The holy grail of linux users has arrived: Core A5 and the return of BigAdv

By , the in Folding@Home Project - 1 Comment
The holy grail of linux users has arrived: Core A5 and the return of BigAdv
It's official now, Linux is back in favour for BigAdv after a long absence.

For many months, folding BigAdv units on Linux was not possible due to a curious bug that caused the results to be incorrect.

The new A5 core appears to have fixed the gremlin allowing Linux BigAdv once again, and indeed significant performance gains.

To use the new core:
It is necessary to have a new version of the client (6.34) plus all the usual pre-requisites for BigAdv as before.
All BigAdv units will migrate to A5 in the coming weeks, so it is necessary to migrate to the 6.34 client on all platforms or you will only receive standard SMP units using the a3 core.

To compare your performances with the new and old cores the forum has set up an official comparison thread: here.

nVidia CUDA 4.0 arrives, optimised for MultiGPU systems

By , the in Drivers - No comment
nVidia CUDA 4.0 arrives, optimised for MultiGPU systems
Up to now, it was possible to perform calculations using multiple GPUs, but due to the way CUDA is designed transfers of data between the cards would have to go via system memory, incurring a significant performance penalty that all-but negated the performance bonus from using multiple cards.
The latest version of CUDA changes all that, allowing GPUs to access each other's memory directly. This simplification could possibly allow Folding@home to be redesigned to fold one WU using multiple cards, in the same way as the SMP client does with CPU cores. As usual, speed is everything (so long as the results are accurate ;) )

It's not a revolution, but its a new feature which could be very useful!

Source : PC INpact

Publication #78: the role of water in membrane fusion

By , the in Folding@Home Project - No comment
The research team at Folding@home recently published the 78th peer reviewed paper to be produced using the project's work, in the Journal of the American Chemical Society (JACS). The paper, entitled "Water Ordering at Membrane Interfaces Controls Fusion Dynamics" is the result of a collaboration between Peter Kasson (University of Virginia), Erik Lindahl (University of Stockholm and the Royal Institute of Technology), and Vijay Pande, of Stanford University.

The ordering of molecules at the interface controls the dynamics of membrane fusion.

Based on previous simulations of membrane fusion, the Folding@home team used the network to systematically analyze the fusion reaction between two cells, and specifically the role of water in this reaction. To achieve a reasonable simulation time, many simulations use an implied model of water to save the processing time that would be required if the water molecules were explicitly modeled. In this simulation, the researchers have shown that water does in fact have a direct effect on both the thermodynamics and dynamics of the reaction. This has important consequences for future simulations of membrane fusion, and the researchers provided a further example of two flexible hydrophilic surfaces. This structure of water could be important in many cellular interactions.

For the record, the units that contributed to this paper were the BigAdv p268x series.

Source: JACS *

*Access to the paper is paid for the first year under the publishing agreement with JACS unfortunately. For those at academic institutions it may be worth checking if your university/college has paid for access to the site.

The Core i7 990X arrives quietly... but shakes up prices

By , the in Hardware - No comment
The Core i7 990X arrives quietly... but shakes up prices
Several weeks ago, Intel released a new technological showcase to top its range of consumer processors: the Core i7 990X.

Much like its predecessors it is made on a 32nm process, but this time features 6 cores at 3.46GHz (compared to 3.33GHz for the 980X). This processor is designed for the LGA1366 platform on motherboards equipped with the X58 chipset.

As usual for Extreme Edition models, the price is set at $999 and the multiplier is unlocked.

The good news for most of us does not concern this chip however, but its effect on chips further down the range. Specifically, the Core i7 960 and Core i7 970 have had their prices (per batch of 1000) reduced substantially.

The result is that the Core i7 960, a hyperthreaded quad core running at 3.2GHz with 8MB of onboard cache can now be had for $294, rather than $562. The hexcore 970 also runs at 3.2GHz but with 12MB cache onboard, and now costs $583 rather than $885.

Source : TechConnect

Sandy Bridge is released ... but the 1st generation Core i7s strike back!

