Friday, May 19th 2017
AMD Talks Improved Ryzen Memory Support, Ryzen 3, and Game Optimization
AMD, in an interview with Forbes, confirmed that it is working to improve DDR4 memory support of its Ryzen series processors, to enable higher memory clocks. AMD Ryzen users find it difficult to get DDR4 memory clocks to run above 3000 MHz reliably. With memory clock being linked with the chip's Infinity Fabric clock (the interconnect between two CCX units on the "Summit Ridge" silicon), the performance incentives for higher memory clocks are just that much more.
AMD confirmed that its AGESA update for May improves DDR4 memory compatibility, although it also stressed on the need for motherboard manufacturers to improve their board designs in the future, with more PCB layers and better copper traces between the DIMM slots and the SoC socket. The company assures that more updates to AGESA are in the pipeline, and would improve performance of Ryzen processors at various levels. The AGESA updates are dispensed through motherboard vendors as BIOS updates.
The company also talked at length about game optimization, and how it has a dedicated team tasked to work with game studios on improving game performance on Ryzen processors, particularly at lower resolutions such as 1080p. At 1080p, today's games begin to get CPU-limited, and this is one area where Ryzen processors are losing ground to Intel Core processors. AMD has already worked with Oxide Games to improve Ryzen machines' performance with "Ashes of the Singularity," and is working to improve 1080p performance of other titles such as "Total War: Warhammer" and "DOTA 2." AMD also credits the recent Ryzen-balanced power plan, introduced through a chipset driver update, to improve performance in games by another 5-10 percent, which hands power-management from the OS over to CPU silicon-level Precision Boost features.
AMD also dispelled some misconceptions about the 20°C temperature offset noticed with Ryzen "X" series processors (eg: 1800X, 1600X). The Ryzen Master software places a 20°C temperature offset, which had some enthusiasts thinking that the "X" series processors have a lower throttling temperature. AMD clarified that the temperature offset never affected thermal throttling on these processors, and that it has updated Ryzen Master software to show the correct temperature.
AMD touched upon the lack of mini-ITX socket AM4 motherboards. While acknowledging the market successes of the recent mini-ITX motherboards by BIOSTAR, which are based on the X370 and B350 chipsets, AMD hinted that the broader availability of mini-ITX motherboards could be linked to its upcoming X300 chipset. Given that the mini-ITX form-factor mandates fewer expansion slots, onboard devices, and storage connectivity, the form-factor can make do with the SATA and USB ports put out directly by socket AM4 Ryzen processors, given that they are full-fledged SoCs. The X300 chipset lacks bandwidth-heavy downstream connectivity of its own, and only has the low-bandwidth machinery to keep the platform running. It also has extremely low thermal and power requirements that make it perfect for mini-ITX.
Lastly, AMD talked about the roll-out of its cost-effective Ryzen 3-series processors. The company is looking to launch quad-core and perhaps even dual-core Ryzen 3-series processors based on the "Summit Ridge" silicon, by Q3-2017 (after June). Ryzen 3 quad-core chips differ from Ryzen 5 series quad-core chips such as the Ryzen 5 1400, in that they lack SMT. Find the full interview in the link below.
Source:
Forbes
AMD confirmed that its AGESA update for May improves DDR4 memory compatibility, although it also stressed on the need for motherboard manufacturers to improve their board designs in the future, with more PCB layers and better copper traces between the DIMM slots and the SoC socket. The company assures that more updates to AGESA are in the pipeline, and would improve performance of Ryzen processors at various levels. The AGESA updates are dispensed through motherboard vendors as BIOS updates.
The company also talked at length about game optimization, and how it has a dedicated team tasked to work with game studios on improving game performance on Ryzen processors, particularly at lower resolutions such as 1080p. At 1080p, today's games begin to get CPU-limited, and this is one area where Ryzen processors are losing ground to Intel Core processors. AMD has already worked with Oxide Games to improve Ryzen machines' performance with "Ashes of the Singularity," and is working to improve 1080p performance of other titles such as "Total War: Warhammer" and "DOTA 2." AMD also credits the recent Ryzen-balanced power plan, introduced through a chipset driver update, to improve performance in games by another 5-10 percent, which hands power-management from the OS over to CPU silicon-level Precision Boost features.
