Paul King
Forum Replies Created
-
This is one long thread full of people who do not have the full story when it comes to enterprise vs desktop.
Most unreliable enterprise drive – Seagate SCSI. Had our greatest percentage of failures with Cheetahs both 10k and 15k.
Back then there was a clear difference between enterprise and desktop – SCSI vs PATA.Along comes SATA with a better connection than PATA and much better capacity and potential than SCSI.
Controller manufacturers adopted unified controllers early (SAS/SATA) and back then (2005) none of them had an issue with desktop drives for RAID. David is right, most workstation storage box vendors used desktop drives and most were Hitachi.
Back then reliability was more an issue of where you bought the drives than what model or brand they were. We bought all our drives through Synnex and we had a 25% failure rate, regardless of brand. They mistreated the drives in transit (a WD rep who witnessed this told me).
We switched to different supplies and the failure rates dropped to 1%.Around 2005 we used desktop drives in RAIDs, most predominantly WD. At the demise of SCSI and after the rise of SAS, the drive venders brought out enterprise variants with SATA interfaces. AT that time they started to BS about enterprise drives. WD took it one step further and crippled their Black series to stop people using them in RAIDs. Adaptec had 2TB FAUX Blacks on their compatibility list, I bought 24 of them, had issues and so Adaptec took them off their compatibility list. WD could have fixed them but sighted that they were desktop and not suitable for RAID. I hacked their firmware and made them work anyway, but never bought another WD drive again.
We switched to Hitachi. Bob – the desktop drives 7kx000 were rated for 24/7 use not 8/5. That’s why we went that way, and they worked fine until WD finished them off.
So why do we have enterprise drives? Greed. Same reason Avid sold Newscutter for $1000 more than Media Composer – “they’re broadcasters so they will pay more for it”, direct quote from Avid.
Everyone here that has advocated enterprise drives – go to every vendors compatibility list and have a look at how many desktop drives appear. Most of them are early models (2005-2008) but Adaptec and LSI have current desktop drives on their lists. They preface this by saying they recommend enterprise though.
Vibration? A current desktop drive would have less operational vibration than an enterprise drive from 3 years ago.
WD Red – what a crock of s… Only up to 5 bay NAS – what BS and WD here in Australia contradicted this during their sales tour.
Only fundamental difference is TLER (WD term) and this is a real issue. However from 500 Hitachi drives in RAID 5 we have only had 3 drop out from a TLER event. Solution? Put them back in and let the RAID rebuild. However if I use desktop drives now (Seagate Barracuda is the only model left) then I set them up as RAID 6.
How many guys here have had to rebuild a 60TB RAID5? It’s a nervous wait whether it’s desktop or enterprise so RAID6 is the way to go.To sum up, enterprise is a crock and motivated by greed. The physical drives are likely the same (albeit the enterprise probably test better out the factory) with only the firmware difference. It’s an excuse for making money and probably a sound corporate strategy with so much storage going cloud based.
So if you’re on Mac it’s Areca (they are basically LSI controllers with a Mac driver) and Constellations. If you’re on a PC I’d do Adaptec 7 and Barracudas. My most recent RAID is Adaptec 7 with 16 Barracudas in RAID6. 2100MB/sec from desktop drives seems good enough for a workstation. Beyond this I wouldn’t bother with mechanical HDD. Adaptec 7 with SSd is good for 6GB/sec.
So everyone here has been correct in part, but there is still a lot of misinformation which only strengthens the HDD vendors bottom line.
-
Frankly everything said here has been inconsistent and at times conflicts with Resolve published information.
The official config says 1 x GPU use 680, 2 x GPU use 680 or 1 x 690 (although they are not recommended due to a lack of RAM – they are still on the list for 4k) 4 x GPU use 580. So they don’t recommend 680 for 4 x GPU. Problem with the config is they say what to do but not what not to do (same here in this thread). They say use MB for 2 x GPU and expander for 4 x GPU but not why that is. As far as Davinci are concerned, more than 2 x GPUs are not recognised, however we have Al here saying that they can be.
Also, config says going from 1 x GPU to 2 x GPU is a double in performance (doubling of CUDA cores). It also says don’t use Quadro unless you have to as GTX is far better performance. So why? All that they have over Quadros is CUDA count. Look at the new Quadros, doubling of CUDA cores. Why?
Nvidia could not justify the price hike for Quadros and give us a bunch of marketing speak as to why they are better. But here we are with one of the most professional and GPU intensive apps in the industry and consumer cards are recommended.
