SoylentNews is people
SoylentNews
AMD announced its first Zen 3 (Ryzen 5000 series) desktop CPUs on October 8.
Compared to Zen 2 (Ryzen 3000 series) CPUs, the Zen 3 microarchitecture has higher boost clocks and around 19% higher instructions per clock. A unified core complex die (CCD) allows 8 cores to access up to 32 MB of L3 cache, instead of two groups of 4 cores accessing 16 MB each, leading to lower latency and more cache available for any particular core. TDPs are the same as the previous generation, leading to a 24% increase in performance per Watt.
AMD estimates a 26% average increase in gaming performance at 1080p resolution, with the Zen 3 CPUs beating or tying Intel's best CPUs in most games.
Ryzen 9 5950X, 16 cores, 32 threads, boosts up to 4.9 GHz, 105W TDP, $800.
Ryzen 9 5900X, 12 cores, 24 threads, boosts up to 4.8 GHz, 105W TDP, $550.
Ryzen 7 5800X, 8 cores, 16 threads, boosts up to 4.7 GHz, 105W TDP, $450.
Ryzen 5 5600X, 6 cores, 12 threads, boosts up to 4.6 GHz, 65W TDP, $300.
You may have noticed that these prices are exactly $50 more than the launch prices for the Ryzen 3000 equivalents released in 2019. The 5600X is the only model that will ship with a bundled cooler.
The CPUs will all be available starting on November 5. AMD will stream an announcement for its RX 6000 series of high-end GPUs on October 28.
See also: AMD Zen 3 Announcement by Lisa Su: A Live Blog at Noon ET (16:00 UTC)
AMD Teases Radeon RX 6000 Card Performance Numbers: Aiming For 3080?
Previously: AMD's Zen 3 CPUs Will Not be Compatible with X470, B450, and Older Motherboards
AMD Reverses BIOS Decision, Intends to Support Zen 3 on B450 and X470 Motherboards
AMD Launching 3900XT, 3800XT, and 3600XT Zen 2 Refresh CPUs: Milking Matisse
AMD Zen 3, Ryzen 4000 Release Date, Specifications, Performance, All We Know
(Score: 5, Interesting) by richtopia on Friday October 09 2020, @01:27PM (11 children)
I build an AM4 system with the 3600 about 8 months ago. I've been happy with it, but I've expanded my hobby of recording my Friday gaming sessions and would like something with more cores.
With that context explained, I'm curious what will happen to the prices of the 3xxx chips already on the market. AMD has been pretty good with reducing Ryzen prices as new generations arrive, and I wouldn't be surprised if we see the new Ryzen 5 5600X competing with the existing Ryzen 9 3900X. I'm really excited for the efficiency gains of the new chips, but I'm wondering how twice the number of older cores will compete.
Everything is speculation until reviews arrive in November. And for my personal use case the CPU probably isn't what I should upgrade first: I'm still waiting for AMD and Intel to release their new GPUs before I make a decision there. Exciting year for PC hardware across the board.
(Score: 2) by takyon on Friday October 09 2020, @03:16PM
That's the gold price and the iron price. I wouldn't expect huge price drops on Ryzen 3000 chips right now since the launch prices of Ryzen 5000 are higher, creating a gap in the first place. Ryzen 5000 prices will probably drop by March, when Intel's Rocket Lake is released. Until then, AMD has free reign.
The efficiency gain seems to be a direct result of the performance gain (IPC and clock gain) at nearly the same power usage. Process node benefits (it's assumed to be TSMC N7+) were put into raising the clock speeds. Though undervolting/underclocking your CPU could show great efficiency benefits.
For you, if your motherboard is compatible, you could consider getting a Zen 3 with higher core count in 2022+ after Zen 4 is out. 8 cores for better gaming performance at that time (games will start using at least 8 cores because of the consoles), or 12+ cores if it matters.
AMD is definitely aiming for RTX 3080 performance with its top Big Navi GPU. It could land just short of that, or into RTX 3090 territory (since the gap between 3080 and 3090 is not much).
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Friday October 09 2020, @03:16PM (1 child)
> I build an AM4 system with the 3600
I think I may have one of those, or a 3400G or something. It used to be easy with Pentiums (I, II, III, IV) and Athlons, now I'm never quite sure which part is which and that makes discussing them on a site like this difficult. At a certain point reading the parts lists, my eyes glaze over, I mentally detach and begin staring into the middle distance.
Perhaps it's me but wouldn't it be better all round if CPUs were named like $marketing_name-$process-$cores-$speed-($graphics_core)?
