[biznatch]

The larger the amount of RAM you want you will have to invest in ECC/Registered RAM (which is not cheap) to avoid problems.

Is this just because with more RAM there's a greater chance of something going wrong, or is there something inherent about having lots of RAM that requires ECC/Registered RAM?

[GenoMax]

http://www.overclock.net/hardware-ne...rd-8-dimm.html

64GB RAM PC will soon be reality

Ymc,

There are several workstation models (from Dell, Lenovo, HP to name a few) that can have > 100 GB RAM (this is a representative link for Dell workstations: http://www.dell.com/us/enterprise/p/precision-desktops).
For example, the T7500 workstation from Dell can have a maximum of 192GB of DDR3 RAM. If you are going to invest in a machine with that much of RAM you should consider getting a proper workstation (or better still a proper server). The larger the amount of RAM you want you will have to invest in ECC/Registered RAM (which is not cheap) to avoid problems.

[ymc]

http://www.overclock.net/hardware-news/1089713-dh-gigabyte-x79-motherboard-8-dimm.html

64GB RAM PC will soon be reality

[DZhang]

64GB should work. De Novo assembly (or the memory requirement) also depends on data quality. So you may go aggressive on quality trimming (and/or error correction) to reduce the memory requirement and speed up the assembly. Removing redundant reads is another way.

[ymc]

Hi ymc,

For microbial genomes, 32 GB should be more than adequate in de novo assembly. It should work (but a little tight) for de novo assembly of ~100-mb genomes. Anything bigger, you need more than 32 GB memories. Keep in mind, if you have few and slower CPUs, the assembly just takes more time; if you do not have sufficient memory, the program simply can not run.

How much RAM is enough for de novo ~250mb??? Is 64GB enough? I heard that there might be single CPU workstation X79 board that might allow you to put 8 DIMMs.

[ymc]

I would also consider using a fast ~64GB SATA III SSD as a dedicated swap disk, so if you are RAM limited while sequencing, operations can be swapped to disk with minimal latency. Currently, in that category, the OCZ Vertex 3 60GB is the fastest mainstream device available.

Also, if you're running Windows, consider purchasing a pair of these SSDs, a Western Digital Velociraptor 10k RPM primary HDD, and a Z68 motherboard. Use one of the SSDs for paging and the other as a disk cache with Intel SRT. You'll want to make sure the HDD and disk cache SSD (though not during initial OS installation - read Intel's SRT documentation carefully) are connected to the chipset SATA III controller (in RAID mode) and the paging SSD is connected to the Marvell SATA III controller. You can't build a faster IO subsystem than this with mainstream desktop hardware.

One last note on paging: again, if using Windows, be sure to set a fixed pagefile size of > 48GB on the paging SSD.

Good luck!

I think I am going to have a 2 SSD RAID0 setup for program and swap and then big HDDs for data.

Swap is important but I find that when I tried to run truly memory intensive programs, the frequent swapping can crash the program. So more RAM is unavoidable if your application needs them.

[DZhang]

Hi ymc,

For microbial genomes, 32 GB should be more than adequate in de novo assembly. It should work (but a little tight) for de novo assembly of ~100-mb genomes. Anything bigger, you need more than 32 GB memories. Keep in mind, if you have few and slower CPUs, the assembly just takes more time; if you do not have sufficient memory, the program simply can not run.

[looselyhuman]

I would also consider using a fast ~64GB SATA III SSD as a dedicated swap disk, so if you are RAM limited while sequencing, operations can be swapped to disk with minimal latency. Currently, in that category, the OCZ Vertex 3 60GB is the fastest mainstream device available.

Also, if you're running Windows, consider purchasing a pair of these SSDs, a Western Digital Velociraptor 10k RPM primary HDD, and a Z68 motherboard. Use one of the SSDs for paging and the other as a disk cache with Intel SRT. You'll want to make sure the HDD and disk cache SSD (though not during initial OS installation - read Intel's SRT documentation carefully) are connected to the chipset SATA III controller (in RAID mode) and the paging SSD is connected to the Marvell SATA III controller. You can't build a faster IO subsystem than this with mainstream desktop hardware.

One last note on paging: again, if using Windows, be sure to set a fixed pagefile size of > 48GB on the paging SSD.

Good luck!

[ymc]

CLCBio Genomics workbench (just using it as an example) can do the reference human genome alignments (if that is what you are referring to by "reference assembly") within a few hours on comparable hardware (see benchmarks: http://clcbio.com/index.php?id=1240).

Other free software (on *unix) should be able to do this too.

Thanks for your reply. Looks like 32GB is good enough for reference genome alignments.

[GenoMax]

Are you talking about de novo assembly?

What if I only want to do a human genome reference assembly?

CLCBio Genomics workbench (just using it as an example) can do the reference human genome alignments (if that is what you are referring to by "reference assembly") within a few hours on comparable hardware (see benchmarks: http://clcbio.com/index.php?id=1240).

Other free software (on *unix) should be able to do this too.

[rwenang]

It should do okay. Though, I think what you need to worry about is the RAM if you are going to assemble big genome.

[ymc]

Not for a very big genome.. Smaller genomes and many transcriptome projects should be fine.

Are you talking about de novo assembly?

What if I only want to do a human genome reference assembly?

[peromhc]

But is 32GB enough?

Not for a very big genome.. Smaller genomes and many transcriptome projects should be fine.

[ymc]

Sounds very interesting as I am after a good "home" desktop as emergency back-up. What make/model and where buying please?

My experience with Corei7s is very positive; I have a 940 corei7 desktop with 12Gb of RAM, which is great for mapping and assembling transcriptomes. BLAST, too, is extremely fast.

However, sorry, does not answer your question and I can't as never tried to assemble a genome from scratch on mine (12Gb never going to work anyway). I guess it may be possible for Human or similar size but could take a long time.

Can you get 32Gb of ram for a desktop i7?

yea, the new Sandy Bridge CPU supports quad channel DDR3-1600 RAMs.
A-DATA is selling 2x8GB DDR3-1333 RAM for US$400 now. Buying two packs
of these can make it a 32GB desktop.

[poisson200]

Sounds very interesting as I am after a good "home" desktop as emergency back-up. What make/model and where buying please?

My experience with Corei7s is very positive; I have a 940 corei7 desktop with 12Gb of RAM, which is great for mapping and assembling transcriptomes. BLAST, too, is extremely fast.

However, sorry, does not answer your question and I can't as never tried to assemble a genome from scratch on mine (12Gb never going to work anyway). I guess it may be possible for Human or similar size but could take a long time.

Can you get 32Gb of ram for a desktop i7?