AMD Fires Back with 7 New Chips

https://www.tomshardware.com/news/amd-ryzen-7-5800x3d-5700X-Ryzen-5-5600-5500-4600G-4500-4100

by ItsTotallyOn • 4 years ago

161 155 4 years ago

piinbinary• 4 years ago

Depending on how the performance vs. Intel shakes out, I could see the Ryzen 5 5500 becoming the new default option for mid-range PC builds

komuher3• 4 years ago

i3-12100f is almost 100 usd cheaper in most builds you would use extra cash for better gpu or just get i5-12400f with the same / lower price and higher clocks but lets wait for benchmarks.

Retric• 4 years ago

Ryzen 5 5500 is a $159 6 core chip.

i3-12100f is a 100$ 4 core chip, which should line up with Ryzen 3 4100 at $100 or possibly Ryzen 5 4500 at $129.

WithinReason• 4 years ago

How much more expensive are Intel motherboards though?

jeffbee3• 4 years ago

I am interested in your belief that most builds needs GPU more than CPU, or that most builds even have GPUs. iGPUs have ~70% of the market. Most builds are going to spend the cash on CPU performance.

noselasd• 4 years ago

How can we get these vendors to do versioning of their products that we can relate to, or atleast somewhat grasp ?

gruez• 4 years ago

The only "new" chip is the X3D one (with 3d v-cache). The others are the same chips as last year, just at a different price point (eg. the 5700X is basically a cheaper version of the 5800X, only being marginally slower).

zekica3• 4 years ago

They also seem to have released a bunch of CPUs that are APUs with disabled GPUs

  Ryzen 5 5500 - 5600G with broken and disabled GPU
  Ryzen 5 4500 - 4600G with broken and disabled GPU
  Ryzen 5 4100 - 4300G with broken and disabled GPU

kibwen• 4 years ago

I concur, as someone who only pays attention to the processor market twice a decade I'm clueless when it comes to navigating the matrix of CPU market segmentation. It makes me yearn for automobile-style naming where you just have a year, a make, and a model.

kemotep5• 4 years ago

Largely those style names do exist in the current product names. Since Intel styled the Core i-series 3, 5, and 7 (and the later 9) over 12 years ago, AMD is following their lead with Ryzen 3, 5, 7, and 9.

The first digit is the generation and then the second digit and suffix letter designate additional features. So a Ryzen 7 5700X is a 5th gen Ryzen 7 meaning it has 8 cores. It is a step down in price and performance to a 5800X and also has no Graphics capabilities (those have a G suffix). Intel uses different letter suffixes such as the Core i7 12700KF being a 12th generation Core i7 with overclocking features (K) and no graphics (F).

It is far worst with GPUs. Certainly an industry wide naming convention reset would be nice but everyone is not going to cooperate like that.

It does take a minute to read up on but at least Intel has *mostly* followed this naming convention for 12 product generations so far.

tormeh• 4 years ago

Just look at benchmarks and prices. Names can't solve anything.

TillE• 4 years ago

You would still want to refer to specs and benchmarks to make good comparisons, so a "better" naming scheme doesn't actually solve any problems.

flatiron2• 4 years ago

Remember athon 1800+ naming scheme?

I’m not sure what is better. Amds weird numbers of apples ultra max pro m1

IE63• 4 years ago

Even within one vendor it's super confusing. i3? i5? i7? Y series? U series? I think 12th gen Intel may even use different naming conventions already... A Y series i7 of olde delivered a significantly different experience than an i3 U series (literally the difference between being able to run some workloads comfortably vs. locking the machine up). And then marketing materials make it harder because they will drop the CPU model name for the nondescript "i5 CPU". I think it's probably somewhat intentional.

paulmd14• 4 years ago

if they don't call out mobile/ULV processors with their own series (Y and U series) then people whine that they're trying to "sneak lower-performing processors into the lineup using the same numbering".

similarly, marketing a processor as being a "5700U" or "5700G" despite it being slower than a desktop 5700X could be seen as equally deceptive. That's not really any better than "i7" vs "m7" or whatever.

unfortunately you're just going to have to learn the naming convention, there really isn't a good solution for that, given the wide range of applications that a given series might be applied to. You just happen to think that AMD's naming convention is worth taking the time to learn while throwing your hands up at the Intel naming. Same with people who think Intel is just awful for internally codenaming all their products after Lakes and Coves while eagerly memorizing every single painter and city name that AMD uses in their codenames - says more about your priorities than their naming scheme.

n00bface• 4 years ago

At least it's not monitors. I just pre-ordered a AW3423DW. The sexiest monitor version/model ever.

