[HN Gopher] AMD Reveals Next-Gen Desktop Processors for Extreme ...
___________________________________________________________________
AMD Reveals Next-Gen Desktop Processors for Extreme PC Gaming
Author : doener
Score : 120 points
Date : 2024-01-10 10:02 UTC (1 days ago)
(HTM) web link (www.amd.com)
(TXT) w3m dump (www.amd.com)
| dralley wrote:
| These aren't Zen 5 CPUs, they're Zen 4 chips with better
| integrated graphics (which gamers likely won't care about anyway)
| and some AI accelerators (which probably won't see wide adoption
| for at least another year)
| kkzz99 wrote:
| Do these AI accelerators even support common implementations
| for LLM inference or stable diffusion models?
|
| Last time I checked, AMD's support was terrible.
| diggan wrote:
| I'm guessing no, as these newly announced processors are the
| first with those accelerators.
|
| > AMD is also bringing the power of a dedicated AI neural
| processing unit (NPU) to desktop PC processors for the first
| time with the introduction of Ryzen(tm) AI
| wtallis wrote:
| They're merely the first _desktop_ processors from AMD with
| these NPUs. This is last year 's laptop silicon repackaged
| for their desktop socket. They also re-branded the laptop
| version with new model numbers, because laptop OEMs
| basically demand new model numbers on an annual cadence.
| This year their marketing for these chips is heavily
| emphasizing the NPUs, because they actually have some
| tooling for them now, but they were basically dead weight
| when the silicon first shipped.
| zamalek wrote:
| AMD's support is terrible, if you believe their
| documentation. ROCm/hip works just fine (well, with Torch -
| Tensorflow is shamelessly an NVIDIA shill) on many
| "unsupported" GPUs if you enable an override envar.
| NekkoDroid wrote:
| To add a bit more context: I remember reading somewhere
| (may have been in the Phoronix forum) from an official AMD
| engeneer that "supported" means validated and actively
| tested in their pipeline, while "unsupported" generally
| just means "we don't do any or minimal testing on these,
| they should work and we don't explicitly prevent them from
| working but we don't guarantee anything" (at least when it
| comes to same die/gen cards).
|
| In the same post they also wrote that they are gonna look
| at integrating more of those "unsupported" cards into their
| suite.
|
| Honestly, I hope they change their wording for this to
| something like "validated", "supported" and "unsupported"
| with actual explenations what each of these mean (fully
| tested, works theoretically, does not work even
| theoretically)
|
| Edit: I actually found the post I was talking about
| https://www.phoronix.com/forums/forum/linux-graphics-x-
| org-d...
| makomk wrote:
| ROCm is also incredibly fragile and buggy at the best of
| times, so anything not actively tested by them stands a
| good chance of not working. Hell, I remember that a while
| back people were having problems with machine learning
| code giving garbage results on one of the few consumer
| GPUs that was officially listed as supported and AMD
| eventually replying to the bug report and declaring that
| actually, no they weren't going to support it and they'd
| remove it from the list ratther than try and fix the
| issue. I think this was back when newer consumer GPUs
| were genuinely unsupported as in the code simply wasn't
| there too. Integrated GPUs have also alwayu had a lot of
| problems.
|
| There's also questionable OS compatiblity. ROCm is Linux-
| based and has extremely limited and rather experimental
| Windows support. Their fancy new neural processing unit
| is Windows-only, tied in with a Microsoft framework, and
| they don't seem to have any kind of definite plan for
| supporting it elsewhere. So there's quite possibly no
| single OS where all the hardware that theoretically could
| be used for machine learning and AI in these chips
| actually has even vaguely functioning code that works on
| that OS to make it work that way.
| wmf wrote:
| Does ROCm even work on XDNA? I don't think it does?
| jiggawatts wrote:
| Tensorflow is open source on GitHub. What's stopping AMD
| engineers from contributing with Pull Requests? A casual
| search through the open PRs shows nothing of interest being
| submitted by team red.
| antx wrote:
| ah yes, the good old HSA_OVERRIDE_GFX_VERSION=10.3.0
| switcheroo
| caycep wrote:
| as long as PyTorch supports it, does it matter?
| antx wrote:
| for as long as I'll need to recompile PyTorch for it to
| adequately support my card, yes, it matters.
| coffeebeqn wrote:
| Extreme integrated GPUs? Marketing department going wild
| asmor wrote:
| They are enough to run the Ally, and they're closely related
| to the one in the Steam Deck. So maybe they're extreme in the
| segment of integrated graphics??
| bonton89 wrote:
| They're just trying to compete with Intel Extreme Graphics
| from 2002. Must be they're reaching performance parity.
