AMD has been overrated since at least 2006, IMHO. Maybe they have done better in areas other than ‘Desktop Duty’ and I just don’t get into those areas. I don’t know, but they have never been ‘Snappy’ on any of my desktops when compared to Intel’s. Anyway, my Ryzen 9 3900X (for $398.75 w/ free shipping ‘n no tax) arrived the other day ‘n I’m very happy with performance, but just not impressed.
I have some performance comparisons coming up. Note on the ASUS Prime A320M-K MoBo, it was a budget board, to be used in a budget build for a Ryzen Test Computer, which was to be used mainly for testing Linux Distros in a computer now named ‘Apevia’. Thusly, the AMD Ryzen Master Utility wasn’t going to work with Linux, i.e. only supported OS is “Microsoft Windows 10 64-bit version.” I don’t generally get into overclocking CPU’s, beyond an Auto setting in BIOS, and the A320M-K doesn’t allow any overclocking. I may upgrade that board in the future, but doubtful.
How Many Cores Is Enough? (oops! Lost that link – UPDATE: found it! )
snip… (desktop and laptop varieties alike) have at least four cores, which is what we consider the sweet spot for most mainstream users.
Mainstream users would be desktop ‘n laptop users.
Cores: These are the processors within the processor. Modern CPUs have between two and 64 cores, with most processors containing four to eight. Each one is capable of handling its own tasks. In most cases these days, you’ll want at least four cores–or at least four threads (see below).
Again, depending on use, 4 cores (“or at least four threads”) is a great starting point.
A single number
The effective CPU speed index approximates performance by distilling hundreds of data points into a single number. It is weighted towards typical consumer tasks.
The first few threads
Desktop tasks such as surfing the web with multiple tabs, watching videos and listening to music rarely use more than four threads. Very few of today’s popular games benefit from more than six threads. There is not much difference in fps between a 4 thread i3-9100F and an overclocked 16 thread Ryzen 2700X, in fact, the 9100F is 10% faster. CPU latency has more impact than core count. Gaming performance is primarily influenced by the GPU rather than the CPU.
Higher thread counts are useful for workstation tasks such as cryptography or virtual machine hosting. If dedicated GPU hardware (NVENC/QuickSync) is not an option, streamers and video producers could, suboptimally, use additional CPU cores for encoding. On the 27th of March, 2020 UserBenchmark’s six core database server averaged 10,000 queries per second with a CPU load of just under 10%. High data throughput is more sensitive to latency than core count.
CPU memory latency
Lower latency results in quicker data retrieval and faster computations. Multi socket server CPUs have much higher latency than single socket CPUs. In a multi socket configuration, data often has to pass through an additional memory controller before arriving at its destination. Desktop architectures also exhibit different latency characteristics. The Zen CPU architecture has significantly higher latency (70 ns) than Skylake (45 ns) which is partly why Skylake delivers superior gaming (fewer frame drops) and higher database throughput despite having comparable processing cores.
Four threads are enough for regular ‘Desktop Duty’, six threads covers most all of the popular games, and gaming is “primarily influenced by the GPU” ‘n *NOT* the CPU. “Higher thread counts are useful for workstation tasks such as cryptography or virtual machine hosting,” and I will add it is also useful for a ‘Developer Workstation’ like the one Linus Torvalds recently built to help him with the Linux kernel.
UserBenchmark is an excellent site that focuses a little on everything computer orientated ‘n with a strong focus on gaming. They provide a Gaming ‘n a Desktop ‘n a Workstation benchmark in their comparison results (Note: Test doesn’t work with Linux).
That was the AMD Ryzen™ 5 3400G, an APU with Radeon™ RX Vega 11 Graphics, but I had the GeForce GTX 1660 installed for the testing. All processors with their own graphics are incredibly slow in UserBenchmark tests for Gaming…and GPU’s change so fast that the 1660 is slow compared to the new ones that come out all the time. However, most processors with their own graphics work fine for most Desktop users…the Radeon™ RX Vega 11 Graphics also is a little faster than the Intel® UHD Graphics 630, with 17% ‘Surfboard’ performance compared to Intel’s 13% ‘Tree Trunk‘ performance for Gaming.
Ryzen™ 9 3900X had 10% increase over Ryzen™ 5 3400G in Gaming score, 10% increase in Desktop score, and 25% increase in Workstation score. The Ryzen™ 5 3400G is around $140-150 ‘n and has its own graphics if you’re on a budget…the Ryzen™ 9 3900X is holding steady at $430 (unless you can find one w/o the cooler/fan – tray type). Here’s the Intel® i9-9900:
Ryzen™ 9 3900X has 4 more cores ‘n 8 more threads. The Intel® i9-9900 ‘Antec Jr.’ main computer has 32GB ram instead of 16GB ‘n a M.2 PCIe SSD…it also has the GeForce GTX 1660 Super. The Intel® i9-9900 has 7% increase over the Ryzen™ 9 3900X Gaming score, a 22% increase in the Desktop score, and a surprising 34% increase in the Workstation score.
Other benchmark tests on CPU Multi Thread performance had a huge advantage for the Ryzen™ 9 3900X, but I’m not into cryptography, virtual machine hosting, and nor am I a developer working on the Linux kernel…just a desktop user.
I’ll add this to the Ryzen™ ‘Stuff‘ page…
Those results are from 10 months ago, and benchmarks can change as time passes so I doubt that the i9 9900 would show that result now…probably closer to the Ryzen 9 3900X’s present 61%.