Sohcahtoa82• 4 years ago

Intel and AMD use a similar naming scheme, but get a little confusing with suffixes, especially AMD.

Both include a target market, generation, and performance level. With Intel, it's "i{market}-{generation}{performance}", with market being 3 (budget), 5 (mainstream), 7 (enthusiast), 9 (high-end). The numbers used in the "performance" level varies, but higher is always better. For example, i7-3770 is the third-gen chip targeted towards enthusiasts and was the highest performer in its generation. My i9-9900 is the top end model for the 9th gen. Intel will also use suffixes "K" to mean it has an unlocked multiplier (making overclocking easier) and "F" which means it does not have an on-die GPU, so you'll need a discrete GPU card.

AMD is similar, they'll call them "Ryzen {market} {generation}{performance}", ie, Ryzen 5 5600. But where AMD goes crazy is with the damn suffixes that express an additional performance level that is impossible to decipher.

zekica3• 4 years ago

Desktop CPUs:

  X - sightly higher performance one than the one with no suffix
  XT - even higher performance

Desktop APUs:

  G - includes graphics, different architecture than non-G CPUs (65W)
  GE - lower power than equivalent without E (similar to T with Intel) (35W)

Mobile APUs:

  U - lower power (15W, configurable)
  H - higher power (35W, configurable)
  HS - lower clocks than H
  HX - higher clocks than H

dsr_• 4 years ago

Sadly, AMD is even more opaque than that.

A Ryzen 2700 is a second-generation CPU with 8c/16t.

A 3700 is a third-generation CPU with 8c/16t

A 4700g is a second-generation CPU with 8c/16t and a GPU

A 4500h is a second-gen CPU with 6c/12t and a GPU

There is no 4600G.

An X suffix generally means higher clockrate; an H suffix generally means high efficiency (lower power draw); a U suffix means ultra-efficient (very low power draw).

In the 5000 and 6000 series, possible suffices include X, H, U, HX, HS, but not G.

Some CPUs are OEM-only.

Some CPUs are only sold in packaging for laptop/tinybox manufacturers.

In general, if you get integrated graphics, you lose an entire processor generation, but more recently you only lose top-end features.

sliken7• 4 years ago

Heh, I think the best we can do is embarrass them and complain when they do something stupid and confusing like some of the Ryzen 5XXX series being Zen 2 and others are Zen 3 cores. Not that Nvidia and Intel hasn't done the same thing.

Using http://ark.intel.com has been very useful for figuring out the details of Intel CPUs. Not found anything as useful for AMD or Nvidia.

culopatin3• 4 years ago

I thought it was pretty clear no? The higher the number the better and the G means graphics? Am I missing something?

JohnTHaller• 4 years ago

I'd love to see proper testing of all the current chips from both AMD and Intel with Specter v2 mitigations applied. Performance hits of up to 50% are being shown in some workloads.

paulmd14• 4 years ago

AMD really also should be benched with KPTI turned on, since they've got a Meltdown-style vulnerability (discovered by the same team) they've left unpatched because "KPTI fixes it", but they also want KPTI off-by-default because it tanks their benchmarks. It also completely breaks KASLR (and that's been broken on AMD for a while thanks to prior work from the same team).

https://mlq.me/download/amdprefetch.pdf

Sadly people don't seem to take AMD vulnerabilities very seriously, Specter/Meltdown were trumpeted from the rooftops but when AMD leaves vulnerable defaults because mitigations would tank their benchmarks then it's fine, and everyone continues to benchmark in vulnerable configurations since it's "manufacturer-recommended".

There seems to be a mindset for many that because the initial vulnerabilities didn't affect AMD that they're invulnerable forever.

throwoutway• 4 years ago

There will be many mort such security issues. So dual benchmarks should be standardized from here on out:

With protections turned on

With protections turned off

jeffbee3• 4 years ago

I can't think of anything less relevant for gaming workloads than spec. ex. security. In that use case you'd obviously disable all of the mitigations, to get the highest possible speculative performance.

kemotep5• 4 years ago

The biggest part of this announcement is the bios updates for 300 series motherboards in my opinion. People who bought into the AM4 ecosystem 5 years ago should be able to update to these brand new zen3 cpus.