|
| https://en.wikipedia.org/wiki/List_of_Intel_graphics_process.
| ..
| seabrookmx wrote:
| Parity? AMD integrated GPU's have outperformed Intel's for
| many years.
| dathinab wrote:
| only APU like the 5700G, 8700G or the recent Steam Deck
| APU (in reasonable comparison targets)
|
| The integrated GPUs of the 7000 serious (not APUs) are
| maximal minimalist to a point I think they couldn't have
| shrunken them more without running into unusual issues
| (assuming not changing the architecture/design). They are
| only suited for web browsing, office use-cases and
| debugging (but also kinda where made only for that use-
| case). Well I guess Dead Cells and similar "low
| requirement/high optimized" games still run nicely on it
| on 1080p.
|
| Still nice to have them if you only have such use-cases
| saves a ton of money.
| ahoka wrote:
| They did not say which one of extremes.
| Dalewyn wrote:
| >they're Zen 4 chips with better integrated graphics (which
| gamers likely won't care about anyway)
|
| Integrated GPUs are becoming far more appealing to the general
| consumer, including (especially?) gamers, with how fucking
| expensive discrete GPUs are getting these days (AMD is part of
| that problem).
|
| We might just witness dGPUs becoming the sound cards of this
| decade.
| mirsadm wrote:
| Extremely unlikely. The types of GPUs these compete against
| don't cost a lot of money.
| Dalewyn wrote:
| An RTX 4060 (with a pitiful 8GB vram) is about $300.
| Something a bit more practical like an RTX 4070 (with 12GB
| vram) is about $550.
|
| That is bullshit expensive. It's still nowhere near as bad
| as during the cryptomining craze, but remember a GTX 1080
| MSRP'd for about $430 back in the old days. An RTX 4080 for
| context is around $1,500 to $2,000 right now.
|
| If I'm just looking to game and only have a reasonable
| budget like most people, I'll just grab some AMD APU and be
| done with it. Better yet just go and buy a console or game
| on my phone.
| BenjiWiebe wrote:
| How do the best integrated graphics compare performance-
| wise though? If they are way behind in performance, you
| might be able to buy a cheap 2GB VRAM GPU for less than a
| cutting edge integrated graphics CPU.
| justsomehnguy wrote:
| > you might be able to buy a cheap 2GB VRAM GPU
|
| Lol?
|
| It's another ~$100 what would not yield you any
| performance gains.
|
| Older 5700G was ~$300 so it was hardly a part of a budget
| build, but 5 5500GT is _$125_ (and would probably drop
| below $100 in a half /year) and would clearly offer at
| least 30fps on FullHD[0] and you _would buy it anyway
| because it 's a CPU_.
|
| https://www.tomshardware.com/reviews/amd-
| ryzen-5-5600g-revie...
| Dalewyn wrote:
| >and you would buy it anyway because it's a CPU.
|
| This is the part that needs to be emphasized more, that
| iGPU is practically _free_ because it comes with the CPU
| we 're buying anyway.
|
| It's already hard enough to compete with _free_ , but on
| top of that dGPUs are asking for Keksimus Maximus monies.
| The value proposition for gamers, let alone most consumer
| users, is very strongly favored towards iGPUs.
| sofixa wrote:
| > How do the best integrated graphics compare
| performance-wise though
|
| A bunch of handhelds (Steam Deck , Asus ROG Ally) that
| can comfortably run even AAA games with 30fps+ on small
| screens (as an example Red Dead Redemption 2 runs with
| 50fps on my LCD Deck) with battery and thermal
| limitations would imply that desktop versions would be
| totally acceptable, unless you're looking at 4K on the
| latest big titles.
| BirAdam wrote:
| The best integrated graphics are about equivalent to a
| 1650. This assumes that the machine also has decently
| speedy RAM.
| whatwhaaaaat wrote:
| Your last sentence clearly shows you are not a pc gamer
| which is fine but you cannot use a persons opinion who is
| not a pc gamer about .. pc gamers.
| bryanlarsen wrote:
| Lots of PC gamers are very happy with "slightly better
| than Steam Deck" levels of performance.
| Arrath wrote:
| Yo! I tolerate relatively poor performance out of my
| laptop (a several year old model with a Ryzen and iGPU)
| while having a fairly kickass home desktop[0]. So chalk
| me in that group.
|
| [0]Which currently sees me playing a whole bunch of Rule
| the Waves 3..... poor underutilized 3080.
| Dalewyn wrote:
| Don't worry, the 3080 Ti in my Kickass Desktop(tm) also
| spends most of its gaming time playing _Princess Connect!
| Re:Dive_ and _Uma Musume_. :V
| wiseowise wrote:
| PC gamer as a Reddit stereotypical le gaming masterrace?
