The GTX560 Ti arrives priced at €229, let the price war begin!

By , the in Hardware - No comment
The GTX560 Ti arrives priced at €229, let the price war begin!
The GTX560 Ti is the 5xx series evolution of the GTX460. While similar to the older card, all the locks and disabled components present in the GTX460 have been unlocked or enabled for the GTX560 in the pursuit of ever more power. As a result, the full complement of 384 shaders is active at a clock speed of 822MHz (versus 336 and 675MHz for the older card). The new revision of the chip is called GF114, and is still manufactured on a 40nm process with nearly 2 billion transistors.

Physically the card looks different to its predecessor, as it is considerably longer (3.5cm) than the GTX460. However this still does not make it an extremely long card as the GTX460 was a short card to begin with (21cm). Power consumption increases slightly from 160W to 170W, however when viewed in light of the potential performance gains available this would seem a reasonable increase. The power circuitry has been beefed up considerably, which should prove a boon for overclockers wishing to extract the maximum possible performance from the card.

In short, if you loved the GTX460, you'll love the GTX560 Ti.

The more observant among may have made the connection with the card's suffix: it was last attached to the GeForce 4 Ti, one of nVidia's best-selling cards in the past.

Source: PCINpact (in French)

New projects roundup : p6801, 6805, 6883 and 6887

By , the in Folding@Home Project - No comment
During the second half of December, not many new projects are typically added for processing due to Stanford's winter break, though the first half of January usually helps to redress the balance somewhat. In the last month, only Yu-Shan Lin has added new projects, as follows:

p6801: OpenMM for Fermi
This project simulates the amyloid beta in a mutant configuration, as with other p6800 series projects. The units are worth 1348 points and simulate a system of 634 atoms. The preferred deadline is 5 days, with the final deadline 5 days later. These projects are served from

p6805: OpenMM for Fermi
This project continues the p6800 series of amyloid beta studies, but this time not in a mutant configuration. The units are worth 1280 points and comprise 598 atoms. Deadlines and the server are identical to p6801.

p6883: Uniprocessor Gromacs 3 (Core 78)
This project simulates the abeta peptide in explicit water. The unit comprises 12,226 atoms, is worth 69 points and has a preferred deadline of 14 days, with the final deadline a week later. These units are served from

p6887: Uniprocessor Gromacs 3 (Core 78)
This project also simulates the abeta peptide in explicit water. The unit comprises 12,219 atoms and is worth 69 points. Deadlines and the server are the same as p6883.

Happy Folding :)

GTX560 unveiled in the latest drivers! The Fermi range update continues

By , the in Hardware - No comment
GTX560 unveiled in the latest drivers! The Fermi range update continues
nVidia's latest drivers (266.44) have revealed the existence of a card called the GeForce GTX560 Ti. This is the logical continuation of card updates in the Fermi range, and as the name suggests, would be intended to replace the GTX460. It seems that this card is very similar to the GTX460, but with all the available SPs enabled (384 rather than 336), and an increased core clock frequency of 825MHz, compared to 675MHz for the older card.

Knowing the performance of the GTX460, this card looks very promising for Folding@home!

Source: TH France (in French)

Rumor: 8-core Bulldozers 50% faster than i7

By , the in Hardware - No comment
Rumor: 8-core Bulldozers 50% faster than i7
Intel has been flexing its muscle in the past few weeks with the arrival of Sandy Bridge, but AMD are so confident about the competitiveness of their upcoming CPUs that they have leaked some very intriguing information.

8-core bulldozer should be more than 50% faster than a 4-core i7 (with HT enabled)

Under what conditions these figures were produced, we do not know. However it would mean that 8-core Bulldozer would be aimed as a competitor to 6-core products from Intel, quite possibly at a lower price point.

We await benchmarks with impatience!

Source: TH France (in French)

Have machines based on EVGA's SR-2 become dinosaurs?

By , the in Hardware - 1 Comment
Have machines based on EVGA's SR-2 become dinosaurs?
Until recently, the huge (war?) machines based on the EVGA SR-2 ruled the roost for energy efficient folding, that is high PPD/W.

Previously folders would find that multiple single-socket machines would be the most effective way of gaining points. This technique was rendered useless by the introduction of the bonus system, which made it more economical from a PPD/$ perspective to just buy one system with a view to turning in units as quickly as possible.

The SR-2 meant an easier way to overclock very powerful dual-socket machines in order to extract the most points, and quickly became the most efficient method of earning points using any client. The downsides are clear, mainly upfront costs (the motherboard, processors and RAM alone could easily cost $2000, pricing most people out of buying such a system.

However now there is a new kid on the block ready to challenge the SR-2's dominance. Sandy Bridge has proved to be a very able overclocker, and the i7 2600K can be overclocked to 4.5GHz on air (though not using the standard cooler, obviously). A system based on this processor and the P67 chipset can easily make 35,000-40,000PPD. Suddenly the scenario of using multiple cheap systems would appear to once again be the most cost-effective way of producing points. It should be very interesting to see how the second wave of Sandy Bridge processors performs, which are apparently due to arrive in 3-4 months and will be the more high-end processors. This should see much more advanced 6-core processors (and a price tag of less than $800 for such chips at last).

