I've been running last few days without problem at 1140 MHz. I saw specs for M5500M (same GM204 GPU) and are:
Base Clock: 1140 MHz
Boost Clock: 1165 MHz
So I guess that's the safe limit zone.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
iM 4,1-14,4 iMac 2011 Maxwell and Pascal GPU upgrade
- Thread starter G3llings
- WikiPost WikiPost
- Start date
- Sort by reaction score
- Status
Base Clock: 1140 MHz
Boost Clock: 1165 MHz
So I guess that's the safe limit zone.
I ran games for about 3 hours at 1175mhz and seem to be stable. Maximum core temp 65c and 78c hot spot temp. Any other issues and I'll bump down to 1165mhz.
I found a seller at the same time have p5200 rtx4000 each need 1088 dollars, not a brand new card is dismantled card put a long time to fall ash looks dirty. The price is too expensive I am considering whether to buy.
Attachments
These cards look in horrible dirty condition 😳 , I would not buy them without some kind of full refund warranty by the seller. The rtx4000 has TDP of 110W and the P5200 TDP of 100W. Also, maybe too much money to invest for the old iMacs...
I'm still considering about buying one P4000M for testing, is the only one I can find at reasonable (but high) price.
Please tell me, does the P3000.rom file in your folder work in the iMac 27 computer?
I tried the P3000_PWR.rom file, the computer is working, the MAC OS system is working, the backlight is working via OPENCORE.
I think the only one who tried this vbios is @jay508, both roms are the same but the PWR version is modded with the Mobile Pascal TDP Tweaker. Probably both vbios roms will work the same for you.
Windows uses its own Direct3D11 API while macOS is using the former industry standard OpenGL to achieve this results. Because of this (and many other reasons) Apple decided years ago to introduce its own low level high speed graphics API called Metal/metal.
Unfortunately I know of no graphics benchmark using Direct3D11 on Windows and Metal on macOS to compare the implementation details.
UNIGINE Valley is obsolete - and nowadays only helpful when used on the same platform to compare results.
The only meaningful benchmark for the GPU in the latest MacOS is the Metal score.
And that is only applicable to AMD GPUs with graphic and video acceleration.
Thus, getting even the low end more affordable AMD GPU like WX4130 would be much more preferable than the so called high end nVidia GPUs.
And that is only applicable to AMD GPUs with graphic and video acceleration.
Thus, getting even the low end more affordable AMD GPU like WX4130 would be much more preferable than the so called high end nVidia GPUs.
My WX4170 arrived today. I attach its pictures below.
It lacks an EEPROM chip to function normally.
It will need the EEPROM chip soldered onto it, according to the experts about WX41xx in this forum.
Right over the top left hole is the missing EEPROM chip.
If yours are the same, you might need to solder the EEPROM chip as well.
Attachments
Thanks a lot for the update.
Mine was a really funny one (you can see photos)
and my Imac 2011 didn't even boot with it
so I've opened a dispute with Ali and returned it back.
Attachments
It was sold as AMD WX4170 made for HP Zbook17 G4 but as you can see it's pretty hard to say what it really is
HP Zbook17 G4 G5 amd wx4170 are blue pcb, your photo is red pcb really rare. I also purchased three hp blue pcb wx4170 from HP Zbook17 G5.
10.23 can probably receive I will shoot the front and back of the high-definition large picture to see the lack of those resistors.
I'm still looking for the right Turing rtx3000 rtx4000 t1000 t2000 mxm
Amps a2000 a3000 mxm very rare are on board bga's. For testing
Translated with www.DeepL.com/Translator (free version)
Attachments
Take a look at the PCB and check the ident string 109-C99581-01 - this is a variant of the WX4170. Mine has this ident string 109-C99581-00B_00 (blue, AMD version, including BIOS chip).
I finally managed to find a "reasonable" priced P3000 card which should arrive in the following weeks.
I also ordered some shunt resistors and smd 0805 sized ones to play with power modding the card, let's see if I can squeeze it a bit over the 75W limit.