| Because most gamers I know don't give a shit about this
| stuff. They care about games.
| wmf wrote:
| The 4060 is really bad though. I would guess that a
| 12100F + A750 or 6600 XT would be faster than the 8700G
| with its "extreme" graphics.
| zokier wrote:
| More appropriate comparison point would be something like
| Arc A580 which retails around $180 and is already
| dramatically faster than this igpu.
| toast0 wrote:
| If you want to buy new (which I don't think is
| unreasonable), there's not much in the market that doesn't
| cost a lot of money.
|
| Maybe a Radeon 6500XT or a Arc A380; Geforce 1650 are still
| kind of a lot of money, and you're looking at at Geforce
| 1030 if you want something for not a lot of money. But why
| buy a Geforce 1030, unless you really just need something
| low profile.
| 0x457 wrote:
| Well, look at the steam hardware survey, you're going to be
| surprised.
| sva_ wrote:
| People playing on laptops and the steamdeck? All legit uses
| for the iGPU (I have one myself)
|
| https://store.steampowered.com/hwsurvey/videocard/
| dathinab wrote:
| I think they meant people playing on really old dedicated
| GPU which likely are slower then the 8700G to a point where
| a 8700G might be a good replacement if their setup breaks
| or similar.
| wiseowise wrote:
| That's obviously not-legit. /s
| shmerl wrote:
| Steam Deck users would disagree about integrated graphics.
|
| Besides, APUs really got good for gaming unless you push
| resolution high.
| sva_ wrote:
| I think they were hinting at the fact that the iGPU is more
| interesting to users of the mobile CPU (Steamdeck/Laptops),
| since this is a desktop CPU, which usually have dedicated
| graphics.
| jabroni_salad wrote:
| I used to play WoW on a Llano APU. Plenty of gamers are
| addicted to something like that and just want a serviceable
| cheap machine. Pretty sure my current 3070 cost more than the
| entire computer I was killing Deathwing with.
| Arrath wrote:
| I played healer so I could just point my camera at the floor
| zoomed in as far as possible and suffer the least bad
| performance possible. And scramble to get back in-game after
| guaranteed disconnects on big events e.g. every time Nefarion
| spawned a wave of ads. What a struggle.
| dathinab wrote:
| The 5700G had for some time been one of the best options for
| budged gaming.
|
| The 8700G should be seen as a successor of that I think.
|
| Enough to play a huge amount of games nicely under 1080p.
|
| But in a certain way at the lowest end of desktop gaming.
|
| Still better then the low-mid range of gaming on a laptop, tho.
|
| But for a lot of people that is exactly the right balance of
| what they get to what they pay. Weather that's because they
| don't have a lot money or they don't do a lot gaming and don't
| want to wast money.
|
| The main "problem" the 8700G might run into is I guess that the
| cost of AM5 motherboards and RAM.
|
| Also AMD (and I think Intel, too) have in recent years slowly
| worked to take advantage of integrated graphics even if you
| have also dedicated graphics (of the same vendor) if that sees
| some more improvements the 8700G could also become interesting
| in some use-case you wouldn't expect today. We will see.
| NohatCoder wrote:
| A quick check puts a Ryzen 5 5600 and a Radeon RX 6600 at
| $360 combined, where the 8700G is set to $330. And a RX 6600
| will deliver around double the graphics performance. So even
| without factoring in motherboard and memory the 8700G is hard
| to justify for a cheap gaming rig.
| chx wrote:
| Source:
|
| https://www.newegg.com/amd-
| ryzen-5-5600-ryzen-5-5000-series/... $150
|
| https://www.newegg.com/p/pl?N=100007709%20601394871&Order=1
| starts at $210 indeed.
|
| People might argue the 5600 is Zen 3 where the 8700G is Zen
| 4.
| kube-system wrote:
| > New AMD Ryzen(tm) 5000 Series Desktop Processors Bring More
| Performance to Legacy Socket AM4 Platforms
|
| Wait, what? New chips for the old platform? I wonder what the
| driving factor even is to compel them to produce these?
| diggan wrote:
| > Wait, what? New chips for the old platform? I wonder what the
| driving factor even is to compel them to produce these?
|
| A lot of people are already using AM4 socket and might not want
| to do a socket upgrade, but want a new CPU.
|
| Also, I seem to remember AMD made a commitment before/at AM4
| launch that they would support it for at least N years, but
| don't remember the details about that. Maybe that could be
| related?
| borissk wrote:
| Nope, AMD promised to support AM4 socket until 2020.
|
| AM4 came out in 2016, supporting it with new CPUs for over 7
| years is unheard of in the PC industry.
