Home Page Login Register

Web Address: www.SETIUSA.Net

Supporting BOINC Distributed Computing Projects

Raw Horsepower

One of the big initiatives over the last few years has been the field of stream computing. This fledgling industry leverages the power of today's GPUs - which are more like CPUs than ever before - to handle tasks like physics, video coding, and game AI. In essence, GPUs are now crunching the numbers that once required a CPU.

Both ATI and NVIDIA have worked to develop the stream computing initiative. Between NVIDIA's CUDA language and ATI's Stream Technology, both major industry players now have a solution to do stream computing. But just like the processor, heavier workloads and higher performance call for faster chips, and the HD 5000 series is no slouch in that department.

According to AMD, the new Radeon HD 5870 offers 544 double-precision GFLOPS of processing power. The FLOP stands for FLoating point Operations Per Second. A floating point operation is a basic calculation used by the CPU to process code, especially scientific ones like computer AI, video encoding and physics. Double-precision FLOPS ensure a high degree of accuracy in these calculations, which translates to more accurate rendering or encoding. The Radeon HD 5870 can perform 544 billion such double-precision calculations every second; in comparison the Intel Core i7 975 XE chip can only perform 70 billion of them.

In terms of raw numbers, the new Radeons are at least 7 times more powerful than the CPU. In other terms, that's twice the capacity offered by the Radeon HD 4870. That is some seriously righteous throughput, and the rise of DirectX 11 will ensure that it will increasingly be put to very good use.

Power Consumption

Power efficiency has been a big trend in the computing industry. Terms like performance per watt, dollars per watt and green computing all point towards the market's desire for more horsepower on less juice. ATI has listened, and the Radeon HD 5800 series is one of the most power-efficient enthusiast GPUs the company has ever produced.

The biggest reduction in power consumption is owed to the transition to a 40nm process node. GPUs built to 40nm quite simply use less power than their 55nm predecessors. As a result, the new Radeons consume just 12% more power (188W peak) than the 4870 (160W peak), but at twice the performance. By any measure, this directly translates to an incredible improvement in power efficiency.

The move to 40nm has reduced idle power consumption as well. While the Radeon 4870 consumed around 45W idle, the 5870 consumes just 27W with no GPU load. ATI has also taken advantage of Windows Vista and Windows 7's Linked Adapter mode to detect underutilized GPUs in a CrossFire environment. These underutilized GPUs can achieve an even lower power state, which takes consumption well below 27W.

The other major reduction in power consumption is owed to a brand new memory controller. The Radeon 5870 packs 1GB of high-frequency GDDR5, and that can gobble up juice in a hurry. To keep VRAM power consumption in check, ATI has implemented a so-called low strobe  mode. This mode forces the GDDR5 to act much like GDDR3, which requires less power. This low power mode is perfect for non-3D instances where the GPU's full 4.8Gbps of memory bandwidth isn't required. In this state, the GPU has just 0.6Gbps of memory bandwidth, but it can be kicked up to full blast in an instant.

Noise and Heat

The switch to 40nm has also had an effect on the GPU's heat output in the form of a big reduction over the previous generation. ATI has followed this improvement with a commensurate improvement in cooling, resulting in lower temperatures and lesser noise.

While the Radeon 4870 could easily reach temperatures north of 70 C under full load, the 5870 should regularly stay within 40-50 C in a well  ventilated case. These frosty GPU temps, says AMD, are owed in part to a larger GPU with an increased heatsink surface area. A bigger heatsink simply dissipates more heat. The larger player in reducing heat, however, is the 40nm chip itself. Smaller chips require less voltage than bigger ASICs, and less voltage means less heat.

Some would think that higher performance with lower temperatures leads to increased fan noise, but that's not so according to AMD. The firm has redesigned the fan to use newer bearings that produce a lower tone. These new bearings are scarcely audible at low speeds, and aren't  quite as piercing at maximum RPMs.

Lastly, the fan's velocity has also received an overhaul. The new cooling system on the 5870 can run as low as 1200 RPM, and that's a lot  less than the 4870 before it. Even at maximum velocity, says AMD, the fan is still slower than the one used on older Radeons.

All of these changes to thermal output, the heatsink and the fan mean the 5870 is a whisper compared to the noisy blast furnace it's  replacing.

Intel now has 3 different high end desk top CPU chips now on the market. These processors are the, I7 9xx series, the i7 8XX series, and the I5 7XX series. Each one of these processors were designed for different purposes, have internal differences, require a different mother board type, memory type, CPU cooler, and use a different CPU socket. It is very confusing - you need a program or crib sheet to keep up with the players. What are these chips, what are the differences between them, and what are the pros and cons of each?

First a short history, Intel developed the Nehalem class of chips released under the names I7 and now the I5. These CPU's are very simply high end quad core processors with advanced instruction sets and an operating system controlled speed boosting technology called Turbo-Boost. The advanced instruction set and other CPU advances give a reported 15% to 30% increase in performance per processor cycle over previous Intel models. Turbo-Boost technology allows one or multiple CPU cores (CPU temperature allowing) to temporally increase speed to run high priority tasks quicker. In the I7, a process called Hyper-Threading is used to simultaneously run 2 tasks on one CPU core at a time. This allows 8 tasks or threads to be run at the same time on a quad core processor.

