|
|
本帖最后由 金鹏 于 2010-4-26 12:21 编辑
OCTOBER 22, 2009
Updates on new cores: Protomol and GPU3 core
There has been a lot of work on updating the Protomol core to bring it in line with the other cores in Folding@home. I'm happy to say that a lot of progress has been made and it's looking much better. Joe will continue to test it, but it looks close to be moving to the next levels of QA.
The GPU3 core is also moving along. It will be called core_15 (the natural next number in the GPU series). The main changes there have been to incorporate the updated GPU code from OpenMM. OpenMM was based on our FAH GPU MD code to start, but has had several enhancements and additions. In particular, it should be much more stable than the previous FAH GPU MD cores. However, this stability does come at a mild cost in performance. We will address this at the benchmarking stage, since all core 15 (GPU3) projects will start fresh, not continuing existing projects.
April 24, 2010
Prepping for the GPU3 rolling: new client and NVIDIA FAH GPU clients will (in the future) need CUDA 2.2 or later
As we've discussed in previous posts, due to its great computational abilities, our GPU client has had a great scientific impact so far. In our most recent FAH paper (also see the movie), the GPU clients play a star role in allowing Folding@home to push to unprecedented levels, simulating protein folding on the millisecond timescale in an atomistic model.
We are prepping for the rollout of the next generation GPU client (GPU3). As mentioned in previous posts, GPU3 will allow for greatly enhanced science (including more accurate models, new science can be done, 2x faster execution of the science, more stable simulations, OpenCL support for run time science optimizations, and greater flexibility for adding new scientific capability). This is accomplished through the use of the OpenMM GPU library (which originally came from FAH GPU code, but has been significantly enhanced by Simbios staff).
We would like to give donors a heads up of what's coming. We are doing internal testing now and will do closed beta testing hopefully soon. With the rollout of the new GPU3/OpenMM-based core (core15) for NVIDIA GPU clients, we will need donors to do two software installs (please note that this is not required to be done immediately, since the new client is not openly available):
1) In order to get WUs using this new core, donors will need to make sure their CUDA level is least CUDA 2.2, but ideally 2.3 or the most recent. To know which version of CUDA you have, you can find out based on your driver version:
CUDA 2.0: 177.35+
CUDA 2.1: 180.60+
CUDA 2.2: 185.85+
CUDA 2.3: 190.38+
CUDA 3.0/OpenCL: 195.36+
2) A new client will be needed to access GPU3 WUs. This new client will report the CUDA level to the assignment server, so it can assign around machines with less capable CUDA levels. Note that "assigning around" the issue means that if your client can't do the work available, it won't be assigned a WU, so it's best to make sure your CUDA drivers are reasonably updated. We feel this is better than giving a WU which will crash the core, etc.
While the new client has not been openly released yet, we wanted to give this heads up to donors so they have time to upgrade their drivers.
Thanks to all of the GPU folders. We have done some great work so far and the best results are yet to come!
大意：
GPU3准备开始公测
GPU客户端为FAH做出了巨大的贡献。具体可参考论文和影片。
GPU3可以提高计算精度，增强稳定性，同时也更方便代码优化，可以使研究速度提速2倍。主要是因为GPU3使用了OpenMM（由FAH GPU衍生出来的）。
GPU3当前还在内测中，估计不久就会开始公测。在这里我们先‘剧透’一些信息。GPU3/NVIDIA客户端将使用core15计算内核，为此需要大家做好2个准备工作。
1、新内核对CUDA的最小需求是2.2版，不过版本越高越好，所以您最好安装最新版的驱动。下面是版本信息：
CUDA 2.0: 177.35+
CUDA 2.1: 180.60+
CUDA 2.2: 185.85+
CUDA 2.3: 190.38+
CUDA 3.0/OpenCL: 195.36+
2、安装最新版客户端，客户端会向AS（任务分配）服务器报告您当前的CUDA版本，并取得最合适的任务。所以当您的驱动版本较低时，很有可能收不到任务。
因为GPU3还没开始公测，所以您有足够的时间，更新驱动，不用急。
NVIDIA 驱动程序下载 |
|
发表于 2010-4-26 08:28:30