| zamalek wrote:
| And it earns them a lot of good will, and sales.
| MPSimmons wrote:
| I might buy one for one of my old desktops that's still
| kicking around as a server. 16 cores isn't something to
| shake a stick at!
| FirmwareBurner wrote:
| _> 7 years is unheard of in the PC industry_
|
| LOL. Intel's LGA-775 which span support for CPUs from
| single core Pentium 4, to dual core Core-2-Duo, to quad
| core Core-2-Quad would like to have a word with you.
|
| And 7 years back in those days was equivalent to an
| etternity in today's tech progress.
|
| https://en.wikipedia.org/wiki/LGA_775
| borissk wrote:
| Kind of - early socket 775 motherboards don't support
| later Core CPUs.
| universa1 wrote:
| Not sure if the Mainboard Chipsets initially released
| with the Pentium iv supported the later released core2
| quads... Which is actually the case for the am4, where
| even the cheapest, oldest Chipset (a320) can support the
| newest am4 CPU, as long as the MB maker provides a bios
| update.
| FirmwareBurner wrote:
| Sure, but the technological leaps from Pentium 4 to Core
| 2 Duo and then to Quad that Intel made in those days over
| 15 years ago, were massive enough to justify the
| limitation of the same chipset from the Pentium 4 era not
| supporting the later multi-core Core CPUs, compared to
| the Ryzen 2-3-4 jumps that AMD made in a similar
| timeframe which aren't as radically different to each
| other.
| Arrath wrote:
| And since then has an Intel socket supported more than
| one revolution of a tick-tock CPU generation?
|
| Genuine question because I've now built several desktops
| in the Intel 'Core iX' era and have never truly had on
| opportunity to reuse a mobo.
| FirmwareBurner wrote:
| Of course not, but I was only pointing out that 7 years
| of socket support is not unheard of as grandparent
| claimed.
| noobface wrote:
| Motherboard prices have increased pretty significantly. First
| comment on this hardforum post a entitled: AMD and Intel
| motherboard prices skyrocket past surging inflation rates
| thanks to 35-40% ASP increases[1]:
|
| "Motherboard prices is why I haven't upgraded my CPU. More
| important things going on right now."
|
| [1]https://hardforum.com/threads/amd-and-intel-motherboard-
| pric...
| treprinum wrote:
| AM5 mobos are more expensive than Threadripper x399 mobos
| when they were new...
| kvemkon wrote:
| Hope AM5 is the last such an obsolete 2-channel memory
| platform. Was it 2004-2005 as we got 2-channel memory with
| single or first dual-core CPUs? Now we have 16 cores
| with... still 2-channel memory. Looking forward to
| 4-channel AM6 socket.
|
| Memory bound software like OpenFOAM saturates memory
| bandwidth already on 6-8 cores with 2-channel memory [1].
| So actually we should have got 4-channel with first
| 16-cores CPUs long ago.
|
| [1] https://www.cfd-
| online.com/Forums/hardware/198378-openfoam-b...
| wtallis wrote:
| Same chips, different binning. Zero design effort and almost
| zero QA effort went into those new products, and it lets them
| fine tune their margins.
| dotnet00 wrote:
| AM5 is a more expensive upgrade, not only are the motherboards
| more expensive, they also take DDR5 and the overall performance
| improvement is not so meaningful. So, for many people who
| already have an AM4 PC, AM5 is not yet a justifiable upgrade.
|
| There may also be some hesitance since while AMD technically
| did support AM4 for longer than the promised duration, they
| tried to pull out of it halfway through and overall it was kind
| of a mess for a while in terms of compatibility with chipset
| revision.
| phkahler wrote:
| >> New chips for the old platform? I wonder what the driving
| factor even is to compel them to produce these?
|
| I might upgrade my 2400G, but I'd want 8 cores in 65W which is
| sadly not an option.
| toast0 wrote:
| > I might upgrade my 2400G, but I'd want 8 cores in 65W which
| is sadly not an option.
|
| 5700G is 8 cores, with graphics, 65w tdp, released in 2021.
| This release also includes the 5700 (no letters) with no
| graphics at 65w tdp. Looks like 5700X is also 65w tdp? Also,
| you can usually set a 65w power limit on a cpu with a higher
| tdp and get most of the performance.
| NohatCoder wrote:
| They are produced on an older processing node than the AM5
| lineup. So if AMD switched to only produce AM5 CPUs and Radeon
| 7000 graphics chips, the old factories would lack something
| worthwhile to produce. Therefore AMD can negotiate much lower
| rates for using those older facilities, making it viable to
| continue production as long as the products sell. There aren't
| any truly new chips, it is just new binning and marketing of
| the old models.