Have we ever had such a long lived socket and chipset before? Supporting brand new products for 5 years?

toast0• 4 years ago

Socket 7 was around for a long time and supported a multitude of different processors from different companies. You could usually put a socket 5 processor into a socket 7 board as well (but not the other way). And that was a lot larger jump between the first and last processor.

kemotep5• 4 years ago

It looks like AM3+ went from 2011 to the first generation of Ryzen launch in 2017. Still you are right, socket 7 takes the crown for longevity and performance boost.

Anecdotally, the increase in performance from my Ryzen 5 1600 to the Ryzen 7 5700G was roughly a 70% increase in performance. Despite that feeling like a solid upgrade it is nothing on the practical doubling between Original Pentium/AMD k6 releases in performance year over year.

xattt• 4 years ago

What was the shortest lived socket? Socket 423?

flatiron2• 4 years ago

That would be my guess as well. Only a single p4 generation iirc

rsynnott• 4 years ago

Socket 4 or 5, probably (first two Pentium sockets). Socket 5 had a new socket replace it faster than 4 did, but new chips were made for 5 for longer.

paulmd14• 4 years ago

Probably depends on your definition of "shortest lived". There's quite a few sockets that only got one generation.

AM1 comes to mind, there were only ever about 5 processors compatible with it and really only two were ever intended for retail market and only one of them ever really existed off the drawing board. That's probably one of the rarest consumer sockets ever produced, at least recently.

There have also been a few HEDT sockets that saw short life and small numbers. The W-3175X platform and AMD Quad FX (Quadfather) both were exceptionally short lived in themselves, but on paper supported a decent number of server processors due to socket compatibility. They are probably some of the smallest numbers sold. Quadfather never lived past a single board and neither of them probably sold more than a few thousand units.

TR4 and TRX40 both went down as exceptionally short-lived HEDT sockets with only one generation each and no cross-compatibility with server chips that share their sockets. But they probably sold higher numbers than W-3175X and Quadfather.

WRX80 is probably one of the lowest-volume sockets around since it's basically OEM only and a niche of the niche HEDT market, but again then you've got two generations, even if the volume sold is small.

Most Intel products really aren't contenders here, "one generation for the socket" is table-stakes here, it's volume that really decides it imo. Nobody can really say Intel doesn't produce volume for their stuff, a short-lived socket for Intel probably sold 100x the amount that Quadfather sold. Weird niche products like W-3175X or Kaby Lake-X are the exception but again then you've got other products in the same socket.

fomine33• 4 years ago

Not shortest but weird one was LGA1356.

tehbeard• 4 years ago

People who bought into the AM4 ecosystem 5 years ago should be able to update to these brand new zen3 cpus.

As long as people recognise thay a brand new CPU in a 5 year old board will have some compromises...

Which given the rates of #internetdrama these days, is probably not likely.

kemotep5• 4 years ago

It reduces potential e-waste but depending on the board there certainly will be issues getting the best performance out of the newer zen 3 cpus. No pcie 4.0 support being a big one.

jotm2• 4 years ago

Only thing I can think of is RAM/PCIe speed... which isn't much of an issue. What else is there?

awill5• 4 years ago

I get it's good, but what took them so long? Wouldn't many people who had 300 series motherboards already have upgraded? They were stuck for 18 months without a path forward.

blihp3• 4 years ago

They're trying make switching to Intel less appealing when upgrade time comes. Until 12th gen, they didn't seem to view that as much of a threat. Now it apparently is. Competition is good!

belval4• 4 years ago

I bought a 1700 in 2017 when they launched, upgraded to a 3800X 2 years ago and now I might even get to run a 5800X on my B350 board that I paid CA$130 for.

This is what good competition looks like. As a customer I love it!

mikepurvis• 4 years ago

Seriously, yeah— this is great for consumers. Though I wonder if the mobo manufacturers feel quite as thrilled about it.

delusional• 4 years ago

This isn't just good competition. It's damn good engineering. We have to remember that this wasn't something that AMD was forced to do by the competitive landscape. Someone pushed hard for this decision internally, and that person or team deserve recognition. Someone also had to actually build it, and they deserve even more recognition.

