maybe time for a different sort of boot screen card.
     
  2. Mr. Zarniwoop thread starter macrumors demi-god

    Mr. Zarniwoop

    Joined:
    Jun 9, 2005
    #102
    Very interesting.

    I thought the newer 290X gets properly initialized by Yosemite? Or is the driver support lacking?
     
  3. MacVidCards Suspended

    Joined:
    Nov 17, 2008
    Location:
    Hollywood, CA
    #103
    2 very different ways for a card to work.

    Esp AMD cards.

    With an EFI and assigned a frame buffer personality.

    Or self-inited and using generic "Radeon Framebuffer"

    Currently there is no personality (other than the one for upcoming nMP 7,1) to assign an R9 290X.

    Go digging in 8000 controller, you will understand what I mean.
     
  4. Mr. Zarniwoop thread starter macrumors demi-god

    Mr. Zarniwoop

    Joined:
    Jun 9, 2005
    #104
    I guess we'll have to hold out hope for new personalities in a future Yosemite release (maybe an iMac update?).

    In the meantime, Yosemite with a R9 280X really has made my 2006 MacPro1,1 quite useable. I probably bought another year or more of use out of this Mac for the cost of the graphics card!
     
  5. Earl Urley, Nov 24, 2014
    Last edited: Nov 25, 2014

    Earl Urley macrumors 6502

    Joined:
    Nov 10, 2014
    #105
    more good news

    My other R9 280x seized up for some unknown reason, so I returned it to where I bought it.. it was an open box with some nicks and cuts on it, eh, anyway MicroCenter credited my cc, so I ordered another MSI R9 280X GAMING 3G refurbished for $199 from Amazon, even less than what I paid for the first card.

    The box had an 'official MSI refurbished' seal which consisted of a grey laminated label with the MSI logo that said PREOWNED / OPEN BOX / REFURBISHED on it. My heart sunk when I opened the antistatic bag.. the EAN and UPC were totally different, and at the bottom of the board instead of "V277 Rev 1.0" it read "V277 Rev 1.1!" This wasn't the same exact card that Mr. Zarniwoop had mentioned!

    Oh well, let's try it and see.. double checked the BIOS switch was set to Legacy, plugged it in, hooked up the 8-pin to 6 pin adapter and the regular 6 pin adapter, booted to Boot Camp and the video came on easily enough.. saved the old flash bios, then flashed with Zarniwoop's EBC ROM. Man, that was the longest flash I ever had to sit through! Luckily it didn't report any errors.

    Used Boot Camp control panel to boot back to Mac, shut down.

    Powered on machine, waited for RAM check, then BAM! Grey screen followed by Apple and twirly wheel! Yeah!

    Card is working just fine even though it's a newer revision. HWSensors says the new card is pretty much pulling the same wattage from slot 1 and power supply as the other card did. It's also chewing up BOINC GPU work units as fast as the old card did too.

    Thanks again Mr. Zarniwoop! :D
     
  6. Mr. Zarniwoop thread starter macrumors demi-god

    Mr. Zarniwoop

    Joined:
    Jun 9, 2005
  7. Earl Urley, Nov 24, 2014
    Last edited: Nov 24, 2014

    Earl Urley macrumors 6502

    Joined:
    Nov 10, 2014
    #107
    Thanks, all I did was swish around some money, still haven't tried removing the R17 resistor, figure I'll just get a good X-Acto blade and be reeeeeal careful one of these days.

    Rebooted again to make sure I would get the option boot screen, and yep, I got it. The layout looks the same as the 1.0, there are some small modifications to the coolant piping and fan layout, the board looks much the same with some components shifting only slightly.

    I have a bunch of pics I took of the older card to compare to the new, but I figure as long as v1.1 works who cares, really? If I get another really cheap (there are used ones for less than $180 on Amazon, a couple used on eBay for $150/160) and it burns out or something (or if this one does) I'll keep it around for spare parts cause I bet that heatsink / fan combo would be pretty hard to find and you know that's probably what's going to fail first..