| flumpcakes wrote:
| AMD historically supported their sockets a lot longer than
| Intel's two years. Recently they moved to AM5 and only had a
| single generation on a threadripper socket. This greatly
| angered tech media and consumers, despite it still being better
| than Intel support.
|
| Perhaps releasing new SKUs on the "old" platform is a way to
| bump up the numbers even more: "see we supported AM4 for 5+
| generations!"
| bloopernova wrote:
| Having just recently built a Ryzen 7 7800X3D CPU based desktop, I
| was worried I had bought just before a new line came out, but
| from my reading of the press release, none of these CPUs are
| considered a replacement for that chip. I think?
| dawnerd wrote:
| Literally did the same last month but heard rumblings about
| what was coming and felt much better. The 7800x3d is really
| awesome though.
| slowmovintarget wrote:
| Same with the Ryzen 9 7900X3D. I got a System76 machine, and
| it is really good. Nvidia card for GPU / local inference.
| asmor wrote:
| The 7900X3D is by far the worst value AMD currently offers.
| It only 6 cores per CCX, but two of them, so not only do
| you get the scheduling issues of the 7950X3D where
| sometimes games don't use the X3D CCX, you also only get 6
| cores if it works. And you get a "7600X3D" (not a real
| thing) if you disable the other CCX, something 7950X3D
| owners would sometimes do to benchmark the scheduling
| difference to get what's essentially a 7800X3D.
| theogravity wrote:
| This is the reason why I went with the 7950X and not the
| 3D variant. Too much additional work involved to get the
| other cores to play well.
|
| Also I read that only one set of 6 cores as full access
| to cache while the other doesn't or has partial access.
| asmor wrote:
| The cache is on a single CCX, yes. Which is why the
| 7800X3D is so good: There is no second CCX.
|
| Another big advantage is that to get them to work, they
| need to use higher bin chips. CCX overheats easily
| because the stacked cache acts like a heat shield, so you
| get the most power efficient ones.
| kvemkon wrote:
| The disadvantage of single CCD is only half of available
| write memory bandwidth (using at least 1 computing thread
| on each CCD) [1].
|
| [1] 16 B/cycle write vs 32 B/cycle read per CCD
| https://www.servethehome.com/amd-ryzen-7000-series-
| platform-...
| asmor wrote:
| I'd wager the number of workloads that'd actually take
| advantage of operating on entirely different data on a
| per-thread basis is much, much lower than ones that
| benefit from massive L3 cache.
|
| And note that the infinity fabric and memory controller
| don't run on the same clock. The fabric tops out at ~2200
| (though testing shows sweet spot at 2033 usually) and the
| memory controller usually tops out at ~3100 (DDR5-6200).
| kvemkon wrote:
| Those might be not necessary 2 threads of one program,
| but 2 independent programs.
| slowmovintarget wrote:
| The interesting thing was that in testing it turned out
| that having 3D cache for all cores didn't increase
| performance. The sweet spot for heat/performance was for
| half.
|
| Granted, I don't think I'm running anything so CPU
| intensive that I'll be able to tell. For gaming, I was
| more interested in the DDR5 memory than in worrying about
| the L3 cache. For everything else I do (including running
| LLMs and compilers) it's beyond adequate.
|
| For a mere $20K I could have had an overall slower gaming
| experience but be able to locally run things like the
| Goliath 120B model, but that seemed like a poor trade-
| off.
| asmor wrote:
| No, the "extreme gaming" line in the press release must have
| been chosen by someone that have no relation to AMD products
| and were just told "it does gaming without a GPU pretty good".
| Which is true, it does do that.
|
| These are less important CPUs in this lifecycle though, usually
| the G series are pretty good for a small computer where you
| don't want a dedicated tiny GPU, but all but one Ryzen 7000
| desktop CPUs had integrated graphics (though, much fewer CUs).
| dragontamer wrote:
| AMD is fleshing out the $100 to $250 market with this
| announcement.
|
| Top end $300+ chips (like 7800x3d) are safe.
|
| ---------
|
| I'm seriously considering the $250 5700x3d for my nieces, or
| maybe something cheaper. They don't need the latest-and-
| greatest PC. But having something "pretty good" (and the
| 5700x3d will be pretty good) is surely going to be appreciated.
| PartiallyTyped wrote:
| It's a bit funny how the 5800x3d was creeping in the 7xxx
| series during launch.
| piinbinary wrote:
| The press release has a strange title for new chips targeted at
| lower-end PCs (those without a dedicated GPU).