AMD also dispelled some misconceptions about the 20°C temperature offset noticed with Ryzen "X" series processors (eg: 1800X, 1600X). The Ryzen Master software places a 20°C temperature offset, which had some enthusiasts thinking that the "X" series processors have a lower throttling temperature. AMD clarified that the temperature offset never affected thermal throttling on these processors, and that it has updated Ryzen Master software to show the correct temperature.
AMD touched upon the lack of mini-ITX socket AM4 motherboards. While acknowledging the market successes of the recent mini-ITX motherboards by BIOSTAR, which are based on the X370 and B350 chipsets, AMD hinted that the broader availability of mini-ITX motherboards could be linked to its upcoming X300 chipset. Given that the mini-ITX form-factor mandates fewer expansion slots, onboard devices, and storage connectivity, the form-factor can make do with the SATA and USB ports put out directly by socket AM4 Ryzen processors, given that they are full-fledged SoCs. The X300 chipset lacks bandwidth-heavy downstream connectivity of its own, and only has the low-bandwidth machinery to keep the platform running. It also has extremely low thermal and power requirements that make it perfect for mini-ITX.
Lastly, AMD talked about the roll-out of its cost-effective Ryzen 3-series processors. The company is looking to launch quad-core and perhaps even dual-core Ryzen 3-series processors based on the "Summit Ridge" silicon, by Q3-2017 (after June). Ryzen 3 quad-core chips differ from Ryzen 5 series quad-core chips such as the Ryzen 5 1400, in that they lack SMT. Find the full interview in the link below.
74 Comments on AMD Talks Improved Ryzen Memory Support, Ryzen 3, and Game Optimization
I hope AMD really improve this line-up and stick with the AM4 for a few years unlike intel's strategy which new procs must be used with new boards to get the most of the mediocre improvement.
Remember this is not an artificial limit or a result of programmers being lazy. They've already worked hard to utilize 2-3 cores in processes that are basically single-threaded.
As for Office, Excel is the only relevant applications here (with long and heavy enough jobs to make a difference). As I've said earlier: it's about the proportions of VBA and native calculations in a file.Single-core performance means exactly what the name says. It's always be very important. :)
At this point Ryzen is a way more modern architecture, so it's close on single-thread front. We'll see what happens when Intel launches new architecture and new process node.
games are not the way to test true power of a cpu. single core is very important for most users the workload is 25% when you have 4 cores. i have a lot of web pages with scripts that overload the cpu to 25% (1 core full workload) and makes everything slow.
At the moment I'm on a PC with 7-year-old CPU (fairly weak even by 2010 standards, a Pentium E5400 - 2 cores). Everything I need is running: 3 cloud storage apps, firefox, excel, music player, steam, antivir, ccleaner and possibly a dozen more. CPU load? 7-10%. :)
Overclocking is a silly argument, when the alternatives overclock much better.Actually, in reality Intel have >20% better performance per core. But in theory Ryzen have more ALUs, it's a "better" superscalar and should have much more throughput than Intel. With the right mix of integer and float it should achieve >33% more throughput. But Intel have a much better prefetcher, and better AVX support (which helps some applications). Ryzen is simply not able to feed it's execution ports well enough.
Rendering in games typically iterates a long list of objects, invoking a number of functions on each one. This causes cache misses, which stalls the CPU. Overclocking the CPU or using faster memory doesn't help, since the penalty is a time constant. The only way to improve this is to reduce the cache misses, and a better prefetcher makes a difference.
As always, real world performance matters, not theoretical figures. An i7-6800K beats Ryzen 7 1800X in most use cases, even though it only have six cores. Better cores is always better than more cores.
self ryzen calculated for 6500 mhz ram and 5 , 1080 graphic card !!!!!
they put limit on motherboard for ram above 2400 mhz = high latency and price
we buy I7 laptops exclusively (quadcore HT low clock and or quadcore no HT)
I5 desktops...
users using excel, web browsing and meetings still complain that they're slow for office use... I think the time is right..
