CPU Benchmarks: Synthetic
All of our benchmark results can also be found in our benchmark engine, Bench.
Dwarf Fortress 0.44.12: Link
Emulating the ASCII interfaces of old, this title is a rather complex beast, which can generate environments subject to millennia of rule, famous faces, peasants, and key historical figures and events. The further you get into the game, depending on the size of the world, the slower it becomes as it has to simulate more famous people, more world events, and the natural way that humanoid creatures take over an environment. Like some kind of virus.
For our test we’re using DFMark. DFMark is a benchmark built by vorsgren on the Bay12Forums that gives two different modes built on DFHack: world generation and embark. These tests can be configured, but range anywhere from 3 minutes to several hours. After analyzing the test, we ended up going for three different world generation sizes:
- Small, a 65×65 world with 250 years, 10 civilizations and 4 megabeasts
- Medium, a 127×127 world with 550 years, 10 civilizations and 4 megabeasts
- Large, a 257×257 world with 550 years, 40 civilizations and 10 megabeasts
Dolphin v5.0 Emulation: Link
Many emulators are often bound by single thread CPU performance, and general reports tended to suggest that Haswell provided a significant boost to emulator performance. This benchmark runs a Wii program that ray traces a complex 3D scene inside the Dolphin Wii emulator. Performance on this benchmark is a good proxy of the speed of Dolphin CPU emulation, which is an intensive single core task using most aspects of a CPU. Results are given in seconds, where the Wii itself scores 1051 seconds.
3D Particle Movement v2.1: AVX2/AVX512
This is the latest version of this benchmark designed to simulate semi-optimized scientific algorithms taken directly from my doctorate thesis. This involves randomly moving particles in a 3D space using a set of algorithms that define random movement. For v2.1, we also have a fully optimized AVX2/AVX512 version, which uses intrinsics to get the best performance out of the software.
Tiger Lake wins here as it has an AVX512 unit.
y-Cruncher 0.78.9506: www.numberworld.org/y-cruncher
If you ask anyone what sort of computer holds the world record for calculating the most digits of pi, I can guarantee that a good portion of those answers might point to some colossus super computer built into a mountain by a super-villain. Fortunately nothing could be further from the truth – the computer with the record is a quad socket Ivy Bridge server with 300 TB of storage. The software that was run to get that was y-cruncher.
This is another AVX-512 test.
Linux OpenSSL Speed: SHA256
One of our readers reached out in early 2020 and stated that he was interested in looking at OpenSSL hashing rates in Linux. Luckily OpenSSL in Linux has a function called ‘speed’ that allows the user to determine how fast the system is for any given hashing algorithm, as well as signing and verifying messages.
Conclusion
Having integrated graphics in a desktop processor saves the need for a discrete GPU when a screen output is needed. Having that as a fallback is always handy, however the question as to whether anyone needs anything more than that is an important paradigm to explore.
In mobile platforms, having integrated graphics is absolutely important to keeping overall power low based on the various synergies that are made when both CPU and GPU are built on the same piece of silicon. Mobile platforms can also take advantage of high-bandwidth low power memory, unlocking a lot of performance.
Consoles are basically big mobile processors, bridging the gap between mobile and desktop by having desktop-class performance and die sizes, but still using the mobile philosophy of high bandwidth and low power. Consoles also benefit from having a heavily optimized driver stack and constant hardware capabilities over the lifetime, enabling developers to get the most out of what is available.
On the desktop is where it gets messy. Desktop platforms by design are limited to DDR memory, which is higher power and lower bandwidth, but enables a lot more customization. It doesn’t take much for a low level discrete card ($100-150 ) to surpass the integrated graphics, but that $100 level means that discrete solutions below this price are more for function than performance.
With integrated graphics on the desktop, there is less opportunity for users to customize – the moment you put in a discrete graphics card, the extra money, die size, and power spent for the integrated graphics is suddenly worth very little, except for times when debugging without a discrete card is needed. However, integrated graphics does enable smaller form factors.