| phkahler wrote:
| >> The press release has a strange title for new chips targeted
| at lower-end PCs
|
| It's a pretty high-end CPU though, and the graphics support AV1
| encode/decode. This would be perfect for software development,
| video production, or pretty much anything but high-end gaming.
| piinbinary wrote:
| True, I should have said "low-end gaming PCs"
| KennyBlanken wrote:
| By the looks of how deep they had to dig to get frame rates
| to show off, this chip struggles at almost any 3D game that's
| been made in the last few years. It's equivalent to a six
| year old mid-tier discreet GPU.
| trynumber9 wrote:
| > the fastest integrated desktop processor graphics in the world
|
| I'm pretty sure the Mac Pro is a desktop. I'm pretty sure the M2
| Ultra configuration includes an integrated graphics processor.
| And I'm pretty sure it's faster than the 780M in almost every
| workload.
|
| AMD marketing team needs to be a bit less wrong. At least add a
| "PC" caveat.
| MacNCheese23 wrote:
| Right, the topic "Extreme PC Gaming" - PC - PC - PC - yeah just
| for you spelled again PC - is extremely misleading - next time
| they should include "Not Overpriced" that people like you know
| their beloved Hardware isnt meant.
| trynumber9 wrote:
| It's in certain spots but not the text where they claim that
| nor the footnote. And in any case only the best by the
| technicality of Apple selling desktop personal computers that
| aren't IBM PC lineage.
| ClassyJacket wrote:
| You know Mac is a specific brand of PC, right?
| fourteenfour wrote:
| Yeah, this "article" is terrible. "users can expect immense
| power and dominant performance for intensive workloads
| including gaming and content creation."
| shmerl wrote:
| Kind of a misleading title. It's not next gen processors (you'd
| expect Zen 5). It's still Zen 4.
| brucethemoose2 wrote:
| Very misleading.
|
| As I have said before, HN needs a "misleading title" flag.
| flumpcakes wrote:
| Is it actually misleading? AMD never makes claim that their
| consumer branding relates to uArch. AMD still has Zen 2 in
| their 'modern' lineups. Intel also has previous generation
| uArch in current SKUs, and let's not talk about the 10->11 and
| 12->13 gen "refreshes".
|
| Maybe the technically inclined of us would expect Zen 5 because
| we're keeping note of CPU uArches but to the general public
| they won't care. They will just see it's the biggest numbered
| AMD chip and expect it to be their best offering - which for
| integrated graphics on the desktop it is.
| GeekyBear wrote:
| If you consider "Extreme PC Gaming" to be AAA games at 1080p on
| low settings, sure.
|
| https://www.anandtech.com/show/21208/amd-unveils-ryzen-8000g...
| declan_roberts wrote:
| For a desktop without an additional $200-$500 graphics card
| this is impressive.
| brucethemoose2 wrote:
| Not really, since the APUs themselves are historically quite
| expensive.
| Moto7451 wrote:
| That hasn't been true for the Zen based models at least.
| The 5700G and now the 8700G occupy the $350-370 price slot.
| My comparable 13th gen Intel CPU with lesser graphics was
| the same price. That's a great value for a casual gamer.
|
| I'm sure at some point we'll see Intel based APU
| equivalents on desktop with some teeth. Intel needs to
| solve their driver issues first.
| brucethemoose2 wrote:
| > That's a great value for a casual gamer.
|
| It depends. If you're truly performance insensitive and
| _have_ to buy new, maybe, but these APUs are extremely
| slow compared to even older, low end discrete GPUs.
|
| AM5 is an expensive platform. The total cost of a DDR4
| platform is peanuts in comparison.
| toast0 wrote:
| The cpus are currently significantly more expensive, yes,
| but is the AM5 platform that much more expensive than
| AM4?
|
| On AM4, pcpartpicker says I can get an A520M board for
| $70, and 2x16GB DDR4-3600 CL18 for $55; total $125
|
| On AM5, an A620M board is $75, and 2x16GB DDR5-5600
| starts at $76; total $151.
|
| The real question is where the Microcenter bundle pricing
| will end up, if you live near a Microcenter.
| brucethemoose2 wrote:
| Mmm, A620 has come down considerably since I last
| checked. That's good.
| CivBase wrote:
| That price range might have been accurate 7 years ago, but
| modern graphics cards for gaming range from $300 to $1000
| with the highest level pushing $2000 thanks to the AI craze.
| The higher prices make quality iGPUs all the more relevent -
| especially since the "gamer" demographic continues growing
| and many of the most popular games are not particularly
| demanding.
| brucethemoose2 wrote:
| Platforms, especially AM5, are very expensive too. Ryzen
| 8000 is far slower than even the slowest new discrete GPU.
|
| The real value option is buying an older platform (AM4) and
| a used discrete GPU, from back when prices were more sane.