I also ordered some shunt resistors and smd 0805 sized ones to play with power modding the card, let's see if I can squeeze it a bit over the 75W limit.
That's going to be an interesting test. Looks MXM-A and with a TDP of 50W it may work on 21'' iMacs.
If it works on iMac it should score slightly above the P3000 card.
Also, I'm curious about boot screen as Santa's mentioned.
Already pm @Santa's Little Helper. waiting for his reply, t1000 mxm-a price is more expensive than p3000 and p4000 same price, does not support macOS can only use uefi Windows10, a very risky experiment.
If I can't use it in iMac I still have hp laptop I don't know if I can use it.
All the results are in I received the test feedback results.
At the beginning is the intention to buy rtx3000 rtx4000 helpless price is really too expensive.
Translated with www.DeepL.com/Translator (free version)
If I can't use it in iMac I still have hp laptop I don't know if I can use it.
All the results are in I received the test feedback results.
At the beginning is the intention to buy rtx3000 rtx4000 helpless price is really too expensive.
Translated with www.DeepL.com/Translator (free version)
Just tested the MXM-A OEM Adlink version of the T1000 4G results are as follows, using a small two heat pipe heat sink temperature control is good performance than the P3000 can hard decode 4k 5k video.
No boot screen installed the driver internal screen also did not show all the tests are through the remote desktop view test.
Comparison of the two photos show the model is R9 M375 W5170M T1000 4G DDR6
Waiting to assist @Santa's Little Helper experiment, respectively refreshed techpowerup downloaded T1000. 90.17.42.00.3B.DELL.rom T1000.90.17.42.00.44.NVIDIA.rom gpu-z are not properly installed driver gpu-z can not be correct recognize the information of the graphics card.
The T1000.90.17.46.00.41.Adlink.rom version can install the public version driver normally Windows 10 system remote desktop passed all tests.
Translated with www.DeepL.com/Translator (free version)
No boot screen installed the driver internal screen also did not show all the tests are through the remote desktop view test.
Comparison of the two photos show the model is R9 M375 W5170M T1000 4G DDR6
Waiting to assist @Santa's Little Helper experiment, respectively refreshed techpowerup downloaded T1000. 90.17.42.00.3B.DELL.rom T1000.90.17.42.00.44.NVIDIA.rom gpu-z are not properly installed driver gpu-z can not be correct recognize the information of the graphics card.
The T1000.90.17.46.00.41.Adlink.rom version can install the public version driver normally Windows 10 system remote desktop passed all tests.
Translated with www.DeepL.com/Translator (free version)
Attachments
I know I’m too late to reverse that return. (Apparently another thread should be on my Watched list…) Your card, even if it is an engineering sample, is (was) probably fine. It is the lack of two jumpers (0R resistors) preventing it from POSTing in 2011 iMac. Marked in red in this post.
I've been testing a P3000M last couple of weeks with @Santa's Little Helper vbios, so I'll make a couple of posts with detailed info.
First, as Santas said, eeprom programming has to be done at 1.8v. Since I already had a 5v/3.3v ch341a programmer I bought an adapter from 3.3v to 1.8v. Seller did not give detailed instructions on how to connect, so after some research and getting it working I attach some detailed pictures of how to connect.
Then I tried linux flashrom, but it did not detect eeprom (that may be something specific to my card). I even compiled latest version from github to no luck. I have not tested forcing eeprom with -f -c option, maybe that works, but as I said not tested.
Then I tried latest AsProgrammer on windows. It did also not detect my eeprom type, but selecting it as W25Q40EW 1.8V made it work. I can read/erase/programm eeprom using ch341a. Pictures attached also.
First, as Santas said, eeprom programming has to be done at 1.8v. Since I already had a 5v/3.3v ch341a programmer I bought an adapter from 3.3v to 1.8v. Seller did not give detailed instructions on how to connect, so after some research and getting it working I attach some detailed pictures of how to connect.