As for laptops, most Intel-based models use the -U series of CPUs, which are all 2C/4T (from i3 to i7) - just the clocks differ. So the i7-7600U (2,8GHz -> 3,9GHz turbo) is pretty brisk, but the i3-7100U (2,4 GHz, no turbo) isn't.
The -H models have 4 cores (with or without HT), but they're designed for gaming/workstation laptops that are meant to challenge desktop performance.
Heat dissipation and battery life dictate how much speed you can pack.
Corporate laptops are different story with all the security like drive encryption, heavy AV, web filters firewalls and other stuff that provable makes more of a difference that having a i5 or i7.
AMD has about 4-7% smaller single thread IPC than Kaby Lake.
However, we should not forget that in multi threaded scenarios it achieves higher IPC (perhaps due to the way its cache works).
Now, if someone doesn't run anything that pushes PC to its limits, it doesn't matter. But if not, I want to see those "single threaded apps" he/she runs. If it is notepad, ok, my Z80 would still be ok for you, and if it is Excel, let me surprise you:
fastexcel.wordpress.com/2014/01/29/threading-and-hyper-threading-optimizing-excel-calculation-speed-by-changing-the-number-of-threads/
Ryzen has another hidden bonus: socket is there to stay until 2020, so upgrading to a superior CPU later on is an option, unlike with Intel.
For me, the main issue here is that computers are full of software garbage. Gamers run so much pointless stuff (like the huge, blinking apps for OC or GPU optimization), I'm not surprised they need 4 cores to do anything. :)I'm really looking forward to your comment when AMD releases 2C Ryzen. :)AMD has told us multiple times that AM4 will be supported for N years (N varies over time...).
AFAIK they've never said that it'll be the main socket or that i'll offer full compatibility. Or did they? Can you link something useful?
Also, just the fact that you'll still be able to run a 2020 Ryzen on your motherboard doesn't mean you'll want to do that.
Since Ryzen is a SoC, it'll age faster than Intel CPUs (only feature-wise, because it has a safe margin of performance).
Looks like selling anything below R7 is a waste, disabling working cores just to cover the market.
I wish they made separate silicon for the quad cores.
Provable we would get higher clock lower consumption and cheaper to make processor.
Just because the socket stays around, doesn't necessarily mean that the chipsets will support all the same processors. This is only a cost savings measure (you can rely on sockets, and therefore, connections from socket to other chips and ports on the motherboard to remain the same).
Re: Socket 775, 2011 (yeah, two versions, but same pin-out).
The fact is AMD has three main segments now to push Ryzen cores into, and they have already been vocal about two of them, but we are still missing one.
We have Ryzen 3/5/7 for mainstream > high end
We have Threadripper for workstation/HEDT
Aaaand... drum roll. We have APU's, just like Intel's i3, that are, will be, and always have been cost effective all-in-one solutions, the kind of solution AMD will target i3 with. Ryzen IS NOT THAT. But you did figure out it has no IGP, good job, you're getting there. Slowly.
The day AMD will make 'one processor to rule them all' only exists in fantasy.
Algorithms can be classified as either single-threaded (serial) or multi-threaded (parallel).
If a piece of software is relying heavily on parallel algorithms, it will most likely run on as many threads as it can access.
If it's relying on serial algorithms, it can only be "forced" to do few of them at the same time by very time-consuming and difficult optimization.
I find Photoshop to be a great example (but it's true for other photo-editing programs as well). In Photoshop some things are parallel and some aren't.
Many recognition/editing algorithms are single-threaded. Why? Because what happens to a pixel relies on what happened to those next to it. It's single-threaded by definition.
Photoshop manages to utilize 3-4 threads by forcing multi-threading, e.g. when you run "auto-adjust" it'll process few single-threaded, independent operations (e.g. brightening and distortion correction) at the same time. As a result, it'll use few threads. But more likely 4 than 16. It's obvious that this kind of "optimization" is limited.
Additional threads are put to use when you run batch processing. Photoshop runs a session of each file processed and easily utilizes all available cores.