Every desktop processor on the market today with integrated graphics is the mobile version repackaged with slower DDR memory. If we’re ever going to bridge the gap between a desktop integrated processor and a console, or beyond, then there has to be a suitable system paradigm. A processor with more graphics power would be bigger (increase in die area), but also more memory bandwidth would need to be added. Recently we’ve seen the older Xbox One S processors be repackaged for desktop use (that’s the A9-9820 in our tests, review coming soon), with a good die size for an integrated graphics solution. Even with slower DDR3 memory, the integrated graphics is relatively good for such an old processor. If we had something more modern, with 4-8 channels of DDR4 memory (or an onboard cache / separate cache chip), then integrated graphics could go above and beyond current solutions.
But is there a market for it, on the desktop?
For AMD, repackaging its laptop CPUs is relatively easy. As long as the memory controllers work, the only thing holding it back would be good demand for the processors as laptop processors rather than desktop models (and is in fact the situation AMD currently finds itself in). By making the Ryzen R4000 desktop series available to OEMs only for prebuilt systems, it allows AMD to focus its limited supply on the notebook segment while also supplying specific desktop customers that can more accurately track their own customer demands, rather than have to supply a full ecosystem of individual end-users. The silicon that goes into R4000 desktop APUs might also be dies that don’t quite meet the stricter voltage/power demands of the notebook, but it helps that the silicon can scale to desktop power levels.
For Intel, there has been no inclination for mobile Tiger Lake processors to come to the desktop. The situation as we understand it is a bit more dire regarding supply of the laptop variants: according to a recent report, Intel cannot fulfil the orders from the major OEMs. We have no worries that the silicon can scale to desktop power levels (we see 51 W spikes on the 28 W mode), however Intel is also set to bring an 8-core 45 W version of Tiger Lake to market next year, which might be more desktop suitable.
But back to the products at hand – how exactly have they performed?
Desktop APU vs Desktop APU
Throughout the tests, there’s admittedly not much to choose between the three AMD Ryzen 4000 processors. In a few tests the reduced core count of the Ryzen 3 pegs the performance, however the Ryzen 5 is often just a gnats wing away from the Ryzen 7. In pretty much every case, the new Ryzen 4000 performance surpasses the Ryzen 3000 APUs, although not often by much – this is partly down to how AMD has reordered from Vega11 to Vega8, choosing a different graphics combination for die area and frequency. If we compare to Intel’s best desktop integrated graphics solution, the Core i5-5775C, because it is relatively old now, AMD forges on ahead to lands anew.
Integrated vs Integrated
When comparing absolute integrated graphics performance between the desktop R4000 and mobile solutions, the Ryzen 4000 APUs appear to be ahead at lower resolution/fidelity testing, while Tiger Lake can get the upper hand at the higher resolutions. In some benchmarks Tiger Lake pulls ahead by a good margin, whereas in others it can be behind even the Ryzen 3, or sitting between the three APUs.
When comparing best against best, the differences can swing from a 55% performance to AMD (Civilization 6) to a 40% performance to Tiger Lake (Final Fantasy 14). Overall, at the lower settings, AMD has a 5.5% advantage. At the higher resolution and quality settings, Intel has a 5.8% advantage.
Integrated vs Discrete
This is where it gets a little bit tricky – discrete cards have a lot more memory bandwidth, and so can enable better graphics at times where memory bandwidth is important. If we compare the 4750G against the 2600 GT1030 for example, the integrated graphics wins in 7 titles, but when it loses, the discrete graphics card wins by 30-50% (Final Fantasy 14), especially in low quality settings. In high quality settings, often the reverse is true, and the integrated graphics wins by up to 61% (F1 2019).
When we move up to the GTX 950, which is a more expensive card, everything falls in favor of the GTX 950.
Overall
It’s clear from our data that AMD’s integrated graphics solutions aren’t great for specific games – Final Fantasy 14 being the key one. However, when pairing this level of integrated graphics with this level of CPU compute, titles like Civilization 6 and F1 2019 shine.
While AMD has not launched Ryzen 4000 APUs for end-users on the desktop, there are a number of segments with their fingers crossed that the next generation of APUs will be coming in desktop packaging. There have been rumors as to what that could be (Zen 3 Vega, or Zen 3 RDNA2), and when, and for how much. We look forward to whether AMD plan to push the integrated graphics market further, especially in light of recent launches.