Im freelancing out of a major post house, and came across a weird issue of running out of VRAM. after investigation it seems by policy this company configs their flames in a very unusual way! my machine has a 24GB card as the secondry card and a 8Gb gpu as the primary! i am thinking this is batshit crazy, and im not confused as to why my machine is running out of Vram. but the head engineer told me this cannot be swapped, the Second card needs more vram, and that RGS remote working puts an extra strain on the cards. Have any of you see any method to this madness? i feel like every step of the way lately our speed and efficiency has taken a massive hit and a backseat to ease of configuring a facility. When i use my local flame i am flying! when i use any of the large studios whether i was in the studio or not i feel like im going along at a snails pace. it would interesting to hear if anyone thinks theres a valid reason on putting the least powerful card as the primary…
That is insanity.
i thought so!! but you never know what bigger picture you are missing, its not my job to run a big facility but it doesnt make sense!
thank you Alan
Are they using Resolve on the system as well?
A few versions back (14 or possibly 12) Resolve performed better to have any GPU running the display but then have a more powerful GPU that you specified for compute. Was madness and has not been the case for several years now but I wonder if some tech has hung onto that from several years ago?
Either way, if you plug your monitor cable into the more powerful card Flame will use that card I believe. Haven’t been in that scenario though so can’t say for sure.
purely a flame machine. i suspect they may have done this when we went remote. maybe someone from HP told them to do that… The head of engineering told me the application will use all of what is installed. i did not think that was the case. If i load a batch into memory and it has models or other memory intensive things it doesnt just use the second GPU. as far as i know the second GPU only kicks in when we do a background task and that the 2nd card is configured to be used
Yeah man. That’s fucked. Someone did that machine (and you by association) dirty. That 24gig card should be primary… literally nonsense to leave it for background reactor only. That’s not even a supported config.
thanks Christopher for weighing in . i appreciate it. I can be a little more certain in my gut that its totally bonkers. All i can do is point it out to all the heads of dep that its the wrong way to set up flames. i wonder now if they are all configured as such or jus this one machine. im amazed how this has slipped through the cracks!
is the 8Gb card a quadro and the 24G card isnt? (like maybe its a titan?)
There is a thing with HPZ /RGS and dualscreen AVC and quadros or something. But I always get confused if that was windows host only
Anyhow this config sounds crazy.
Both Quadros, M4000, and a M6000
Yeah he did say RGS is a bit screwy and they were looking for another solution that RGS but it sounded like he configed these flames for a while like this… im going to talk to the head of 2d when he gets back from holidays…
RGS is great…what issues are they having?
That is a bonkers config for Flame. You should just pull the M4000 out.
100% agree with the others here, Flame needs to be running on the M6000 24GB… Personally I’d just take the M4000 out of the system, it’s just burning up power for not much good, since trying to run Background Reactor renders on it is likely to result in VRAM-limited failures anyway.
I don’t have a lot of expertise to share with HP RGS for Flame, except some anecdotal reports from an artist who said they preferred the experience after switching to Teradici Graphics Agent since more of the standard hotkeys seemed to be working for them (getting all the hotkeys Flame is used to having to go through transparently with any remote access solution is a challenge).
From our experience all the hotkeys work as normal with RGS.