Then I tried linux flashrom, but it did not detect eeprom (that may be something specific to my card). I even compiled latest version from github to no luck. I have not tested forcing eeprom with -f -c option, maybe that works, but as I said not tested.
Then I tried latest AsProgrammer on windows. It did also not detect my eeprom type, but selecting it as W25Q40EW 1.8V made it work. I can read/erase/programm eeprom using ch341a. Pictures attached also.
Attachments
About P3000M performance and working:
- External display works on port closest to USB ports.
- Performance is around 15% higher than fully overclocked M4000M, at much lower power consumption and temperature. On 3dMark benchmark, the card hardly reaches 60W power and stays below 70ºC temperature.
- It is VERY difficult to reach the 75W TDP power throttling, I only saw it (green lines PerfCap) a few of times on FurMark benchmark. Never saw PerfCap throttling on 3dMark, Catzilla, Unigine Superposition, GravityMark or real gaming.
Then I tested the P3000M_PWR vbios. As PerfCap is hardly reached. all benchmarks score the same, but it seems like thermal and power caps are effectively raised. I never saw green lines PerfCap limits and thermal throttling above 54ºC does not seem to happen, and GPU-Z TDP stays at around 70%.
Initially I thought card may benefit from power modding, but after testing it seems like the real reason of not so good performance is the GPU clock limit. GPU Clock is limited to 1240 MHz boost and voltage to 0.7250 V, so I see no reason to try a power mod on the card.
As a reference P5000M with same GPU has a base clock of 1278 MHZ and boost clock of 1582 MHz, so my guess is that card may benefit easily from a 15% to 20% overclock and performance. Sadly, overclocking these mobile Quadro Pascal cards does not seem an easy task, I've searched a lot and found no info about it. I've also tried copying P5000M V-P states table from P5000M vbios to P3000M vbios, and while in GPU-Z it shows the 1278/1582 MHz clocks, the card does still not clock beyond 1240 MHz on P0. I guess there is some other clock, pll or voltage table limiting max performance.
Seeing other Pascal cards with same GPU, it seems clock/voltage limit could be changed by tweaking the vbios, hope it's not a hardware limit.
Also, I tested that memory can be overclocked, but does not make a difference on performance at this low GPU clocks.
- External display works on port closest to USB ports.
- Performance is around 15% higher than fully overclocked M4000M, at much lower power consumption and temperature. On 3dMark benchmark, the card hardly reaches 60W power and stays below 70ºC temperature.
- It is VERY difficult to reach the 75W TDP power throttling, I only saw it (green lines PerfCap) a few of times on FurMark benchmark. Never saw PerfCap throttling on 3dMark, Catzilla, Unigine Superposition, GravityMark or real gaming.
Then I tested the P3000M_PWR vbios. As PerfCap is hardly reached. all benchmarks score the same, but it seems like thermal and power caps are effectively raised. I never saw green lines PerfCap limits and thermal throttling above 54ºC does not seem to happen, and GPU-Z TDP stays at around 70%.
Initially I thought card may benefit from power modding, but after testing it seems like the real reason of not so good performance is the GPU clock limit. GPU Clock is limited to 1240 MHz boost and voltage to 0.7250 V, so I see no reason to try a power mod on the card.
As a reference P5000M with same GPU has a base clock of 1278 MHZ and boost clock of 1582 MHz, so my guess is that card may benefit easily from a 15% to 20% overclock and performance. Sadly, overclocking these mobile Quadro Pascal cards does not seem an easy task, I've searched a lot and found no info about it. I've also tried copying P5000M V-P states table from P5000M vbios to P3000M vbios, and while in GPU-Z it shows the 1278/1582 MHz clocks, the card does still not clock beyond 1240 MHz on P0. I guess there is some other clock, pll or voltage table limiting max performance.
Seeing other Pascal cards with same GPU, it seems clock/voltage limit could be changed by tweaking the vbios, hope it's not a hardware limit.
Also, I tested that memory can be overclocked, but does not make a difference on performance at this low GPU clocks.
Attachments
Last edited: