GeForce GTX Titan – The most powerful single GPU card

By , the in Hardware - No comment
GeForce GTX Titan – The most powerful single GPU card
If you’re not shocked by spending a month of minimum wages on a computer component, you should know that nVidia is preparing to release a mythological monster !
The GeForce GTX Titan will be based on a GK110 chip, a development of the Kepler family, which is the first card on sale to use this component. The chip is supposed to be as powerful as 85% of a GTX 690 which is equipped with two high end Kepler chips !

The GK 110 chip is identical to the one embeded on the Tesla K20X GPGPU card. As a reminder, it is fitted with 15 SMX (2880 SP) with 14 activated (for 2688 SP). This chip is capable of delivering 3.95 GFLOPs on single precision calculations and 1.31 GFLOPs on double precision ones.

Here are some numbers, mostly as rumors, for the GeForce Titan GK100:

  • Released on February 18th
  • 2880 SP (15 SMX) with 2688 activated (14 SMX)
  • 7.1 billion of transistors
  • 502 mm² die
  • GPU frequency : 732 MHz
  • 6 GB GDDR5 on a 384 bits bus clocked at 1300 MHz
  • 235W TDP
  • Priced between $850 and $1000

In other words, a WU devourer ! Even if this kind of high end chip might be subject to early defects. nVidia is going to enforce strict design rules for this card on its partners, so no custom coolers!

The rumors about the availability date vary. The vast majority of the sources tends to point toward February 18th, though some of them mention the 25th …

Source : PC World (in French)

Computation errors and GPU temperature : some numbers

By , the in Hardware - No comment
This is something we knew, but without being able to define the real impact. Our GPUs are more likely to produce computation errors than our CPUs. However, temperature seems to be a decisive variable in the growth of error probability. This news (in French) shows a small panel of nVidia cards from various architecture and the testing conditions.

Of course, the cards in desktop cases are the main focus, but the laptop or high performance computing rack cards are more likely to be subject to these issues because they run in a much more confined environment. The good news is that most modern GPUs (Fermi or Kepler) are kept away from erroring by efficient thermal protections.

This experiment consolidates our usual advice : avoid folding on laptop GPUs, and make sure that your sensitive components are correctly cooled.

Source : Tom's Hardware France

Xeon Phi: an x86 coprocessor, but what about OpenCL?

By , the in Hardware - No comment
A few weeks ago, Intel restarted the hype about its abortive "Larrabee" architecture. This time the project is proceeding to the market, as proved by Intel assigning it a trade name: "Xeon Phi".

If you followed the announcement you will be aware that the new chip contains "more than 50 cores" capable of running x86 code, and up to 8GB of memory, all mounted on a PCI-express card rather than the more traditional CPU socket, to negate the need for a new design of motherboard.

Clearly this has changed the aim of the project from a GPU that is capable of GPGPU tasks, to that of a coprocessor, similar to the x87 chips of old.

The card is designed to work alone, or in a cluster as part of a wider supercomputer arrangement.

In terms of software support, details are currently unclear. The card will run a Linux OS, isolating itself completely from the rest of the system (one could ask how it functions as a coprocessor given this restriction...) and Intel's own marketing documents do not mention OpenCL. Is Intel ignoring this standard-in-the-making?

A few days ago, the Parallelis website scrutinized the documentation for the card and raised some questions about its effectiveness, particularly when running x86 code, and the limitations imposed by using PCI-express (which is much slower than interconnects between CPUs on the same chip, or CPUs and RAM, for example).

In short: this is a potentially fascinating development, but there are unanswered questions. Intel, the ball's in your court...

Source: PCInpact (in French)

Dust, our old enemy finally defeated by MSI?

By , the in Hardware - No comment
Dust, our beloved machines tend to accumulate it forever … at a variable pace depending on their situation in the house and the time they spent powered on …
Faced with this plague, we all fight it with our own proven weapons: can of compressed air, tooth brush, paintbrush, everyone has his own style.

Yet MSI seems to have discovered an innovation that could revolutionize our struggle, this innovation is ...

... the fan that removes dust from itself!

The idea is simple: during the first 30s of boot, your graphic card is cold and we are sure that it won't perform any intensive computation (you're still booting your OS).
The spin of the fan is then reversed compared to nominal functioning and the speed it runs is pushed at its maximum value. The intended effect is to get out some of the dust that began to take up residence on the blades and the bearings of the fan.After 30s of reverse spin, the fan spin is reversed again to enter its nominal functioning behavior.

This will never replace a good clean up, but that’s probably something that will help the graphic cards to handle the folding torture longer.