For more information on computer CPU advances in the Intel Nehalem see this web site.
 

URL: http://www.intel.com/technology/architecture-silicon/next-gen

or URL: http://en.wikipedia.org/wiki/Intel_Nehalem_(microarchitecture)

I7 9XX

Intel first introduced the 9XX series I7 in November of 08. The 9XX series performance and over-clocking potential have made it the must have for all top gamers and power users.

Important specifications for the 9XX I7 are as follows:

Has a built in DDR3 memory controller. This controller allows the use of very high speed (very expensive) DDR3 memory. (DDR3 memory supplies the data stored on it faster allowing the system to operate with less latency.)

- Uses the x58 mother board chip set.

- Excellent Overclocking performance.

- Uses the LGA 1366 CPU socket.

- 24 gig memory capacity.

- CPU upgradeable

- Produces a tremendous amount of heat requiring a CPU cooler. Especially when over-clocked.
 

I7 8xx

The Intel 8xx I7 was introduced in September of 2009. When comparing the new 8XX series to the 9XX series, the I7 8XXs; CPU clock speed is faster, Turbo-Boost is more efficient, the DDR 3 memory controller has been removed, and it has been given a single path CPU chip based PCIE controller. The on board PCIE controller reduces the lag time for the CPU to service the computers PCIE devices. The 8XX series, in most applications, is a faster CPU than the 9XX. The 8XX series lower CPU purchase price and use of lower cost ram/mother boards will surely make extremely popular.

I5 7XX

Intel also introduced 7XX series I5 in September of 09. This CPU is exactly the same as the 8XX I7 except for one thing. It has no Hyper-Threading ability. (Instead of running 8 tasks at one time it only runs 4) The 7XX is less expensive than the 8XX series and about 15% more expensive than the AMD 965 BE 3.2GHz Phenom x4. The performance of the 2.66Ghz I5 750 was recently compared to the AMD 3.2Ghz Phenom 965 x4 in the November 2009 publication of Maximum PC magazine on page 27. In 27 different tests the slower I5 750 outperformed the AMD 965 in 25 of the 27 tests. It performed anywhere from 15 to 50 percent better than the AMD.

Important specifications for the I7 8XX & I5 7XX:

- Uses DDR2 memory. (Less expensive & slower)

- Uses the X55 Mother Board Chipset. (Less expensive Mother Board)

- 16 gig memory capacity

- Uses LGA 1156 CPU Socket

- Is less expensive than the 9xx series.

- Excellent Overclocking performance.
 

- Does not have a built in DDR3 memory controller
 

- Produces slightly less heat than the 9XX series.

The I7 8XX and I5 7XX chips do have three potential limitations:
 

Intel's removal of the CPU DDR3 memory controller greatly lowers the price of the of the 8XX series. ($200 to $300 a CPU chip). Using DDR 2 memory also saves some money, because DDR 3 is very expensive. DDR2 memory is slower that DDR 3, so logically there is more memory lag in an 8XX series and 7XX series systems than in a 9XX system. Tests posted on the internet, show there is a decrease in overall memory performance of up to 40% when comparing DDR2 memory to DDR3 memory. This difference is only noticeable to the user when running very memory intensive applications.

The addition of the in CPU PCIE controller is a great advance in processor architecture. This controller lowers the time required for the CPU to respond to the needs of the PCIE devices installed on the computer. It also lowers the cost of the X55 chip set (Around $100 dollars less) by allowing mother board manufactures to do away with the part of the system board that controls PCIE access now controlled by the CPU.

There is one potential problem with this set up. The current 8XX & 7XX PCIE controller only has a single PCIE controller lane. This controller lane is capable of a maximum of 16x communication. This means that in a system that has multiple PCIE slots the available bandwidth to each slot is divided by the number total. (For example: 1 PCIE x16 slot bandwidth of 16x, 2 slots a bandwidth of 8x, 3 slots 5.3x etc. This bandwidth is also divided over all PCI slots x16, x4 and x1.)

This is not a limitation for users running 1 or 2 GPU's. 8x speeds are adequate for all current GPUs. (Yes even the ATI 5870 is ok, but it could be a problem with the 5970). Some X55 mother board manufactures have eliminated the problem by adding a PCIE caching chip. This chip manages the PCIE bus allowing all card slots to communicate using their full bandwidth. The chip (Nvidia 200) does add some cost back into the cost of the mother board, (25 to 35 dollars) but is probably worth the money if you are planning on running multiple video cards.

Upgradeability is also a problem with the 8XX and 7XX CPU's. Within the next 60 days, Intel has announced the release of a new 6 core processor. This processor will be able to run 12 threads instead of the I7's 8. The new CPU uses the 1366 CPU socket and with motherboard BIOS upgrades should be compatible with existing 9XX series systems. This will allow I7 9XX series systems to be upgraded to uses the new processor. The name for the new Intel processor has not been released yet and is currently know by the code name GulfTown.

 

Graphics & Video Cards 101: The PC with a View
Written By Ryan Carter - Submitted By Top Viking - March Edition

Obsessable covers the latest in the world of technology, including cell phones, digital cameras, and HDTVs, obsessively, of course.

Article URL: http://www.obsessable.com/contributor/ryan-carter/