(Score: 2) by takyon on Friday October 09 2020, @03:23PM
https://www.fosslicious.com/2020/06/cpu-x-alternative-to-cpu-z-for-ubuntu.html [fosslicious.com]
AMD might be doing that by skipping the 4000 series naming, but we'd have to observe several more launches to know for sure.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 3, Interesting) by Freeman on Friday October 09 2020, @05:29PM (7 children)
Like you, I've got a 3600 running in my latest machine. I'm going to be waiting for a price drop on the new CPUs, before I upgrade. I can also tell you now, you're likely to get more noticeable performance increases by switching from HDD *retarded slow speed* to SSD *400MB/s* or SSD to NVMe SSD *1600 to 3200MB/s*, than any upgrade you do.
I've been playing Space Engineers lately and it's a pretty demanding game. CPU usage goes back up and down not really hitting a lot more than 25%, while my RX480 has hit around 50-60% usage. Admittedly, I'm running at 1360x768 resolution on an old 32inch lcd monitor. With higher resolution it may increase CPU usage a bit. I would expect it to hit the GPU a lot more, though.
I haven't run Mount and Blade II: Bannerlords for a while, because 40GB+ update . . ., but that would likely be a much better game to test.
Really, just get something like https://www.nzxt.com/camapp [nzxt.com] and you'll see the usage while you play and record. Then, you'll be able to make a bit more of an informed decision on, "Do I actually need anything better than what I've got?". I thought that 32GB RAM may have been a bit of overkill, but looking at the usage in a game like Space Engineers, 16GB is almost too little. 32GB looks a lot better when you factor in running a music player, a browser, a chat program, and / or a streaming/recording program, while playing games.
Joshua 1:9 "Be strong and of a good courage; be not afraid, neither be thou dismayed: for the Lord thy God is with thee"
(Score: 0) by Anonymous Coward on Friday October 09 2020, @06:48PM (3 children)
yeah, when i upgraded from my bulldozer guts recently, i went with a ryzen 5 2600 for $125, but got an nvme ssd at 3000 MB p/s. the cp+ram is 2-3 times faster(IIRC) and the ssd is 6 times faster. bought the asrock hdv MB for like $50. very cheap guts and plenty powerful. maybe in 5 years i'll move up to zen4 or something. all i do is code and surf for the most part and the compile times are fast enough for me at this point. if i start writing huge compiled lang applications maybe i'll care more then.
(Score: 3, Interesting) by takyon on Friday October 09 2020, @07:38PM (2 children)
Zen 4 could have a big (1 GB or larger) L4 cache stacked on the I/O die. That could contribute to the general IPC gains, or just benefit certain applications. There's also rumors [wccftech.com] of a graphics chiplet, so you might not need a discrete GPU anymore. Higher core counts are likely, e.g. 24 cores for a 5950X replacement. A TSMC "5nm" node should allow for smaller 8-core chiplets or 12 cores per chiplet.
If all of the above is correct, Zen 4 would be a great option for someone who wants absurdly high compute performance, but doesn't care about gaming (although an RDNA 3 chiplet might be sufficient for gaming too).
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Saturday October 10 2020, @12:43PM
you can get a pci-e slot card that has a socket for sfp+ module. these modules have 10Gbps versions ... or if you have the estate ... 80km... from your pci-e slot ... errr ...
(Score: 0) by Anonymous Coward on Saturday October 10 2020, @06:53PM
Yep, that's me. For me, programming IS the game.
(Score: 4, Informative) by turgid on Friday October 09 2020, @07:21PM (2 children)
I've got a Ryzen 3600 but I put 32GB of RAM in it. It's enough RAM that I can have a 16GB temp fs all in RAM and run things off that, so I can completely bypass hard disks. That way you can easily saturate the CPU.
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].
(Score: 0) by Anonymous Coward on Saturday October 10 2020, @06:44PM (1 child)
i've done some of this before, but didn't find a robust solution to things at the time. what are you using to sync between ssd and tempfs for persistance?
(Score: 0) by Anonymous Coward on Sunday October 11 2020, @10:00AM
There are a couple solutions you can use. One is to increase the time data spend in the buffer before flushing. Another is to use a union file system. Another is to use a synchronizer like asd or unison or even a properly configured rsync. The most robust solution is a combination of the three.
