|
fwiw greymon is still saying the announcement isn't expected until early september, right before the launch on september 15th and the v-cache parts are over 2 months behind the rest in production and so are expected to launch late november or december
|
#
?
Jul 27, 2022 23:06
|
|
|
|
| # ? Jun 5, 2024 19:32 |
|
|
They are gonna milk the launch, like they have done the last few. Boards and general CPU SKUs are expected next week. Wouldn’t surprise me if they have separate events for CPU details, reviews, and actual launch, too.
|
|
#
?
Jul 27, 2022 23:16
|
|
|
I don't know if I'm interested in upgrading if the 7000 series turns out to be exactly the 5000 series, but Zen 4. The 5000 series itself was already exactly the 3000 series, but Zen 3. With the exception of my desktop. I will absolutely upgrade it from a 5000 series APU to a 7000 series, whenever those become available. DDR5 and RDNA2 are big, obvious wins for APUs. Everything else though, I'm gonna be waiting on hard specs direct from AMD before sitting down to think sober cost/benefit thoughts.
|
|
#
?
Jul 28, 2022 00:25
|
|
|
What are you talking about? 5000 series was a pretty big improvement over the 3000 series.
|
|
#
?
Jul 28, 2022 00:54
|
|
|
I really don't know what "the same as the last generation, but with new architecture" is supposed to mean. Isn't the new architecture the whole point? If the combination of new architecture and a node shrink gives a good boost to overall performance and perf per watt, then what else are you looking for? The Ryzen 7000 series will also support AVX-512, if that's important to you.
|
|
#
?
Jul 28, 2022 01:00
|
|
|
I'm aware of the IPC gains. That's why I said "the same, but Zen N+1" instead of "the same." The products that embodied Zen 1, +, and 2 were an absolutely crazy ride of foundational architectural change. Increasing core counts. Ridiculous amounts of cache compared to anything that didn't cost 5X as much. Fascinating possibilities of chiplet architecture. In a nutshell, more capabilities. It was fantastic. But Ryzen 5000 was exactly Ryzen 3000, except with Zen 3 cores. Same core counts. Same cache amounts. Same layouts. Same capabilities. Just an IPC bump. Well, and a small re-arrangement of the L3 cache, but that mostly just got expressed as a fraction of the IPC improvements. And for the record, I still upgraded half of my machines, and was/am happy about it. But I wasn't excited by it. I'm not saying that anything sucks, or arguing that no one should upgrade, or any such thing. I'm just saying that if Ryzen 7000 is really the Ryzen 3000 designs, again, but with Zen 4 cores, I may not automatically upgrade as I have with past generations.
|
|
#
?
Jul 28, 2022 01:28
|
|
|
Zen+ was just a refresh of Zen 1, your narrative is way off. Zen 3 has less in common with 2 than + compared to 1. If you’re just disappointed that each market segment has the same amount of cores as the previous 5 iterations, just say that.
|
|
#
?
Jul 28, 2022 01:31
|
|
|
mdxi posted:I'm aware of the IPC gains. That's why I said "the same, but Zen N+1" instead of "the same." All of which was only a thing because of a decade of profit-taking and lack of innovation by a dominant monopolist. I cannot imagine a more cursed monkey's paw wish in the tech space. Continual smaller improvements are much more the norm. Crazy architecture changes are what you do when you have to take a big swing. They're big risks, and they fail as often or more than they succeed.
|
|
#
?
Jul 28, 2022 02:10
|
|
|
mdxi posted:But Ryzen 5000 was exactly Ryzen 3000, except with Zen 3 cores. Same core counts. Same cache amounts. Same layouts. Same capabilities. Just an IPC bump. Well, and a small re-arrangement of the L3 cache, but that mostly just got expressed as a fraction of the IPC improvements. A small re-arrangement of L3? I wish you had said this about a change in an Intel architecture so Jawnv6 could rip this post to shreds in style. I top out at snarky one liners so let me just ask: How does one zen 2 core on a CCX talk to a zen 2 core on another CCX in the same CCD? If more and more applications are doing at least light work on 4 cores then what would happen if the most popular parts had less than 4 core CCXs? How is this behavior different from any zen 3 core in a CCD talking to any other core in the same CCD? edit: to cut to the chase: why is the 3300x the “best” zen 2 cpu for many people? hobbesmaster fucked around with this message at 02:56 on Jul 28, 2022 |
|
#
?
Jul 28, 2022 02:44
|
|
|
Expecting AMD to reinvent the wheel every generation is a bit much. They had a lot of groundwork to lay to get where they are now with the Zen architecture, but now that groundwork has been established. There is still room for big changes, though. 3D cache is one of those, and perhaps we'll eventually see other uses for die stacking.
|
|
#
?
Jul 28, 2022 02:52
|
|
|
Zen 3 also added resizeable BAR, which admittedly doesn't amount to that much yet in most cases but it is a new capability if you want to wait for a giant architectural shift then Zen 5 is expected to move to an Alder Lake-style hybrid architecture
|
|
#
?
Jul 28, 2022 03:47
|
|
|
lih posted:Zen 3 also added resizeable BAR, which admittedly doesn't amount to that much yet in most cases but it is a new capability I don't think we'll see a little/big cores from AMD, it will mess up their epyc pipeline too much as all cores gets judges if they can make it in epyc cpus.
|
|
#
?