    This is something of a surprise as I've often seen with older AMD Radeons like the old 9500 and 9600's that if you got a new board or a new configuration that only was altered by a few capacitors or RAM chips you were SOL, (remember the bad old days where a rectangular RAM chip was bad, but a square Hynix was good?) however it could be that the MSI is playing its cards close to the vest and doing hardware tweaks that don't require them to recode their firmware (for if they did you can bet that EBC ROM wouldn't work.)

    Also adopting AMD's dual BIOS feature they probably found that they had to include a lot of flashable memory, unlike the bad old days where PC card vendors tended to skimp as much as they could and only put 32K of flash RAM because they didn't give a spit about dual booting.

    Just for kicks, here's what HWSensors is reporting (and it's under BOINC OpenCL load of two tasks running simultaneously on the GPU):

    PC Slot 1: 33.14W
    PCI Booster A: 71.28W
    PCI Booster B: 28.44W

    So total of a little over 138W so far, fingers crossed.
     
  8. Zwhaler macrumors 604

    Zwhaler

    Joined:
    Jun 10, 2006
  9. Denis Ahrens macrumors newbie

    Joined:
    Oct 10, 2014
    #109
    try the furmark benchmark for some minutes. then you see if everything is working. your numbers are way too low for a real test.

    Denis

    PS: http://www.geeks3d.com/gputest/download/
     
  10. Mr. Zarniwoop thread starter macrumors demi-god

    Mr. Zarniwoop

    Joined:
    Jun 9, 2005
    #110
    I am impressed by your jumper wire patch to evenly distribute the power load of an R9 280X which lets you run furmark without external power or undervolting/underclocking the card.
     
  11. Earl Urley macrumors 6502

    Joined:
    Nov 10, 2014
    #111
    Haven't done the Furmark test yet because I'm a little concerned about burning it up, maybe I'll do that wire mod before I attempt it.

    One thing I will say also is that the card smokes, so far, a GTX 660Ti running CUDA GPU tasks for a BOINC project; if and when it catches up to the GTX 570 also running BOINC I will be pretty impressed.
     
  12. theitsage Suspended

    Joined:
    Aug 28, 2005
    #112
    I ran the Furmark benchmark stress test you suggested and my 5,1 didn't crash. It ran for close to 30 minutes. My R9 280X is the same as yours.

    For the 8 pin power cable, I connect it to PCIe booster A (which draws 7.99A when Furmark is running). The 6 pin power cable is connected to PCIe booster B (3.03A or so with Furmark).

    Mr. Zarniwoop, have you run Furmark?
     
  13. mikeboss macrumors 65816

    mikeboss

    Joined:
    Aug 13, 2009
    Location:
    switzerland
    #113
    it's possible that the card draws more than 7.99 Amps! this is just the limit that the sensor will display...
     
  14. theitsage Suspended

    Joined:
    Aug 28, 2005
    #114
    Very likely. I ran the Furmark stress test again while using HWMonitor for Wattage readings.

    Power Supply 307.56W
    PCIe Slot 1 34.73W
    PCIe Booster A 96.30W
    PCIe Booster B 35.73W
     
  15. mikeboss macrumors 65816

    mikeboss

    Joined:
    Aug 13, 2009
    Location:
    switzerland
    #115
    this still might not be the whole truth: Watts displayed are just 12 Volts * 7.99 Amps -> more or less 96 Watts.
     
  16. MacVidCards Suspended

    Joined:
    Nov 17, 2008
    Location:
    Hollywood, CA
    #116
    Oh, it can definitely show higher numbers than that.

    I have seen a GTX285 post numbers that give you that puckering-up feeling. I have seen more than 120 Watts from a single cable before. To see highest possible reading you need two displays.

    Run Furmark full screen on one and leave HW monitor running on other. Furmark in a small window won't draw as much as full screen. A loud click accompanied by sudden darkness means you passed the magic threshold. Pull plug and hold power button down for 5 seconds or so to reset.
     
  17. Denis Ahrens macrumors newbie

    Joined:
    Oct 10, 2014
  18. Earl Urley macrumors 6502

    Joined:
    Nov 10, 2014
    #118
    Is it a version 1.1 board or a 1.0 board?
     
