Alex Gerulaitis
Forum Replies Created
-
[John Young] “1)-https://forums.adobe.com/thread/662972 This should answer your question, at least it did for me.”
John,
I personally believe some of Harm’s recommendations to be flawed. The discussion on the merits of more than two volumes can be found here, and you will see that the consensus is not in Harm’s favor.
Harm is a moderator of the forum and someone with a (generally well deserved) reputation for his knowledge. It does not help here, given that his advice often resides in locked topics as if it’s Adobe’s official position.
-
[Jonny Webb] “I dont trust the Adobe forums anymore, not because there are so many forum squatters with such fragile egos, but because many of the regulars talk outside their field. They may be experts on specific products, but their IT knowledge is often flawed to the point of dangerous. Numerous times i’ve checked there for a hardware question, only to receive an incorrect answer.”
I hope I am not one of them. 🙂 (Trying my best to keep to my field.)
That said, it’s my impression as well. Advice is often given unequivocally and with no regard to workflow requirements.
Highly questionable and also unequivocal opinions such as “ECC is slow and expensive” (compared to non-ECC), “Quadro is a waste of money”, “Xeons are a waste of money unless you edit 4K” and so on – are all over the place, and often come from forum “heavyweights” – people praised for their knowledge.
It kinda is dangerous.
Alex Gerulaitis
Systems Engineer
DV411 – Los Angeles, CA -
The only way to know is to monitor your memory usage and see if you bump against the limit often enough. That said – I am with everyone else, you can (almost) never have enough memory, especially if you are a heavy Pr user.
-
[Jud Johnson] “Thanks brother! Do you know if the K5000 will work with the Tesla C2075 in the “maximus” configuration?
It’s not clear on the Nvidia website wether it does or not.”Don’t know but from all the signs (including what Greg said), I’d stay away from it with K5000 and wait for NVidia to release a new Kepler-based Tesla card, and see if they support Maximus on those.
Looks like (from NVidia specs and from what Greg said), Maximus is a pretty strict Fermi+Fermi coupling, i.e. unlike that with humans, the consenting cards have to be from the same family. Kepler+Fermi is unlikely to work.
Alex Gerulaitis
Systems Engineer
DV411 – Los Angeles, CA -
Sanrj,
You are comparing TWO 12-bay units that you paid more money for than the 16-bay unit Warren described, and saying your configuration is “much better”?
“Much better” how, exactly?
Maybe your configuration works for you, and this is the best purchase you ever made – for your purposes. Why walk into someone else’s thread and say that your stuff is “much better”? Especially that so far, there is little evidence to that?
Did you also not see that Warren’s setup is optimized for shared editing using 10GbE cards on client machines, something you haven’t even tried on your system?
Alex Gerulaitis
Systems Engineer
DV411 – Los Angeles, CA -
[Sanrj Patil] “i did not want to reinvest in 5 year so Fiber was my choice”
This, too, is a strong statement suggesting SAS is a dead end. Care to back it up?
Alex Gerulaitis
Systems Engineer
DV411 – Los Angeles, CA -
[Sanrj Patil] “they striped together. Raid 5
1TB seagate constellation sata
Fiber card – emulex LP-12004 M8”This is what I am not getting. How do you “stripe”, or more correctly, connect two identical quad-channel units to a single quad-channel card?
Further, each one has its own controller, each would have its own RAID5 volume, unless you actually have a main and and a SAS expander unit – which is a wholly different animal from what you described originally.
Alex Gerulaitis
Systems Engineer
DV411 – Los Angeles, CA -
[Jud Johnson] “Any idea when this k5000 will be released? I’ve asked many Nvidia guys on different forums but everyone keeps dodging my question (very annoying). Anyone on this thread know?”
Newegg says 10/24 for the PNY version.
It’s released and in the supply pipeline, authorized distribution has the SKUs in their systems but no ETA yet. I’d guesstimate about two weeks. Three SKUs I am seeing so far: one PNY and two HP versions (smartbuy and regular). Surely Dell and Lenovo will follow suit shortly.
Alex Gerulaitis
Systems Engineer
DV411 – Los Angeles, CA -
Here’s their response:
“We are still looking into the issue. Our engineers have not seen consistent results making it difficult to resolve.”
I’d encourage sending your benchmarks with slow(er) read speeds to Areca support and/or publishing them here, so that Areca engineers can see a more consistent picture of the problem.
Alex Gerulaitis
Systems Engineer
DV411 – Los Angeles, CA -
[Sanrj Patil] “I got it two FiberRaid AR21F8e for $6243 with 12TB each when they had 50% off few months back
I am running raid 5 write 2200mb read 1800mb much better then is316js “Each one does 2200MB/s read speeds, or are they striped together? How are they connected and to what? Shouldn’t the total also include a Fibre Channel card?