| snvzz wrote:
| ... if electricity was free.
|
| Power efficiency wise, a discrete GPU won't compare
| favorably.
| GeekyBear wrote:
| A used graphics card good enough to play 1080p games on low
| settings would not be very expensive at all.
| ozarker wrote:
| You can find used 1080ti's for like $100 nowadays. I still
| use one for 1440p gaming and get excellent frames with most
| games. Kinda mind blowing value
| readyplayernull wrote:
| Aren't these the remnants of the Bitcoin gold-rush age?
| currymj wrote:
| this GPU was so good at the time that it started
| cannibalizing sales of higher end data center cards, and
| Nvidia had to put terms in the EULA for the drivers to
| try to legally prevent this.
| thot_experiment wrote:
| 100% 1080Ti is insane value, I have two of them and zero
| complaints, the 11 gigs of vram make it great for
| inference as well.
| blibble wrote:
| it's average frame rate too
|
| 99%ile (or even minimum) would be far more interesting
| NavinF wrote:
| I'm pretty sure this product line targets people who can't
| afford a new GPU and aren't aware of the used hardware
| market. 60fps average and arbitrarily low 1%-low fps is
| totally acceptable for this segment.
|
| Recall that a lot of games ran at 30fps in the PS3 era. Back
| then people would unironically say "the human eye can't see
| 60fps". Even today a lotta gamers have never experienced low
| latency
| brucethemoose2 wrote:
| Precisely. This is for when used hardware is not an option.
| blibble wrote:
| I'd agree if AMD's title didn't explicitly say "Extreme PC
| Gaming"
|
| Extreme PC Gaming in 2024 is not 60fps at 1080p, it wasn't
| even in 2018
| KennyBlanken wrote:
| It's six year old mid-tier performance.
|
| AMD marketing had to dig pretty deep to find decent
| numbers...4-5 year old games. Metro Exodus? Shadow of the Tomb
| Raider? GTA 5? And some random kid's game nobody has ever heard
| of?
|
| Given the choices are between "buggy as hell" (Intel) "space
| heater and slightly buggy" (AMD discreet) and "good but
| stupidly overpriced" (Nvidia), a fourth option, for the low end
| of the market, is welcome.
|
| Edit: since I'm being downvoted for claiming the games listed
| aren't relevant: go look at steamcharts. GTA5 and Dota are the
| only top 25 games; Cyberpunk is #26. The rest aren't even top
| 100 games.
| chmod775 wrote:
| That game selection isn't an accident. They're all extremely
| popular and/or have a benchmark built-in.
|
| Look at any third-party review and you'll find a very similar
| selection.
| KennyBlanken wrote:
| With the exception of GTA5 and DOTA2, none of the games
| they list are even in the top 50 on Steamcharts.
|
| "and/or have a benchmark built in" doing a whole lot of
| lifting...
| drzaiusapelord wrote:
| Kid's game? Tiny Tina/Borderlands is a huge franchise in
| gaming. I think its clear these two games were chosen because
| the cell-shaded style they use is less GPU demanding, but its
| still impressive. But note things like Cyberpunk 2077 is
| there too. These are all popular games. I don't think its
| this dishonest ploy you're making out to be.
|
| My 2070 barely handles those games at that fps.
|
| Yes those are older games because this APU is not going to
| play modern AAA at 4k, but it can handle some pretty hefty
| games fairly well and might be tempting to budget gamers
| especially when mid-tier cards start at $500-600 nowadays.
| KennyBlanken wrote:
| There are 2,000 people playing Tiny Tina on Steam.
|
| There are _a million_ people playing CS2. Fortnite sees
| about 2.6 million and peaks at _eleven million_.
|
| The Finals has 70,000+ people playing.
|
| Tiny Tina has about 1/5th of the player count it would need
| to be in the top 100. So yes, AMD marketing was pretty
| fucking desperate when they listed that game.
| wiseowise wrote:
| > There are 2,000 people playing Tiny Tina on Steam.
|
| The game is cross-platform and available on PS4/PS5, Xbox
| X/S, Xbox One, Steam and Epic Store. And on top of that
| it is a paid game. I'm not even aware what this game is,
| but I'm aware of Borderlands franchise and they're
| quality games.
|
| Why would you compare player numbers to top f2p games is
| beyond me.
| fireflash38 wrote:
| Crysis was used in benchmarks for how long exactly? And it
| wasnt like it was a blockbuster either...
|
| Maybe you should make an actual performance argument instead
| of a popularity contest.
| drzaiusapelord wrote:
| 63 fps on Cyberpunk 2077 which when it came out was "unplayable
| but on the most powerful PCs" is incredibly impressive without
| a GPU.
