AMD has a winner with Ryzen
Robert Dow on March 13th 2017 |
AMD provided us with a flashy new PC from Cyberpower PC built around the new Ryzen CPU. As everyone knows, AMD is betting big on this latest processor line, which has been designed from the ground up for performance an efficiency. The company says it has achieved a 52% gain in performance over its previous generation Excavator core processors with no increase in power. The new design features higher bandwide with an improved low latency cache. Simultaneous Multithreading enables the processor to feed the GPU and the company says its energy efficient FinFET design will en¬able it to scale up to enterprise products and down to competitive, low cost clients. Is that enough? AMD says there’s a lot more to come, but our testing puts AMD’s new CPU in the game.
The PC was built for AMD by Cyberpower PC and measured 20-inches deep, 20-inches tall, and 9-inches wide, and was adorned with red lights and very neat wiring harness that made AIB replacement easier because you didn’t have to fight with unwieldy power cables.
We ran through a battery of CPU and PC tests (including graphics). We compared the 8-core, 3.7 GHz AMD Ryzen7 to a 4-core, 4GHz, Intel Core i7- 6700k, which is not exactly an apples-to apples comparison, but close enough.
We ran all the benchmarks at 3180 × 2160, on a 31.5-inch Dell UP3214Q monitor without freesync or gsync (and we did see some tearing in some of the programs as a result).
The tests speak for themselves, and the PC tests are shown in the next chart.
The PCMark work load tests came out about the same, but AMD did very well in the PassMark and GeekBench tests.
However, we wanted to see how it would perform in games, we ran a few games and a gaming-related benchmarks. We sought to neuturalize the effect of the GPU by using the same GPU in both systems. The Cyberpower PC came with an EVGA GTX 1080 AIB, so we also tested it with a single AMD Radeon RX 480, and two RX 480s in Crossfire mode. We chose Crossfire as a means of normalizing the GPU FLOPS performance for CPU comparisons. The results were impressive, and interesting, and are shown in the next chart.
The game test scores indicate Intel is the processor to have if playing Fallout 4, but AMD is on par or better in all other benchmarks.
We took the average of the scores and compared them in the following table.
The above table and charts indicate there is little performance difference in between the two CPUs, and tests with other Intel CPUs will obviously get different results. However, based on the above we chose to use the Pmark to break the tie.
Once again, the scores are close, but Intel’s 18-month-old CPU has the advantage with a 72% lead over AMD’s new CPU. Intel won because it has a lower price and lower power consumption. AMD can certainly adjust its price, but probably can’t do much about the power consumption (power consumption data provided by the suppliers, prices off the web).
What do we think?
AMD didn’t knock the ball out of the park, but did demonstrate that a fabless x86 Supplier can play with the 80-pound gorilla and bring a lot of performance to the market. Also, the Intel processor we chose, based on what we had, probably isn’t the one AMD is targeting, although it is well respected gaming CPU.
More importantly, this puts AMD back in the game (no pun intended) with the gaming community, a place where the company was once the choice for CPU.
We were also interested in seeing if AMD’s claim that two RX 480s were as good or better than one GTX 1080, and for less money. The average scores for all the tests resulted in 54 for AMD Crossfire, and 52 for Nvidia.
The test results indicate that two RX 480s can beat one GTX 1080, and even though Nvidia lowered their price (see GTX 1080 ti this issue) the economic argument is still true. The GTX 1080 now sells for $499 (it was $599), and two RX 480s cost $460. However, the two RX 480s draw 300 watts, compared to the GTX’s 180, so if the performance scores were the same, AMD would win the Pmark with a 35% lead.
AMD’s claim that two RX 480s were as good or better than one GTX 1080, and for less money. The average scores for all the tests resulted in 54 for AMD Crossfire, and 52 for Nvidia.
The test results indicate that two RX 480s can beat one GTX 1080, and even though Nvidia lowered their price (see GTX 1080ti this issue) the economic argument is still true. The GTX 1080 now sells for $499 (it was $599), and two RX 480s cost $460. However, the two RX 480s draw 300 watts, compared to the GTX’s 1080, so if the performance scores were the same, AMD would win the Pmark with a 35% lead. – R.D.