Competition is fantastic, but the people who rise to the competition are, in my opinion, the real heroes.

Tomte• 4 years ago

So please help me understand: I have an MSI X370 mainboard, and as far as I understood until now, the Ryzen 7 3700X processor for around 300 Euros is the newest one I can get that still runs on my system (that would be relevant for being able to update to Windows 11, since I have a Ryzen 1700 now).

Now I can use the 5800X3D, but nothing in between those, or is there some other wrinkle I haven't understood?

neogodless4• 4 years ago

https://arstechnica.com/gadgets/2022/03/amd-reverses-course-...

Basically all Ryzen 5000 series chips can become compatible if your motherboard OEM releases an updated BIOS.

zekica3• 4 years ago

What do you mean nothing in between. They should support ALL Ryzen desktop CPUs.

jotm2• 4 years ago

Yeah, people who didn't get a board with the best VRMs must be kicking themselves :D

Really though, there's some cheap mATX boards with high quality VRMs that can run a Ryzen 9 with no issues. Though tbf, if the processor starts but is not stable, an undervolt or maybe clock cap could work, at the cost of a slight performance drop.

It's nice to have the option to just upgrade the processor.

phkahler2• 4 years ago

> The biggest part of this announcement is the bios updates for 300 series motherboards...

A great upgrade for the old Mellori-ITX running a 2400G. https://github.com/phkahler/mellori_ITX

BTW my system is limited to 4K30hz on the HDMI for reasons I never figured out. Haven't found a good Displayport->HDMI adapter to fix it, and not sure why the CPU/mobo combination won't do it on HDMI. I don't do gaming on it so not much of a problem, but I'd prefer 60hz.

matja2• 4 years ago

Is it because the mainboard is not HDMI 2.0 or does not support DSC (Display Stream Compression?) (https://en.wikipedia.org/wiki/HDMI#Refresh_frequency_limits_...). Raven Ridge supports 2.0, but it's down to the mainboard to also support that.

cehrlich2• 4 years ago

Socket A made it from the first Athlon all the way to right before AMD went 64bit, but there were several different chipsets IIRC

freeAgent• 4 years ago

The first Athlons used Slot A. Socket A came later.

rocky1138• 4 years ago

This has been my biggest pain point with Intel. Why are sockets changed, seemingly, every generation? There's no point in having a socket in that case. Might as well solder the CPU right to the board.

eric__cartman• 4 years ago

Money!

cogman10• 4 years ago

Part of the answer there is it's easier from an engineering standpoint to get a new socket than it is to make a new architecture work with an existing socket.

Sort of the same way in the software world where it's easier to add a new method then it is to enhance an existing method. New methods have no legacy baggage to worry about. Tweak an old one wrong and you run the risk of breaking a bunch of people.

AMD's commitment to forward compatibility is pretty nice, but definitely more work on their part.

Then again, AMD's legacy is making sockets/slots work for their processors :D (That was the whole point of the original athlon)

Nexxxeh• 4 years ago

Does this mean A300 boards may get official support for the Ryzen 5 4600G? Because I'd love to slap one into my DeskMini A300. With a (now pulled) Beta BIOS, an A300 ran a Ryzen 7 Pro 4750G just fine, and it'd be a big jump from my R3 3200G.

kemotep5• 4 years ago

It sounds like it depends on your board’s manufacture providing it for your system but it looks like AMD is releasing the BIOS firmware to allow x370, b350, and a320 boards to support all the new cpus just announced today.

In some cases your existing cpu might not be supported using the new firmware. So if you update the bios, you might have to swap the processor before you can boot the computer again.

paulmd14• 4 years ago

A300/A320/X300 were always the exception to the Zen3 lockout code. You could run a 5950X on a $30 trash A320 board with a 16MB BIOS and a 3-phase VRM, but not a Crosshair VI Hero.

anyway, there's nothing stopping vendors from releasing support for A300, Asrock simply has abandoned the Deskmini A300 and doesn't want to support it anymore. That one isn't on AMD, Asrock just seems to have abandoned it.

bipson3• 4 years ago

AMD was aiming for that when AM4/Zen was announced, knowing that this was a major pain point for enthusiasts and builders/"upgraders".