By , the in Hardware - No comment
Sandy Bridge is released ... but the 1st generation Core i7s strike back!
Sandy Bridge has a problem: admittedly a minor one, with the chipset, but it means that for now the 1st generation i7s are not dead!

Firstly 11 Xeons come to reinforce Intel's server list (see board) :

The big news is that Core i7 990X, which slots straight in at the top of the range to lead the hexcores. As a result, other top-end hardware has a price drop to compensate:
The i7 970 and 960 decrease from $583 to $294, and the i7 950 is to disappear from the face of the earth soon ...
More simply, these are the first hexcore processors (from Intel) for less than $600/€600.

Great news for the power folders :P

Source : PCINpact (French)

SMP v6.34 client available for download

By , the in Folding@Home Project - No comment
Version 6.34 of the SMP client is now available for download on FAH's high-performance clients page, for both Windows and Linux.

It is provided as an executable file, which should be placed in the main directory of an already installed client. For safety and data stability reasons, it is recommended that this update be applied when no partially-completed units are present.

The main new feature in this version is support for the new A5 core, which should be making its public appearance in the near future.

This new core will be used exclusively by BigAdv projects. It's identical to A3 core, but it is built from a different code branch which is specific to BigAdv projects to improve compatibility and stability.

Source: Vijay Pande's Blog

Updated on 02/25 at 11:30 :
Project 6901 has just started being assigned. It is identical to project 6900, the only difference being the use of the new A5 core.

The Gromacs 4 core reaches version 2.27

By , the in Folding@Home Project - No comment
Kyle Beauchamps has released a new version of the Gromacs 4 core, more commonly known as Core a4. This version upgrades the Gromacs software to version 4.5.3, giving the researchers more analysis methods.

This core is available for Windows (32-bit) and Linux (32-bit and 64-bit). For the avoidance of doubt the Windows core runs fine on 64-bit Windows ;)

No projects currently require the new revision, however if you wish to force the update then deleting your copy of the a4 core will cause the new version to be downloaded for the next unit that requires core a4.

If you choose not to upgrade now, it will be performed automatically the first time you receive a project needing the newer core.

Source: Official Forum

Xeon E7 (aka Westmere-EX) specifications revealed

By , the in Hardware - No comment
There have always been processors that make you dream. These processors, destined for the professional/server markets, use the latest technologies but are prohibitively expensive for most people on the street.

The specifications of the Xeon E7 revealed today most certainly put this series of processors in the "dream" class. These chips are intended to replace the Xeon 7000 series and should become available in the second quarter of 2011.

These processors are manufactured at 32nm and feature between 6 and 10 physical cores depending on the model, most augmented with HyperThreading. Their stock clock speeds range between 1.73 and 2.67GHz, with nearly all models able to turbo boost to higher speeds if the chip is not fully loaded. There will be between 18 and 30MB of cache onboard and multiple QPI links to communicate with RAM and the other processors in the system.

The names are fairly simple: E7-2000 series is a dual-socket processor, E7-4000 for quad-socket systems and E7-8000 for eight-socket systems (potentially 160 threads in one machine).

TDP for each processor ranges between 95W and 130W.

Prices are not yet known, however it is reasonable to assume that with this level of power available they will be mildly ridiculous.

If you wish to run Folding@home on such machines, you may get a nasty surprise: for a given size in simulation there comes a point where adding more threads can in fact slow down progress rather than increase it (due to the increased time spent exchanging data between the different threads). Currently the recommended maximum is 24 threads for regular SMP and 64 threads for bigadv, for maximum performance.

Source: CPU-World

AMD's Radeon HD6990 and nVidia's GeForce GTX590: Welcome to the world of excess

By , the in Hardware - 4 Comments
In the world of graphics cards, there are broadly two categories. Cards for everyone, the normal person on the street, and technical showcases which are bought only in limited numbers but attempt to assert one manufacturer's superiority over the others. These two cards definitely fall in the latter camp.