(Score: 2) by fakefuck39 on Friday October 09 2020, @11:04PM (15 children)
I'll approach this from a datacenter pov, since that's what I do and sell. literally no one is buying AMD, irrelevant of how many spiffs are given out. why? let's see:
AMD beats Xeon on performance, and is cheaper. If I'm putting in a compute cluster with 500 xeon cores, I can accomplish the same on AMD for half the cost, and save about $200k on the million dollar cluster. Why does no one do it?
This is because AMD is not taken seriously, and it shouldn't be. Server architecture from Intel always uses an older, more proven node process. This means lower clock, and slower RAM. And you compensate for that by getting more cores. Because guess what - you don't want your important shit to run on new and shiny. You want it to run on something solid. So AMD loses.
Laptops are what people get these days for a computer. AMD sucks donkey balls on laptops compared to Intel. It'll get hotter and use up your battery faster. It's a desktop CPU company.
Who's getting deskops these days? A few gamer nerds who don't like leaving the house in their free time. So viola AMD - nice job picking a solid and large target market.
AMD's a joke.
(Score: 2) by Pav on Saturday October 10 2020, @12:10AM (9 children)
The #1 supercomputer in the world runs neither... it's ARM. Just sayin'.
The datacentres of yesterdays lean-and-hungry success stories (ie. todays big boys) will stick with what worked for them. 20 years ago this was called the "noone was ever sacked for buying IBM" mindset. The new lean-and-hungry guys will stand things up on efficient new tech. It has been forever thus.
(Score: 2) by fakefuck39 on Saturday October 10 2020, @12:26AM (8 children)
Umm, no. All the new "lean" players from Nasuni to Nutanix to Rubrik use Intel, and the new solutions are not "lean" by any means - they are more expensive than the old players. Stop looking at list prices - the old guys discount 60-80% off list, the new guys discount 50%.
There is literally almost no AMD. A specialized "supercomputer" running ARM says nothing about what CPUs people use. Just because someone built 200 nodes and called it a single computer doesn't mean anything. I literally sell this shit. The amound of AMD I've sold in the last decade can be counted on zero fingers.
AMD is not "lean" - they hit the same performance for less because they sell shit that's not as proven or tested. No one is going to risk production on that.
(Score: 2) by Pav on Saturday October 10 2020, @02:13AM (2 children)
You're talking about companies that raised investor cash ten or more years ago. In what world are those the "the new guys"? The actual new guys are the ones building out clusters from their own pockets, and who's idea of being conservative will be replicating their AMD-stocked cabinets once investor cash starts burning a hole in their pockets.
(Score: 2) by fakefuck39 on Saturday October 10 2020, @08:15AM (1 child)
Rubrik is 6 years old, GA products 4 years ago.
Nutanix 11, GA products 7 years ago.
Nasuni 12, don't know when their shit became for sale.
Cohesity 7 years old, GA 3 years ago.
Qumulo 8, started releasing GA products last year
All use Xeon. Pure arrays use Xeon, also a new player. Literally every enterprise appliance available uses Xeon chips, especially the new ones. Because when you have a new product you're trying to work out your bugs and get it out the door, and you don't want to worry about a CPU causing potential problems.
If there are guys building clusters from their pockets, and they're trying to save $1k on a CPU for a product that costs 100k+, those guys don't succeed. AMD has been around since 1970. Whatever they were going to start in the datacenter would have started by now. No investor VC cash is going to stand behind a plan that takes unnecessary product failure risk, to save a penny here and there. It is funny how you can't actually name anything on the market. Something tells me the guy who literally designs and sells this shit (me) and the people who are successful at making this shit, know a lot more than you, with your basement gaming rig. Enjoy mario 4k, and stick to nintendo please, not the datacenter.
(Score: 2) by Pav on Tuesday October 20 2020, @11:22AM
I knew sales guys have to learn to like the smell of their own farts, but you're REALLY huffing. :)
I suppose new startups never leverage newer cheaper (sometimes even inferior in some ways) technology against The Way Things Are Done, and in the process roll the established order. In the next year or three you'll be pulling AMD products from your g-string and waving them around like it has forever been thus. ;)
(Score: 0) by Anonymous Coward on Saturday October 10 2020, @03:34AM (2 children)
Intel is more tested... that's why so many vulnerabilities have been found in it.
(Score: 2) by fakefuck39 on Saturday October 10 2020, @08:06AM
The previous node process is more tested, because it ran longer. New node processes have issues all the time, both for Intel and AMD. So when something just came out, it is less tested and proven than something that has been running for 5 years. What is it you're not getting here Sherlock?
You're on an airplane with your family. Do you want that plane to use engines that have a million miles of proven flight time, or do you want to fly on ones that were released last week?