The first MSI card to use this new feature is based on a GTX 580, the GeForce N580GTX Lightning Xtreme Edition.

Source : Tom's Hardware France

It's time to stock up on RAM!

By , the in Hardware - No comment
SMP2 has greatly improved the efficiency of Folding@home in regard to memory usage, reducing its usage from gigabytes down to a few hundred megabytes of RAM. However if your machine has a limited amount of memory, what follows may please you!

Despite a major downward trend, DRAM manufacturers have massively overestimated the market for their products, resulting in an increase in inventories around the world. The memory type most affected is DDR3 (to be expected, as it is the current technology), it is trading at US$1.60 a unit and could fall as low as US$1.25 a unit (a 2Gb so 250MB chip), a decrease of 22%. The main cause of this decline in demand for RAM is considered to be increased demand for other forms of computer such as tablet computers replacing demand for PCs, which use different designs. This fall in prices cannot continue forever though, as there is already evidence that some chipmakers are moving their production capacity over to non-volatile flash memory for use in cell phones, tablet computers, memory cards etc.

Source : ZDNet (in French)

R.I.P. Socket 775

By , the in Hardware - No comment
R.I.P. Socket 775
If, like us, you were a fan of Intel's Core 2 Quad Q6600, you'll probably shed a tear or two following the announcement of Socket 775's passing. The platform appeared in time with the Pentium 4/D and had its heyday with the advent of the Core 2 Duo and Core 2 Quad. Stocks still exist, but if you want to grab one for sentimental value, you'd better act fast.

The famous socket 775

Source: Tom's Hardware France

Single-core nostalgia?

By , the in Hardware - No comment
Single-core nostalgia?
Too many cores? Too many threads? Do you want to return to the simplicity of the past?

Intel has announced the Celeron G440, a Sandy Bridge generation CPU lacking any special features: no turbo mode; no hyperthreading. This CPU offers up 1.6Ghz clock speed, and 1MB of L3 cache at 1GHz.

In short, it's a slimline CPU for the office environment. The unusually high TDP of 35W classes it above most of the mobile device range, making the CPU more or less destined for the desktop PC market. If you happen to pick one up and are bored, we'd be interested in the PPD yields for this processor on the F@H uniprocessor client. :hehe

Source: Tom’s Hardware France

GeForce GTX590: Warnings, and first Folding@home tests

By , the in Hardware - No comment
GeForce GTX590: Warnings, and first Folding@home tests
If you are the proud owner of a GTX590, it would be wise to pay close attention to the drivers installed on your machine.

It is essential to make sure you have driver version 267.85 (the only version which officially supports the GTX590), as this is the only driver which fully supports the card's OCP (over-current protection). Sadly nVidia have chosen to manage this through the drivers, as opposed to AMD who build a controller into each card to manage it in hardware.

The cards are currently shipped with a disk containing 267.52, which does not fully support OCP. This has led to the possibility of damage to your card, as demonstrated by this video:

On a rather more positive note, PCINpact has tested the GTX590 on Folding@home. The result was 13,000PPD per GPU on project 6801, for a total output of 26,000PPD per card. Power consumption for the system was 360W, giving 72PPD/W. It is still a long way from the efficiency of -bigadv...

Source: PCINpact (in French)

A little more information about AMD Bulldozer is revealed...

By , the in Hardware - 1 Comment
A little more information about AMD Bulldozer is revealed...
AMD plans to launch its first range of processors based on Bulldozer in June this year. The first batch of processors are codenamed Zambezi.

These processors will be manufactured on a 32nm process, using Socket AM3+. The naming of the range is fairly obvious, with two octocore processors named FX-8000, two hexcore processors named FX-6000 and two quad-core processors named FX-4000. For each of these pairs one will be a Black Edition with an unlocked multiplier. TDP is 95W, except for the top FX-8130P which will have a TDP of 125W. All processors have an integrated DDR3 controller, but none will have integrated graphics. Black Edition models will feature a Turbo mode, similar to those found on Intel's processors.

The processors have 1MB of L2 cache per core, so 8MB for octocore, 6MB for hexcore and 4MB for quadcores, which is joined by an 8MB L3 cache on all models. Operating frequencies remain unknown.

The Zambezi platform is intended to compete at the higher end of the market and is in competition with Intel's i7-2600 series. Prices remain unknown.

We hope that AMD will once again be competitive at Folding@home... and note that the FX-8000 series is AMD's first range of consumer processors which meets the 8-core requirement for -bigadv.

Source: Tom's Hardware (in French)

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