When it comes to CPU, if I double the core count or the speed I get a equivalent increase in render performance. However this is only in an app that can make use of multiple CPUs. So if you ask me the question about doubling CPU cores, I can answer it concisely, but qualified by the particular application being used. So the same Q relating to AE is a little bit more complicated.
All I have from this thread is cryptic answers. Nothing here has been useful for building a Davinci in June 2013, except Als suggesting about multi GPUs and the SLI setting.
BTW Al you can compare AMD and Intel CPUs, we always could knowing the architecture and I was always able to explain it to someone. But it’s a moot point now as AMD will new catch Intel in the performance space.
-
OK
I take it from the responses that not many here have enough experience with this hardware.
Win/Mac limit 4 GPU, so bigger expansion chassis makes no difference.
Titans in the release versions all come from the one factory so no variation between manufacturers for the moment.I take it you guys are all Mac guys (with the one exception).
The Supermicro board is specified by Davinci for the purpose of multiple GPUs. The notion that GPUs work better in an expansion chassis doesn’t make sense as rhe expansion chassis has less bandwidth than the motherboard.If there are formulas bandied around such as 1.4 x 680 then the question is relevant comparing CUDA cores. Blackmagic here in Australia have said so to us, they just haven’t released any testing results with the newer GPUs.
It makes no sense to put in as many Titans as possible as they are expensive and there is obviously a limit to the number of CUDA cores required for 4k. My question was how many – 4 x 680 = 6000 cores.
Dont want to seem too ungrateful for the replies but they’re not very helpful for the original question.
Thanks
Paul
-
Hi Juan
Well this holds true for the 690, however my question was in context of RAM on each card. Peter from Blackmagic has already informally qualified the Titans, but the issue here is we have no info on how many are require for 4k work.
All we have so far is 4 x 680s for 4k.
I’m simply asking how many 780s and how many Titans.
Maybe I’ll be the first guy to find out if no one else knows.
But this has become a significantly important question now that we have new GPUs available in 2013. An expansion chassis has always been a very expensive option and only serves to house what will become redundant technology.Thanks
Paul
-
Hi Andrew
PC – Win 7.
WWDC disappointment?
Thanks
Paul
-
Hi Ramin
I’m not a fan of Dynamic link, I got burnt by it when it first came out.
The other issue is that there is double rendering, ram preview in AE and re rendering in Premiere. However if you ram preview in AE and then export that file, it uses the ram preview and so you’re not re rendering.So you put the render file in the Premiere sequence. If you want to change it, go back to AE and re render over the op of that file. Go back to Premiere and it’s updated.
So this works like dynamic link, but you’re left with rendered files rather that dynamic links and you’re ready for Resolve.
From there you can just trim the project and then export that to an FCP XML, which can then be imported into Resolve.
-
Hi Kevin
can you check your CPU usage in task manager while you’re exporting?
See if it’s low – say around 25%. If this is the case, turn off GPU rendering in Project settings and retry. -
Brian
Don’t you hate it when people who don’t know the correct answer, give you an answer like they do.
The correct answer is that you can. When you round trip to Adobe Media Encoder, a copy of your project is made before the encode commences. This has to happen so the sequence is locked from changes.
So making a copy would be doubling up this process.
Take this example – you have a sequence with 10 commercials in it, each at 1:30, 2:30, 3:30. You have to export client dubs and station dubs. All you have to do is set the work area, dial up the correct export settings and send to AME. After each export is sent, you can set up the next export.
Hope this post actually helps.
-
Paul King
January 26, 2013 at 2:08 am in reply to: Black Shadow layer not rendering out when exporting?Hi Dan
It’s in Project settings. There’s a drop down under Video Renderer and Playback.
Thanks
Paul
-
Paul King
January 26, 2013 at 2:04 am in reply to: Premiere CUDA acceleration does not work correctlyHi Ryan
There is no emotion in these emails only facts.
I beta tested for Adobe for 6 years. They fix bugs that are easy and avoid comment on bugs that are hard. They had to be badgered to fig important issues.Avid may have it’s bugs, but when it comes to the quality of the visiual output they are mostly flawless. Mostly the same for FCP.
However Premiere is full of quality issues and they never get mentioned.
I am posting these issues to raise awareness and put pressure on Adobe to fix them (something that didn’t work in beta).Quality issues like inaccurate transparency issues, non-linear dissolves, software DVEs, issues with fields etc. These are all things that they ignore.
Ryan I dont think it’s product to talk about other options and products when discussing significant issues with Premiere.