(Score: 2) by fakefuck39 on Saturday October 10 2020, @08:38AM
(Score: 0) by Anonymous Coward on Saturday October 10 2020, @06:37AM (1 child)
The ARM supercomputer that is the fastest supercomputer in the world is by NEC. NEC was a sparc shop who made big iron while Sun supplied the smaller stuff, and purportedly these new NEC arm processors are based on their sparc processors just using the arm instruction set. The CPU's vector units which are what gives this machine its performance (it does not use accelerators like GPUs that you find in Intel HPC clusters), are of NECs own design. But, to GP's point, it ain't Intel.
(Score: 2) by fakefuck39 on Saturday October 10 2020, @07:59AM
a supercomputer is not a computer. it's a supercomputer. it's a cluster of many computers. just because you put a bunch of computers on a network does not make each node more powerful - it just means you have more nodes. to make a node more powerful, or on the topic of this specific discussion - to make a CPU more powerful, ARM does not come even close to competing with other architectures such as Power or x86.
And by the way, Vector units, in addition to being things like SSE/MMX/etc, are what a GPU is made of.
So all you're saying here is a cluster of thousands of underpowered nodes, each running an OS, perform fast on a highly parallel workload. Which has zero to do with how fast a CPU is. A CPU, in case you didn't know, is a bunch of cores on a die. Not a cluster of networked computers.
(Score: 2, Insightful) by Anonymous Coward on Saturday October 10 2020, @06:29AM (1 child)
grep vendor_id /proc/cpuinfo
vendor_id : AuthenticAMD
vendor_id : AuthenticAMD
From a VM on a large cloud provider.
if you google:
amd aws
amd azure
amd gce
amd oracle cloud
amd ibm cloud
*Every* one of the above searches returns a bunch of links announcing AMD Zen based VM instances.
So, I call bullshit.
If you said that Intel still outsells AMD by a substantial margin, that is easy to believe, change comes slowly, and hardware vendors have been slow to add AMD offerings. But, AMD has a much better offering. Not just performance/$. Their full memory encryption is almost there in the current gen, so expecting it will be good in the zen3 epycs. The memory encryption will make AMD very attractive to customers of shared computing resources. At my work, we were an AMD shop in the early Opteron days when AMD was on top, we moved to Intel Xeon when that was better, and we are waiting until next year to replace all our Intel servers and going all-in AMD again. The value proposition is a no-brainer. If Intel doesn't get its shit together, it is going continue losing customers.
For smaller places that use VMWare, there were changes to their licensing that hurt AMD. VMware had been licensed per socket, but now they treat an AMD 64 core processor as two sockets for licensing. If more follow suit, that will help intel too, but AMD is still better perf and much better perf/$ even in the lower core counts, just less of a no-brainer since now cannot cut licensing costs in half by switching. Because of this licensing B.S., I predict AMD to have some higher clock 32 core chips in their next gen at less attractive pricing, but still better pricing than Intel.
(Score: 2) by fakefuck39 on Saturday October 10 2020, @08:02AM
yes, every cloud provider, and even every VAR I've worked for gives people the option to buy AMD. And that node you ran the grep on is what, .01% of that cloud's CPU footprint? Change is slow you say? How slow? AMD has been around since 1970. Any day now I guess.
(Score: 0) by Anonymous Coward on Saturday October 10 2020, @06:49PM
lmao! i don't give a flying rat's ass what stupid, sycophantic, Suited Whores do. These are the same dumb whores who run their shit on Azure and AWS. fuck all of them.
(Score: 1, Insightful) by Anonymous Coward on Saturday October 10 2020, @07:01PM (1 child)
"Literally no one"
Maybe you haven't been paying attention for the last five years or so. That used to be the case. From 2016 to 2018 they went from zero market share to 5%.In 2019 that went up to 8%, and as of July it had passed 10%.
Intel's "proven" technology is only proven to have more bugs, and Intel only uses older technology because none of their new technology works. Intel is completely dead in the water getting by on nothing but brand name and inertia.
(Score: 2) by fakefuck39 on Monday October 12 2020, @10:31PM
Yes, as you note AMD's market share in the datacenter went down from the 20% it had 15 years ago to the 10% it has now. Thank you for agreeing with me. Proven technology is what it is - proven by running for several years. Which is why they did and do completely dominate AMD. Because no one wants to run production on shiny new stuff that's unproven. I guess "dead in the water" means "completely dominates and always has" to you. Oh the mental gymnastics on this one. Dance clown!