  19. RCMan macrumors newbie

    Joined:
    Nov 30, 2014
    Location:
    North Carolina, US
    #119
    If I get the R9 280X will it boot and everything without flashing like Zarnis??

    I don't really care about boot screens.
     
  20. RCMan macrumors newbie

    Joined:
    Nov 30, 2014
    Location:
    North Carolina, US
    #120
    On you site, you say:

    "The newest AMD card available from MacVidCards, the Radeon R9 280X. Capable of running up to 3 displays in your 2008 or later Mac Pro, the R9 280X will meet all of your monitor needs and ramp up your graphics performance."

    I have a Mac Pro 1,1 and I was going to purchase that card from your site but I need to make sure it will be flashed for my Mac which is 2006 (1,1).

    Is it compatible?

    Cheers.
     
  21. Earl Urley macrumors 6502

    Joined:
    Nov 10, 2014
    #121
    Today, the average score on the BOINC client running on my 280X and OpenCL finally passed the the average score of the BOINC client running on an EVGA GTX 570 with CUDA installed.

    40,637.57 on the R280X, 40,237.76 on the GTX 570. Both are installed in Mac Pros where the CPU is doing none of the work and are only running GPU tasks only. There was a stock OEM ATI 5770 on the R280X machine before; it only took about two weeks for the 280X to catch up to and better the average score of the GTX 570.

    Depending on how much higher the average score gets, if it tops off within 1000 points, I'll just hang onto the GTX 570, but if it beats it by more than 15-20% I'll probably just pick up another 280X and consign the GTX 570 to a backup. They both use both power inputs anyway.
     
  22. MacVidCards Suspended

    Joined:
    Nov 17, 2008
    Location:
    Hollywood, CA
    #122
    Yes, but you need to be using a newer OS then your machine natively supports.
     
  23. Ph.D., Dec 4, 2014
    Last edited: Dec 4, 2014

    Ph.D. macrumors 6502

    Joined:
    Jul 8, 2014
    #123
    OK, it's my turn to ask for help.

    MSI 280X Gaming 3G, unflashed, no R17, on a 3,1 Octo-core 2.8GHz.

    The situation: Everything works well, except I'm not remotely getting the sort of performance that Mr. Zarniwoop, etc., seems to be getting. Indeed, except for luxmark, the MSI card is actually slightly slower than my OEM 5770.

    Measure 5770 280X
    -------------------------
    Luxmark 420 845
    Cinebench 39.3 38.7
    Valley Ave. 15.8 15.3

    R17 made no significant difference in the above. Subjectively, I see no particular difference between the two in games, either.

    Temperatures and fans are fine, stability is perfect, shows up as AMD Radeon 7xxx in System Information and AMD Radeon HD Tahiti XT prototype in LuxMark, etc.

    The only odd thing I notice is that in System Information, if I click on PCI Cards, it states that there was an error while gathering PCI card information. There are no other PCI cards installed. The 5770 is reported correctly. However, I do not know if this means anything given that it's not an Apple-blessed card.

    Any ideas?
     
  24. MacVidCards Suspended

    Joined:
    Nov 17, 2008
    Location:
    Hollywood, CA
    #124
    All well re-hashed issues.

    Netkas figured out issue with 3,1 and 7xxx cards, you need to yank 2 kexts and card will gain 40% performance.

    Netkas.org forums, read up.

    Having wrong name and no info in PCI section is standard behavior for non-flashed cards, also old news. (see title of this thread, for instance)
     
  25. Ph.D., Dec 4, 2014
    Last edited: Dec 4, 2014

    Ph.D. macrumors 6502

    Joined:
    Jul 8, 2014
    #125
    Ah, yes, I remember the 3,1 kext issue being discussed. Thanks for reminding me.

    To save others trouble, go to this page, starting most of the way down:

    http://forum.netkas.org/index.php/topic,8206.180.html

    Luxmark went from 845 to 2234 (card only).
    Valley from 15 to 40-ish (couldn't be bothered to wait till the end).
    Cinebench unchanged for some reason.
    A ridiculous improvement in game performance.
     

Share This Page