This would have been my main reason to build an AMD system, although I was easy to win, I always built AMD systems IIRC... And I ended up postponing it again and again, since I don't game and the reasonable (budget) options were always lagging behind regarding technology/efficiency - most of all, my Phenom (actually a rebranded Athlon with 2/6 cores disabled) was still doing OK. Now there will be AM5... Didn't follow the news, is it supposed to live as long as AM4?

Would it even pay off for AMD to repeat that effort?

leguminous• 4 years ago

The differences between the 5600 and 5500 are interesting. The $199 5600 has 6 cores, 32MB of L3, no GPU, and uses the chiplet packaging. The $159 5500 has 6 cores, 16MB of L3, a disabled/fused off GPU, and is one larger, monolithic die.

I thought the chiplet packaging was supposed to result in lower costs due to better yields on smaller dies? I guess the extra cache takes up more than enough die space to make up for that?

faluzure• 4 years ago

The 5500 is less performant, therefore lower cost. It only exists to sell salvaged dies, and won't likely appear in large quantities (similar to how the 3300 / 3100 were impossible to find).

meragrin_2• 4 years ago

None of their chiplet designs currently have GPU. All of their APUs are currently monolithic. The 5500 is just a neutered 5600g. Probably mainly trying make some money on dies with bad GPUs.

Symmetry• 4 years ago

The yield issue is going to be much more of a factor when a process node is young and defect rates are high. The big savings is in engineering hours where you can re-use a design across multiple products.

blihp3• 4 years ago

Everything I've been reading has been that 7nm has been having excellent yields and chiplets do bring their own costs. Lisa Su (AMD CEO) indicated in an interview a while back that there is a price floor at which chiplets make sense. Another thing to consider is that most of the chiplets used in the consumer parts could be considered rejects from Epyc. (i.e. they need to make a lot of relative duds to get the small number of 'golden samples' they use in their server chips)

xxs• 4 years ago

5500 is APU based, and it has only PCIe-3... you can consider it a failed laptop sale.

paulmd14• 4 years ago

the 5500 is a monolithic die, it's based on the monolithic APU lineup rather than the chiplet enthusiast lineup.

This also implies other limitations like PCIe 3.0 and half the cache of the equivalent enthusiast lineup, because that's how the laptop chips were designed. The APU lineup has always performed a bit worse than the equivalent desktop lineup as a result.

PragmaticPulp• 4 years ago

Lower clock speeds but more cache. Also interesting that they stopped at the 8-core chips and aren’t adding the cache to the 12 and 16 core parts.

It looks like the extra cache die isn’t good for thermals, so maybe it’s not viable on the 12 and 16-core chips without sacrificing too much clock speed.

Looks to be a strictly gaming-focused CPU play. Not a bad move given the way the enthusiast market works, but it doesn’t do much for a lot of non-gaming workloads.

zozbot2342• 4 years ago

Looks to be a strictly gaming-focused CPU play. Not a bad move given the way the enthusiast market works, but it doesn’t do much for a lot of non-gaming workloads.

How so? 8 cores still covers plenty of light-workstation workloads, where the extra cache should lead to an overall performance improvement (despite the minor drop in top clock speeds). With current software architectures, it's still hard to make full use of 12 to 16 cores.

paulmd14• 4 years ago

you're rather missing the point about the exclusion of 3D V-cache on the higher-end processors. If it was a productivity-focused play then you'd see it included on those as well.

The 5800X is a good processor and it's true that people do productivity on it, nobody disputed that and there's no need to rush to its defense. But you've missed the point about the overall way the upgrade is (or rather, isn't) being rolled out.

marktangotango• 4 years ago

Does that amount of L3 enable some memory hard POWs to be mined profitably?

bick_nyers2• 4 years ago

I'm hoping it's because they wanted everyone to save their money for Zen4 16 cores with 3D cache. The reasons are probably thermal related though

awill5• 4 years ago

Is this a sign zen4 is delayed? I get they need new motherboards and socket, and DDR5 etc.., but launching zen3 stuff now is strange. A lot of us already upgraded.

hajile• 4 years ago

I believe this was always the plan. This is the refresh with the next chips due late this year to early next year.

blihp3• 4 years ago

I don't take it that way. Zen 4 is likely to only release high-end parts at first. Given the timing of these releases, I'd expect Zen 3 to serve as the mid- to low-end of the product range at least through mid- to late-2023 if not longer. Given the way AMD sandbagged these parts this generation, they'll probably hold off as long as possible next generation for low-end Zen 4 too.

mjevans2• 4 years ago

I couldn't get past the scalpers in 2020. That was already a _slightly_ delayed refresh driven by a combination of less than stellar AMD GPUs (Linux, so nVidia's not a good option) and cryptominer hell to that point.