Intel Socket 1366, 1156, and 1155 details

By , the in Hardware - No comment
With the advent of Sandy Bridge, the technological situation at Intel is becoming more and more complex. To help us see this more clearly, our friends at PCINpact have compiled the technical specs in a handy and very thorough summary.

You'll find all the details on Socket 1366, 1156, and 1155, which comprise the whole Core iX range of processors. Of particular interest, the multipliers and clock frequencies for Turbo mode have also been included.

Click the image to make it readable.

Source : PCINpact (in French)

Sandy Bridge has arrived!

By , the in Hardware - No comment
Sandy Bridge has arrived!
Sandy Bridge is here!

A read through PCINpact's Sandy Bridge review will give you a good look, but for the impatient, here is a brief overview:

  • 1 billion transistors: The complexity is increasing, and poor control of the manufacturing process would have transformed this CPU into a glorified flamethrower, but Intel has taken advantage of the recent successful 32nm manufacturing process to produce something a little less volatile.
  • Another new socket type; Nehalem's has already come and gone, leaving us mourning the days of Socket 775.
  • A sharp drop in TDP, and by extension a lower power consumption.
  • AVX (256-bit floating point) technology, which could prove to yield great bonuses for folders if the FAH cores are re-compiled to exploit it.
  • The clock frequencies are increased again compared to the previous generation.
  • The K versions have unlocked multipliers, allowing for easy overclocking without stretching the mainboard too much. Consequently, only the quality of the hardware/power supply components will remain a deciding factor.

And finally, we come to the PPD, because that's what everyone's after! A grand total of 32,000 PPD was logged on a stock 3.5GHz i7 2600K by our colleagues at PCINpact. It's a sharp increase compared to the 20,000 PPD gained on an i7,870, and this will obviously only rise still further with a little overclocking.

We're eager to try one, and wishfully remind Intel that while Christmas may be over, the giving spirit is year-round.

Happy New Year! Time to take stock... and look forward to 2011!

By , the in Folding@Home Project - No comment
OK so we were a bit late translating this news and its already the 4th of January. However, we wish you all a belated happy new year and hope you haven't failed your resolutions yet ;)

A new year means it is time to look back at what the project has achieved over the previous year, and whether we predicted the year correctly.

How did we do with our predictions for 2010?

2010: Year of the v5 server code: Correct: (OK this wasn't exactly the hardest prediction to make we admit)
The vast majority of the project's servers have now been switched to the much more reliable v5 server code, meaning fewer dropouts (once some teething problems were ironed out). Some servers even began migrating to the further improved v6 server code in the later stages of the year.

A new, more reliable statistics system: Correct
With both new hardware and new software, the statistics system underwent a major revamp in 2010, and most statistics features that had been deactivated are now enabled once again. Combined with some rather more aggressive IP banning by Stanford University, this has resulted in a major performance boost.

The launch of the long-awaited brand new client: Wrong
The v7 client did not make an appearance in 2010, so unfortunately our prediction that it would was proved wrong. However, the v6 client made some important leaps, particularly the abandoning of MPI for the SMP client, helping to make the project easier to contribute to.

A new version of the Assignment Server (AS) code: Correct
While not visible to donors, the code that the ASes run has been updated. As it is impossible to be assigned work without first being referred to a work server by the AS, the improved reliability of the ASes has helped to reduce headaches for donors.

The GPU3 client to reach maturity: Kind of...
In terms of the algorithms needed to make this happen, OpenMM on CUDA is ready and is being used as a kind of GPU2.5 client for nVidia GPUs (as it stands it is the only way to fold using Fermi cards). Unfortunately OpenMM on OpenCL is not quite ready, despite the fact that it is desperately needed by ATI users to replace the current ATI core.

SMP2 making its appearance: Correct
SMP2 has rapidly replaced SMP and added much to the project: more reliability, more speed, Windows compatibility, simpler operation, the latest Gromacs developments... but we have lost Linux BigAdv!

SMP2 to gain more cores: Wrong
Desmond has not yet appeared. 2011?

So 2010 has been a big year for development behind the scenes at folding@home, but SMP2 excepted, not much of a revolution for users.

Consulting FAH-Addict's crystal ball...

2011 year of v7: this time for sure! The key word: usability.
2011 year of GPU3 on ATI: OpenCL should finally bring up the goods and allow GPU3 on ATI.
2011 year of BigAdv Linux: we predict that the bug that is currently preventing BigAdv on Linux will be fixed this year, allowing it to return. We certainly hope so :P

Folding@home's v7 client expected for Q1 2011

By , the in Folding@Home Project - No comment
It is now official, Vijay Pande has announced that the v7 client has a target release date, namely the first quarter of 2011.

Of course, this gives a wide range for the release, and last minute bugs may well appear that may delay the release beyond the target date. However the fact that a date has been announced in advance at all is a good sign that no such issues are expected.

To recap, the new client has been designed with usability in mind. The current clients might best be described as functional, and the new client is intended to address the usability deficiencies and make contributing much easier. For the developers, the new code is intended to be much easier to maintain and more logical to add to, allowing faster development of further improvements.

As a result it should be easier for the client to be modified to support new client types or methods than it has been with the current clients.

We will keep you updated of any news on this release.

Source: Official Forums