Q & A : Benchmarks : CPU vs. GPGPU: Memory Bandwidth

BenchGpuMem Q & A : Benchmarks : CPU vs. GPGPU: Memory Bandwidth

 

What are they?

A set of benchmarks designed to measure memory bandwidth, both internal and to/from the host system. These are native ports of the traditional memory benchmarks that have been available in Sandra since 1998. They allow us to directly compare CPU with GPGPU performance by using the same algorithms and the same data.

Why do we measure it?

We believe the industry is seeing a shift from a model where the vast majority of workload is processed on the traditional CPU: in a wide range of applications developers are using the power of GPGPU to aid business analysis, games, graphics, and scientific applications.

Coupled with the charts added to the latest version of the software, we can work out whether a CPU or GPU would be faster, more power efficient or cost efficient.

What do the results mean?

  1. The memory results are in MB/s, i.e. how many MB can be transferred in 1 second.
  2. In all cases, as higher indexes mean better performance (MB/s) the higher the result the better the performance.

Typical Memory Bandwidth Results

Testing the bandwidth performance of various current desktop processors and GPGPU-capable video adapters reveals quite interesting results.

Note: Prices fluctuate all the time; the below table was correct as of December 2010, for US market, in USD, via JustRelevant and is provided as an example only. Please check prices in your own region.

Rank Chipset Name Cores / Speed / Power (TDP) / Price Performance Power Efficiency Cost Efficiency Comments
#5 AMD AMD 890FX /
2x 2GB DDR3 PC3-8500 /
AMD Phenom X4 970
4 / 1.06GHz / 20W+2W+125W / 12.8 GB/s 87 MB/s/W 33.2 MB/s/$ Even it’s the last on performance and energy efficiency, nevertheless, it has a decent price.
#1 AMD AMD 890FX /
2x 2GB DDR3 PC3-10600 /
AMD Phenom X6 1055
6 / 1.33GHz / 20W+2W+125W / 13.1 GB/s 89.2 MB/s/W 34.3 MB/s/$ A Best Buy if it is to consider equally all aspects evaluated here.
#2 Intel Intel P55 /
2x 2GB DDR3 PC3-10600 /
Intel Core i5 760
4 / 1.33GHz / 11.5W+130W / 15.7 GB/s 153.5 MB/s/W 23 MB/s/$ Good internal transfer and power efficiency, but the cost is a bit too much.
#4 Intel Intel P55 /
2x 2GB DDR3 PC-8500 /
Intel Core i3 550
4 / 1.06GHz / 5W+2W+73W / 13 GB/s 163.3 MB/s/W 43.5 MB/s/$ Excellent power and cost efficiency of all CPU systems, however only average memory bandwidth.
#1 Intel Intel X58 /
3x 2GB DDR3 PC3-12800 /
Intel Core i7 975
4+4 / 1.6GHz / 29W+3W+130W / 26.3 GB/s 162.3 MB/s/W 20.4 MB/s/$ Without a doubt the best performer in CPU systems, but no match to the internal transfer bandwidth of a modern GPGPU.
Rank GPU Name Shaders / Speed / Power (TDP) / Price Performance Power Efficiency Cost Efficiency Comments
#4 nVidia nVIDIA GeForce GTX 480 480 / 1.4GHz / 250W / 420$ 118.7 GB/s (internal) / 6 GB/s (PCIe) 475 MB/s/W 283 MB/s/$ First generation Fermi chip has a good performance, however the price is not yet in his advantage.
#6 nVidia nVIDIA GeForce GTX 470 448 / 1.2GHz / 215W / 240$ 106.2 GB/s (internal) / 6 GB/s (PCIe) 494 MB/s/W 442 MB/s/$ It’s not the latest generation of GeForce, but it delivers good performance with medium cost efficiency.
#10 nVidia nVIDIA GeForce GTX 460 336 / 1.35GHz / 150W / 140$ 57 GB/s (internal) / 5.6 GB/s (PCIe) 382 MB/s/W 410 MB/s/$ For sure it’s not looking very promising to be the last at memory performance and power efficiency, but the price is at least OK.
#5 nVidia nVIDIA GeForce GTX 570 480 / 1.46GHz / 219W / 350$ 108 GB/s (internal) / 5.9 GB/s (PCIe) 493 MB/s/W 308 MB/s/$ The numbers are corresponding to an average overall result which is not a lot behind of the higher spec’d version.
#2 nVidia nVIDIA GeForce GTX 580 512 / 1.54GHz / 244W / 510$ 126 GB/s (internal) / 6 GB/s (PCIe) 516 MB/s/W 247 MB/s/$ The new iteration of the Fermi has an excellent result, but is by far the most expensive of the cards listed here.
#8 ATI ATI Radeon 6870 1120 / 900MHz / 151W / 240$ 104 GB/s (internal) / 1.7 GB/s (PCIe2) 693 MB/s/W 432 MB/s/$ This new entry has acceptable bandwidth performance coupled with the best power efficiency you can have.
#3 ATI ATI Radeon 5870 1600 / 850MHz / 188W / 250$ 119.7 GB/s (internal) / 2.1 GB/s (PCIe2) 633 MB/s/W 476 MB/s/$ This older generation chip from AMD is kind of a Best Buy looking at all its results.
#9 ATI ATI Radeon 5770 800 / 850MHz / 108W / 120$ 58 GB/s (internal) / 2.4 GB/s (PCIe2) 537 MB/s/W 483 MB/s/$ Maybe it’s an older generation but it scored well in aspects of efficiency from this test.
#1 ATI AMD Radeon HD 6970 1536 / 880MHz / 250W / 370$ 133 GB/s (internal) / 2.6 GB/s (PCIe2) 531 MB/s/W 358 MB/s/$ Best internal memory transfer but not as efficient as the generation it replaced.
#7 ATI AMD Radeon HD 6950 1408 / 800MHz / 200W / 300$ 104.4 GB/s (internal) / 2.3 GB/s (PCIe2) 522 MB/s/W 348 MB/s/$ Cayman Pro has a decent placement here regarding its memory bandwidth with average power consumption.

You can check your own processor or video adapter or view how other processors/video adapters measure up using the Sandra benchmarks.

Most Popular Chipsets, Video Adapters

Most popular Chipsets (and thus Mainboards) as benchmarked by users (past 30 days): Most popular Video Cards as benchmarked by users (past 30 days):
1. 0% Intel Atom Intel Atom 10.79 GBP
2. 0% Intel Intel 10.79 GBP
1. 3% Intel HD Graphics Intel HD Graphics 10.79 GBP
For a complete list of statistics, check out the Most Popular Hardware page. For a list of more products, see SiSoftware Shopping.

Comments are closed.