It's going to be 2 years later, and at this point I'd rather jump directly from DDR3 to DDR5.

awill5• 4 years ago

I thought the Radeon 6800 was a pretty good launch. Didn't it hold up with Nvidia in most benchmarks except RT?

sliken7• 4 years ago

Not heard that, but Alder lake has taken back most of the market share the Zen 3 managed to take from Intel. So a response sooner than later will help protect marketshare. In particular they are doubling down on the lower power chips that ADL is having problems competing with. ADL is fast, but power hungry. Thus chips like the 5700x with a 65 watt tdp instead of the previous 8c/16t chip with a 105 watt TDP.

mastax• 4 years ago

I'm debating whether to buy this. I have an 1800X on a B450 board with DDR4 3000 RAM. I use the PC for gaming and compiling, basically, so I should see huge benefits from the extra cache. I could double my CPU performance and breathe new life into this computer for not much money.

However, that computer is getting long in the tooth. The RAM is slow compared to modern stuff. The motherboard is cheap, has poor UEFI, and I zapped the ethernet port with static - destroying it. Do I really want to invest more into it when I could wait 6 months and build an all new PC with DDR5 and Zen 4? It would be a lot more expensive but I have a lot more money now than when I built this computer.

trillic2• 4 years ago

Wait, buy some ECC RAM for the old machine and turn it into a home server.

KarlKemp• 4 years ago

Because a „home server“ a.k.a. a space heater to store extra porn on, is still a server, and everyone knows servers need ECC?

TheBigSalad• 4 years ago

3000 RAM is good enough. If you upgrade the CPU and GPU you'll get effectively the same performance as any shiny new build and can last another 5+ years. You can get back the ethernet port with a PCI card or USB.

That being said, if you can afford it get the shiny new stuff. You deserve it.

_carbyau_• 4 years ago

I like this advice. Both practical (I agree with all points) and yet acknowledges the emotions we feel when buying stuff with a side of financial sense.

lvl1022• 4 years ago

It’s hard to justify buying these when you have M1 Max or Ultra in Mac Studio.

kcb• 4 years ago

Those devices are completely irrelevant to the vast majority considering one of these CPUs.

lvl1022• 4 years ago

For developers? I beg to differ.

nyadesu• 4 years ago

Crazy how people would prefer spending 300 - 450 USD to upgrade a rig they already own, instead of throwing it away and buying a apple device with a new M1 cpu that costs +700 USD

neogodless4• 4 years ago

Not to mention the M1 Max equipped Studio starts at $1999.

And can't run most of the games I've played over the past two years.

That parent comment was particularly out of place since the 5800XD is a gaming-focused CPU.

sliken7• 4 years ago

Er, studio is minimum $2k, well above the average price of a home built AMD system with the same specs (32GB ram and 512GB SSD). It also runs OSX instead of Windows and Linux,

Not really the same market at all. The 5700X (2nd most expensive on a list of 7 CPUs) costs $300, 32GB ram of DDR4-3600 is $150, 512GB SSD is $60, motherboard, case, and power supply is likely another $300. So $810 for a system, not including a GPU, but $1200 is more than enough for something decent.

speed_spread• 4 years ago

It's hard to justify buying an M1 when all they run is OS X.

KronisLV• 4 years ago

Not to discount how cool healthy competition is, but i recently got a Ryzen 5 1600 (used), which came out in 2017 and its 6 cores and 12 threads have been enough for every workload that i've wanted to deal with, even in 2022.

It was a nice upgrade up from Ryzen 3 1200 that i had previously and honestly, i don't see myself upgrading again for the years to come, unless things become visibly slow and sluggish because of Wirth's law (let's assume that in 2030 most of my desktop apps would run Electron instances).