The red side of the graphics card world, AMD, has come up with the Radeon HD6990. This card is codenamed Antilles and combines two of the Cayman chips previously seen in the HD6950 and HD6970 on one PCB. The card's ability to deal with the heat produced by the 6 billion transistors on the PCB will probably be the deciding factor on whether it is seen as a success or a flop.

AMD Radeon HD6990

This card is physically a monster, to match the monstrous power hidden underneath that black shroud. At 30cm long, it will only fit in the biggest cases, and is a full inch longer than the other cards in its series (6950 and 6970). It requires both an 8-pin and a 6-pin PCIe power connector, for a total potential power supply of 300W (150W from the 8-pin connector, 75W from the 6-pin connector and 75W provided by the PCIe slot). As you can see in the picture, cooling consists of one centrally mounted fan blowing air in both directions. Regrettably this means that the heat output of one entire Cayman chip will end up being dumped inside the case, potentially causing trouble.

For connecting to displays, the card is equipped with one DVI port and four mini-DisplayPort connectors. Prices have yet to be announced, but can be presumed to be astronomical. Availability is given as the first quarter of 2011.

Source: Clubic (in French)

For those of you that prefer the green side of the world, nVidia has come up with the GeForce GTX590. This card contains two of the GF110 chips used in the GTX580, and as they remain unmodified this also results in a complement of around 6 billion transistors. The card will therefore have a total of 1024 shaders and 3GB of onboard vRAM.

The operating frequencies are unknown and nVidia has apparently decided to announce them at the last minute, probably to allow for increases if these are necessary to beat the HD6990. What we do know is that they will be considerably lower than the single-chip cards, as one GTX580 alone consumes 244W. It may be a good time to buy shares in your energy supplier if you wish to purchase this card!

Air cooling will be provided by a centrally mounted 90mm fan, in a design similar to the single-PCB variant of the GTX295. The GTX590 will be even longer though, potentially longer than the HD6990. As with the HD6990, the fan's placement means that a large volume of hot air will be dumped into the case so cooling could be an issue.

We do not have any images of the final card, but apparently it is scheduled to be available in February in an attempt to repeat the coup of the GTX580 and beat AMD to market.

Source: Nordic Hardware

One thing is for certain with these cards: the PPD figures they produce will be stellar, but the PPD/W figure will be questionable, with lower-end cards likely to do better on that front.

GPU3 client reaches version 6.40r1; featuring Compute Capability support

By , the in Folding@Home Project - 3 Comments
Original news on 10/31/10:
Mark Friedrichs has just put version 6.40r1 of the GPU3 client online.

The main change in this version is the replacement of the previous method checking for CUDA support, namely by checking the version of CUDA installed on the computer, if it was at all. The client now achieves this end by directly checking the Compute Capability value on the GPU chip itself.

Possible values for detected versions are:

v1.0 -> 10
v1.1 -> 11
v1.2 -> 12
v1.3 -> 13
v2.0 -> 20
v2.1 -> 21
Unknown -> 0

The main reason for this change is due to an incompatibility between Core 15 (OpenMM) and cards that only support Compute Capability 1.0 (which means a G8x chip). The change should now prevent unpleasant crashes for owners of these cards, and pave the way towards better hardware detection, perhaps with some specialised optimizations based on the chip capabilities.

This new version comes in the form of an executable file replacement. Quit out of your client and replace its executable with that of the new vesion.

Console version:
Systray version:

Remember to remove the trailing _console or _systray from the file name!

Updated on 01/27/11:
The Systray version of the 6.40r1 (also known as 6.41) client is now available on the High Performance clients download page.

Change log:
  • A new value for -forcegpu flag has been added : nvidia_g80_1.0 will force Compute Capabilities 1.0 (species=10)
  • The old -forcegpu nvidia_g80 value will force Compute Capabilities 1.1 (species=11)
  • If the -forcegpu value is not recognized, the client will give a pop-up warning then exit.

Source: Vijay's blog

Updated on 02/15/11:
The v6.41 console client is now available on the High Peformance client download page.