Recent Posts

I have a AMD 7800m willing to ship and pay to have it flash if it works. If it doesnt you can keep the card.

Reporting back...

I've done the surgery and the MacPro is working just fine now.
SMC was detected as a 1.39f11. Fans are running normally.

Success!

Hello!
Can you explain in more detail how to flash SMC?
Thank you with respect.

I think he just swap the chip physically, but not flash it.

on this site, look for thread on AMD 7000 series

where can i find the display definitions in the vbios and how to add the efi part and fix the checksum

The new driver for HS is WebDriver-387.10.10.10.40.108. I didn't find the NVDAGV100HalWeb.kext that was in the WebDriver-387.10.10.15.15.108  which is needed for the Titan V (Volta) so I'm not sure if the titan V will work. It may work using the newer driver by changing the version in NVDAGV100HalWeb.kext to  the latest version number. I haven't tried it yet. No time. I have gotten my 2018 MBP to work with the titan V(sonnet breakaway 650) . I have uploaded my Geekbench benchmarks for my 2018 MBP. I'm having trouble getting my iMacPro to use the titan V. Its recognized in the system report & so far I have only got  Geekbench to recognize it. I have uploaded the one titan CUDA benchmark I ran on my iMac Pro 64 meg, 10 core, Vega 56. My gtx1080 is recognized and used..

Can this help anyhow?

Hi, I have some experience testing MXM cards with iMacs and have created several custom roms to substitute existing cards.  Mojave has obviously increased demand for newer cards supporting Metal so I started looking into that a few months back.  Sadly the A1312 iMacs have big problems initializing any AMD cards after the 6 series. For some unknown reason they require a soft reset before the PCI IDs are fully detected, and even then they flatly refuse to fire the back-light, which is annoying when you can actually see an image on the LCD using a flashlight!  The initialization problem seem like a real show-stopper, I suspect it's related to "Enduro" power management, the iGPU display output multiplexer, or most likely the way AMD started hashing certain values in their newer vbios?  I think we might just have to admit defeat on these. 

Initially the Nvidia cards seemed more co-operative, but for their enigmatic DCB table...  I have the 680 / K5000 EFI loading with boot-screens on some external displays but I'm very curious to know how you fixed the corruption on the internal LCD??  No amount of futzing with the DCB table seems to help with that.  I've even tried modding the VESA tables, LCD centered timings, Display Port drive-strength values etc.   The EFI seems to have a BIT section to, maybe that needs some attention?  Am I getting warmer?   

I also tried getting some of the 2013 iMac's Kepler vbios to work on MXM cards, just naively under-clocking the memory wasn't enough as they seem to be structured quite differently, missing power tables etc?  I was hoping this might shed some light on the brightness control problem as they have a much simpler DCB table in comparison.  I did actually get the 3782 EFI to load but this just crashes the display driver so more research needed there.  I just posted about that in the 2080 thread, it could have implications for the MacPro to so thought you might find it interesting.  I'm away from home for the next week but was planning on getting stuck back into this soon.  Just let me know if I can help with anything, I still have an EEPROM programmer wired into my test machine so that should save us some time at least.   

What is the software "nVidiaInfo" is used on your screenshots?

GTX 765M is reported to work in iMacs (without bootscreen and brightness adjustment): https://www.insanelymac.com/forum/topic/333956-gtx-765m-in-imac-111-late-2009-display-brightness-fix/

$120 : https://aliexpress.com/wholesale?catId=0&initiative_id=SB_20181103045637&SearchText=NVIDIA+GTX765M+2GB

GTX 660 $ 135 : https://www.aliexpress.com/item/Wholesales-nVidia-Geforce-GTX-660M-GTX660M-N13E-GE-A2-for-Dell-Alienware-M17X-R4-M18X-R2/32475001442.html?

Does it have Titan V support or not?

looks based on same code as 105.

10.10.10.40 and not 10.10.15.15 which was the unstable 108. that's encouraging. was worried they'd use that unstable code for the new one.