|
| This is pretty close to my 2070 GPU does, which cost me $400+ a
| couple years ago and uses 215W. My CPU also uses 100W, so about
| 300W compared to 65W for very roughly similar performance (in
| some games) is still pretty incredible.
|
| Now GPUs are almost twice that for that xx70's and xx80's
| cards. I don't know what market this is aimed at, but this is
| very impressive for an APU. There's a pretty strong budget PC
| gamer community that could benefit from this. There are a lot
| of people who can't afford gaming PCs anymore and this could be
| a big seller to the budget community. Also at 65TDP power
| supply and fans and ventilation costs will be low, so they can
| be sold in cheap and modest cases and ps's.
|
| I'm not sure if these chips translate into laptops, but a
| laptop that games well is always desirable in the gaming
| market.
| KennyBlanken wrote:
| > This is pretty close to my 2070 GPU does
|
| Not even remotely close. It's equivalent to an RX570 or 580,
| which is roughly 1050 territory. Your 2070 is equivalent
| roughly to a 1080, plus raytracing.
|
| > Cyberpunk when it came out [..] unplayable but on the most
| powerful PCs is incredibly impressive without a GPU.
|
| The game has seen numerous patches in the last _three years
| since it was released_ that have _significantly_ increased
| its performance.
| vient wrote:
| > 63 fps on Cyberpunk 2077 which when it came out was
| "unplayable but on the most powerful PCs" is incredibly
| impressive without a GPU.
|
| Cyberpunk got a ton of (performance) fixes after release, so
| not exactly relevant.
| LeifCarrotson wrote:
| Not an expert, so excuse me if this is obvious, but would these
| integrated graphics be any good for NLP? A GPU with 24GB of video
| memory costs $2000, you can put one of these in a system with 128
| GB or 256 GB of DDR4 or DDR5 and give your neural network
| training software over 100GB of video memory if you want.
|
| You only have 12 CUs, 768 shading units, 48 texture mapping
| units, and 32 ROPs but huge amounts of cheap memory. I'm not sure
| where the bottleneck is but at least it won't crash and burn if
| you ask it to start neural network training routine that requires
| 100 GB of RAM, and you don't have to take out a second mortgage
| for a video card with the requisite amount of graphics memory.
| brucethemoose2 wrote:
| They're slow, but OK for inference.
|
| In practice no one uses AMD/Intel IGPs because no knows about
| the mlc-llm vulkan backend. llama.cpp is en vogue on the
| desktop, which does not support IGPs outside of Apple, and
| otherwise people use backends targeted at server GPUs.
| eightysixfour wrote:
| They're good from the "do it at home" perspective, not from the
| business or enterprise performance perspective.
|
| One of the ways folks do this now is use the Mac M* chips,
| since they have so much combined RAM. The raw performance isn't
| as high as GPUs, but they can fit substantially larger models
| in memory.
| bb88 wrote:
| The bottleneck would most certainly be memory, as you'll
| quickly overwhelm the on-die cache, without careful
| optimization.
|
| That said, I think AMD's chiplet strategy might come into play.
| I could see AMD release a 4 core 8 thread processor with
| increased on die cache and other chiplets being neural compute
| units.
| brucethemoose2 wrote:
| People keep reiterating this, but in practice one needs
| compute _and_ bandwidth, especially outside of tiny context
| test prompts. On my 4900HS, mlc-llm vulkan is far faster than
| CPU inference on the same memory bus, with less cache, which
| wouldn 't be the case if it was bandwidth/cache bound (since
| the CPU has far more cache as well).
|
| My 7800X3D has 96MB of L3 and a golden-bin DDR5 overclock,
| but its absolutely dreadful for inference.
| throwup238 wrote:
| _> "AMD continues to lead the AI hardware revolution by offering
| the broadest portfolio of processors with dedicated AI engines in
| the x86 market," said Jack Huynh, senior vice president and
| general manager, Computing and Graphics Group at AMD._
|
| I can't believe anyone actually said this with a straight face.
| jasongill wrote:
| It's not an inaccurate statement... I think that Intel's AI
| engine (NPU) is only available in one very-recently-released
| mobile processor, correct? AMD has had their dedicated AI on-
| chip thing for a year or more on multiple lines of CPU's.
| flumpcakes wrote:
| This looks perfect to me. I have a big desktop, but I also want a
| tiny NUC-like PC with a decent integrated graphics for playing
| with Linux and trying gaming on Linux - without having to have a
| full sized PC. I also don't want to bother dual booting, or
| running hypervisors with passthroughs.
___________________________________________________________________
(page generated 2024-01-11 23:00 UTC)