"The motherboard limitation is for all general purpose motherboards. Some vendors like ASUS supports maximum 7 GPUs, some can support 8.
All GPUs requesting IO space in the limited low 640K RAM. The motherboard BIOS allocated IO space first for the on motherboard peripheral and then the space that left can be allocated for GPUs.
To be able support 7-8 GPUs on the general purpose motherboard sometimes requested disable extra peripherals to free up more IO space for GPUs.
The server type motherboards like Super Micro (for example X9DRX+-F) can support 12-13 GPUs in dual CPU configuration. It is possible because Super Micro use on motherboard peripheral that doesn’t request IO space."
now that seems a problem =) not so easy to find a way around or I'm missing something?
Here're the answers from Amfeltec [see my earlier posts on pages 10 & 11, above]:
1) First, unsolicited Amfeltec caution:
“Please note that the general purpose motherboard can supports maximum 7-8 GPUs.
If you are planning to have more than 7-8 GPUs on your mother board we recommend
to confirm with motherboard vendor regarding support this numbers of GPU.”
2) (A) My first query following caution:
“I forgot to ask you a question that I have regarding your cautionary statement about general purpose motherboards being able to support a maximum of 7-8 GPUs. Are you aware on any particular motherboard(s) capable of supporting up to 16 GPUs and, if so, who is/are the manufacturer(s) and what is/are the motherboard model(s) that you are aware of that can handle that many GPUs? Since I’m a 3d animator who uses GPUs for 3d rendering, that information is very important to my purchasing decisions.”
2) (B) Amfeltec’s response:
"The motherboard limitation is for all general purpose motherboards. Some vendors like ASUS supports maximum 7 GPUs, some can support 8.
All GPUs requesting IO space in the limited low 640K RAM. The motherboard BIOS allocated IO space first for the on motherboard peripheral and then the space that left can be allocated for GPUs.
To be able support 7-8 GPUs on the general purpose motherboard sometimes requested disable extra peripherals to free up more IO space for GPUs.
The server type motherboards like Super Micro (for example X9DRX+-F) can support 12-13 GPUs in dual CPU configuration. It is possible because Super Micro use on motherboard peripheral that doesn’t request IO space."
3) (A) My last relevant query:
“Regarding the motherboard GPU limitation for all general purpose motherboards, does it matter whether the graphics cards are dual GPU cards (like the GTX 590, 690 and Titan Z)? In other words, would five GTX 590s, 690s or Titan Zs (all dual GPU processor cards) be counted and thus be subject to the same limitations posed by ten GTX 570s, ten GTX 670s, or ten GTX Titan Blacks (all single GPU processor cards)?”
3) (B) Last relevant Amfeltec response:
“I think it will be no difference. You can check it by plug in each board (one CPU and dual CPU) to the motherboard and in Device Manager check the size of the IO space resources requested. I think it will be the same. Very important that GPU board requests just one interval or multiple intervals in one 4K page. The IO space is the legacy request for GPU (support VGA mode) and I think this will be the same for 1 or dual [G]PU ar[d]s.”
Last edited by Tutor on Wed Jan 07, 2015 7:35 am, edited 1 time in total.
Because I have 180+ GPU processers in 16 tweaked/multiOS systems - Character limit prevents detailed stats.
1) Setting PCI Rom Priority to EFI Compatible Rom;
2) Setting progressively larger values for PCI Latency Timer;
3) Enabling 4G Decoding;
4) Enabling PERR# and next SERR# Generation;
5) Setting progressively smaller Maximum Payload values; or/and
6) Setting progressively smaller Maximum Read Requests.
I will perform these tweaks, each separately at first, to try to solve any GPU recognition issues, and if that is unsuccessful, I will make tweaks in combinations.
hope You'll manage to make GPUs work. I believe, this is very interesting for some users,
who consider building higher end machines to avoid separate boxes.
thanks for sharing all the info/findings =)
& hope these "roadblocks" (GPU limit)
will be fragile enough to crush! =)
Now there's room for soda pops. I'm being facetious because they'd be hard to reach and I wouldn't want to risk their exploding or otherwise leaking. But, there's no room for Alphacool NexXxoS Monstra Dual 120 mm Radiator that was originally inended to go into the upper portion of the freezer. Now I'll store Silica there to control ambient moisture. I underestimated the room that would be required by the piping for the coolers in the freezer. As a result, the Monstra will become an room air cooled unit, placed in-line between the 2 extremely large Mo-RA3 radiators.
Here are the main reasons why I chose to run the piping for the four Aquacomputer Airplex Modularity System 240 Radiators - now the only freezer-based chillers - through the door:
1) the door is easily replaceable;
2) because the door is easily replaceable, it's also one of the least expensive parts of the freezer unit for me to replace should I decide to use the freezer for it's manufactured intended purpose;
3) that location (the door) is more convenient because of where the chillers sit, the orientation of the connectors on the chillers, and when I open and close the door, it's easily for me to control the placement and behaviors of the piping and to gain access to the quick disconnects to remove the chillers completely from the freezer;
4) getting the piping through the door posed no risk of damaging some important (though hidden) essential part of the freezer because the door obviously functions only as a door (it's connected by hinges) and I could easily see that no essential electrical/cooling part was run through it; and
5) the portions of the door that I cut into are made of plastic, was easy to cut and didn't take long to cut into and modify and it was easier to work the piping through the door.
Additionally, as I got farther into this project, I found that I had to build a home for the cooling components and that pushed everything else off schedule.
Here're some pics of Mr. Freeze(r) at various stages.
Next comes the wiring; then the final piping; then piping and wiring management/aggregation for more visual appeal, then finally the 24 hour no-leak pre-testing.
You do not have the required permissions to view the files attached to this post.
Last edited by Tutor on Wed Jan 07, 2015 7:30 pm, edited 1 time in total.
Because I have 180+ GPU processers in 16 tweaked/multiOS systems - Character limit prevents detailed stats.
smicha wrote:Tutor,
I wish you the very best in 2015 and hope to see Mr Freeze in action soon
Smicha,
I wish you and your immediate family the very best for 2015 and beyond. That also applies to your extended family and that includes billions.
I hope that you can see Mr. Freeze(r) in action by the end of this week.
That name - Mr. Freeze(r) was partially inspired by a favorite movie of my son - Mr. Freeze and Batman (and, of course, it is a freezer). One day when I came home from work, my son, who was then about four years old, put a video tape of the movie in the VCR; pressed play; then stood in front of the TV screen and proceeded to say every characters' line for the entire duration of the movie. That was the second or third tape of that movie because the earlier ones had self-destructed from my son's previous overexcessive play. I've also been inspired by (and a fan of) Arnold Schwarzenegger, as well as many other top body building contenders, to continue developing my body throughout my life's journey. I see computer modding as having many parallels to body building.
Last edited by Tutor on Wed Jan 07, 2015 7:34 pm, edited 2 times in total.
Because I have 180+ GPU processers in 16 tweaked/multiOS systems - Character limit prevents detailed stats.
@Tutor, this is so Cool man! (obvious repsonse, but I could not resist)
But really... respect!!! And very curious where this will end.
greetings and all the best for 2015 and beyond
)