Alex Gerulaitis
Forum Replies Created
-
[Chris Knight] “The GPU (2GB nVidia 540M) doubles h.264 encoding speed in Media Encoder”
How do you turn GPU acceleration on or off in AME to benchmark its effect on encoding speed?
Alex (DV411)
-
GPU doesn’t really accelerate all that much – a loaded MBP without MPE GPU acceleration will still quite fast.
A Windows laptop with Adobe-approved graphics is at least $2K, and more like $3K:
HP EliteBook 8760w XU100UT: https://goo.gl/IbTvo
HP EliteBook 8740w XT910UT: https://goo.gl/tjJiZBoth are $3K and the just-released 8760w has yet to ship. You can easily upgrade them to 16GB for around $140 or so; upgrading to 32GB will be more serious money.
So staying with MBP won’t be that bad considering the alternative. However much I hate to say that. 🙂
Alex (DV411)
-
It’s true that with MBP you won’t get the GPU acceleration part of the MPE – the rest will still work however: 64-bit and HyperThreading.
If your workflow involves a lot of things that can be GPU-accelerated, then you’d need a Windows laptop to get GPU acceleration.
Alex (DV411)
-
The card would have to have a hardware H264 decoder to make up for it missing in FCP – and I can’t think of any card that does it and is compatible with FCP7.
Alex (DV411)
-
Can’t think of any. Haven’t done it myself yet – but seen other people do it and have yet to hear of any problems.
Alex (DV411)
-
Well put. I am also impressed that Adobe does all this with an air of humility – i.e. it doesn’t go around bragging about it – they do their work and let it speak for them.
-
[Norman Willis] “But my main question is if the CPU will handle both dynamic content generation and software RAID 1 sufficiently, or if I need an inexpensive RAID 1 card (and which one)?”
There isn’t much of a difference between software RAID1 in Win7 or Server 2K8 (R2) and hardware-based RAID1 with the exception of bootable volumes. Those can’t be software-RAIDed if memory serves me right. You can stay with the software RAID1 on non-bootable volumes.
Don’t know jack about dynamic content generation in webservers so can’t answer any questions about that – sorry.
Alex (DV411)
-
You are asking hard questions. 🙂
RAID1 can’t be compared to RAID5 directly because these use different number of drives: 2 for RAID1 (theoretically more than 2 is possible – but I haven’t heard of anyone doing this except for me :)) and a minimum of 3 for RAID5.
Even if they could be compared, there is no single or simple answer percentage-wise: it depends on the (total number of) drives, the controller, and the performance profile: e.g. streaming vs. database server.
Each scenario will be different.
If we compare RAID0 or 10 to 5, then it’s a little more straightforward (but still too many factors in play). On, say, 4 standard desktop 7200rpm drives and ICH10 (hybrid) controller, RAID0 will produce sequential read/write numbers of 400 to 500MB/s, and RAID5 – 350-450MB/s reads and 150-250MB/s writes. These numbers are for empty / unfragged volumes; performance will decline to about 50% of it when fragged/full.
Phew. 🙂
Alex (DV411)
-
[Norman Willis] “Please correct me if I am wrong, but there is a serious performance hit on software RAID 5?”
Yes, at least two hits: computational (if it’s software) and disk throughput.
Alex (DV411)
-