It does gaming, it does programming, it renders videos, it does 3D modelling and other content creation, it compiles code and runs Docker containers and oftentimes multiple of those at the same time with no overclocking and no issues (e.g. render a video in kdenlive while playing a game in the background).

Somehow it feels like either hardware has scaled to the point where new improvements are pretty incremental, or maybe software's ability to be wasteful with the provided resources thankfully still hasn't caught up with these improvements - you no longer really need to get a new CPU every 2 or 3 years, even the Ryzen 3 1200 which also came out in 2017 was adequate (i just got a good deal on the upgrade), which is really nice to see!

Even my homelab servers still run used Athlon 200GE CPUs from 2018 because of the 35W TDP and are on par with cloud resources that i can otherwise afford for my 24/7 available cloud stuff, just much cheaper when you consider how long they last.

Also, there's something really cool about every device in my home (apart from laptops) running on the same CPU socket and all of the parts being interchangeable with no driver update weirdness. Though it'd be even better if the Ryzen CPUs were the variety with iGPUs, given the GPU prices still otherwise being pretty unfortunate (in case my RX 570 would break).

The only reasons that i see for upgrading my setup in the next 5 years would be one of the following:

  - the stocks of used CPUs drying up or prices rising (AliExpress still has them pretty affordable)
  - the motherboards themselves getting deprecated or there being other unforseen issues

Here's hoping that won't happen and that i can enjoy CPUs that are suitable for my needs (and help avoid the e-waste issue at least somewhat), while those more well off financially than me can dabble with the more recent releases to their heart's content!

culopatin3• 4 years ago

For the first time in 2022 I maxed out the 3600x I got when it came out (2018?). And it wasn’t a game, it was my girlfriends R script to do metabarcoding of some bacteria. 100% usage across every core sustained for 15min.

However what shocked me was that my base M1 pro was a few minutes faster. (Same amount of ram)

Still fast enough that it’s not worth an upgrade, but we’ll see once she starts transcriptomix

sliken7• 4 years ago

I do wonder if the days of dimms and long lived sockets of AM4 is over. AM4 maxes out at DDR4-3200 x 128 bit = 50GB/sec and more importantly just 2 pending cache misses. A bit more with overclocking, but not much.

Apple M1 = 128 bit x LPDDR4X-4266 = 68GB/sec and I believe 8 memory channels (8 pending cache misses). A modest core count (4 fast + 4 slow) helps keep the memory bandwidth from being a bottleneck.

The M1 pro doubles this to 256 bits, 16 channels, and 200GB/sec, which is a significant help for the integrated GPU and hits levels that AMD APUs can not match.

M1 max doubles again to 512 bits, 32 channels and 400GB/sec.

M1 ultra doubles again to 1024 bits, 64 channels, and 800GB/sec.

Not sure AMD can really compete and the APUs are severely limited by memory bandwidth, unless you buy a PS5 of XboxX. I'm hoping that AMD takes a page from the Apple playbook and ships a consumer CPU with dramatically more bandwidth and allows users to skip the current curse of discrete GPUs that run hot, are not available at MSRP, and are hard to get even at 2x MSRP.

bick_nyers2• 4 years ago

I think you would find an APU that powerful would suffer the same fate as discrete GPUs in that it would be low-stock and/or expensive. Same thing with the M1 Ultra, starts at $2k it looks like

sliken7• 4 years ago

Dunno. Mining is all about hashrate/$ and scalpers are all about the % profit. Seems like AMD could easily make today's APU with 2x the bandwidth (less than 1/5th of the Apple Studio/M1 max) make a tidy profit and provide 2x the APU performance and make the APUs a good fit for a much larger fraction of the GPU intensive applications.

Not like they aren't shipping by the millions in the XboxX and PS5, there's obviously a demand for them and AMD is obviously capable of making them (they make the CPU in both).

GekkePrutser• 4 years ago

Too little too late. The 5600 should have been out from the start. I've basically wasted 100 bucks on the X version I didn't want.

Only went for AMD so I could reuse the motherboard. But next time I'll go back to Intel.

Koiwai• 4 years ago

Those are lower end chips that were hold back to clear older 3000 series stock, so maybe it's a sign those were finally cleared out.

AMD Fires Back with 7 New Chips

Related Stories