Jul 28, 2022 04:00
|
|
|
wargames posted:I don't think we'll see a little/big cores from AMD, it will mess up their epyc pipeline too much as all cores gets judges if they can make it in epyc cpus. We will see! Floating something out of my rear end completely: 8c Zen 5 CCD die + 16c Zen 4D CCD die from Bergamo connected using TSMC's InFO-L (aka Apple's "Ultra Fusion"), plus an IO/GPU/accelerator die via CoWoS. And hell, 3d stack a vcache die on it while we are here. The patents (where AMD calls their big.LITTLE variant "slim.FAST" lol, don't think that term is gonna get past the corporate lawyers) do seem to suggest that they are planning something with small cores. Early Zen 5 rumors had Granite Ridge, the desktop part, as a single 8c Zen 5 CCD and a single 16c Zen 4D CCD. The multi-die strategy would also match intel's plans for Meteor Lake in 2023, so it seems like the true 3d chiplet future is indeed here next year. Also, frankly, i don't think there is any way either Apple or AMD can hope to stay competitive with Apple if they do not aggressively pursue bigLITTLE. It is not pretty at the moment.
|
|
#
?
Jul 28, 2022 04:35
|
|
|
Apple m1 was so good because it was on tsmc 5nm and everyone was comparing it to older 7nm stuff, and from what we can tell from the 800w nvidia gpus the new 5nm process is truly bonkers on how good it can be, but we will see when everything comes out this year.
|
|
#
?
Jul 28, 2022 04:45
|
|
|
wargames posted:I don't think we'll see a little/big cores from AMD, it will mess up their epyc pipeline too much as all cores gets judges if they can make it in epyc cpus. I mean, they've already announced two different types of Zen 4 cores, so it seems plausible.
|
|
#
?
Jul 28, 2022 04:47
|
|
|
gradenko_2000 posted:budget CPUs are in a pretty good place nowadays, what with a Pentium capable of driving a lot of stuff, and even an i3 is gonna be plenty for anyone not doing productivity/professional-type work That's what Mendocino is meant to address. 6nm "last gen process" 4c/8t + 2CU RDNA 2. It's the "six bullets, more than enough to kill anything that moves" of craptops, and not accounting if they separate into several bins of Mendocino out to like, 4/4 or 2/4. Espcially in light of recent discoveries that AMD is advancing their GPU drivers by leaps and bounds, which benefits everything that AMD touches, from the Steam Deck to the Playstation/Xbox back catalogs. SwissArmyDruid fucked around with this message at 05:38 on Jul 28, 2022 |
|
#
?
Jul 28, 2022 05:23
|
|
|
lih posted:Zen 3 also added resizeable BAR, which admittedly doesn't amount to that much yet in most cases but it is a new capability IIRC resizable BAR is a PCIe thang and is ISA independent (and supported since PCIe 1.0?). It was some AMD GPU/driver that introduced the huge BAR because we aren't running 32bit XP anymore and you can direct map 16GB into some address space as needed. Some goons here know the dirty PCIe details for sure.
|
|
#
?
Jul 28, 2022 05:45
|
|
|
karoshi posted:IIRC resizable BAR is a PCIe thang and is ISA independent (and supported since PCIe 1.0?). It was some AMD GPU/driver that introduced the huge BAR because we aren't running 32bit XP anymore and you can direct map 16GB into some address space as needed. Some goons here know the dirty PCIe details for sure. Correct. It's part of the PCIe spec, and in order for it to work, the PCI root bridge, any bridges in the middle, and the end device must all support resizable BARs, and the operating system's PCIe subsystem and the device's driver have to as well. To oversimplify it, without resizable BARs, a BAR's size can only be up to 256 MB wide, so if you have to transfer more than 256 MB of data, you have to change the BAR address every time you want to do another chunk of the read/write. With a resizable BAR you can scale the BAR's size up to 512 GB. There's a spec for Extended Resizable BARs that I haven't read because I don't want to pay PCI-SIG several hundred dollars out of curiosity, and it seems to let you scale up to 263 bits.
|
|
#
?
Jul 28, 2022 05:53
|
|
|
oh i remembered it being pushed the first generation of AMD CPUs to actually implement it but it does look like they eventually released driver updates for older chips too, had completely forgotten about that
|
|
#
?
Jul 28, 2022 06:31
|
|
|
AMD has been floating the idea of combo X86 and ARM cores on a single CPU for servers for ages now, and I have to imagine the reason they never did it was because the market didn't say "yes we will buy that". The heterogeneous cores idea is a big deal for laptops and mobile though. wargames posted:Apple m1 was so good because it was on tsmc 5nm and everyone was comparing it to older 7nm stuff, and from what we can tell from the 800w nvidia gpus the new 5nm process is truly bonkers on how good it can be, but we will see when everything comes out this year. An M1 Pro also has more transistors than a 5900X and a 3070 Super combined, so calibrate expectations accordingly. It's a super-wide design that is incredibly strong in some areas and not particularly amazing in others. Not a "Intel and AMD are both doomed" thing.
|
|
#
?
Jul 28, 2022 14:21
|
|
|
Klyith posted:An M1 Pro also has more transistors than a 5900X and a 3070 Super combined, so calibrate expectations accordingly. It's a super-wide design that is incredibly strong in some areas and not particularly amazing in others. Not a "Intel and AMD are both doomed" thing. Yeah that's kinda the thing that gets forgotten when most people bring up M1 comparisons- Apple doesn't give a poo poo (well, they do, a little, but not like intel and amd do) about the individual CPU unit cost because they're never selling them on their own, so they are absolutely titanic size wise.
|
|
#
?
Jul 28, 2022 14:44
|
|
|
lih posted:rumour is also AMD has internally tested DDR4 for Zen 4 as well and will consider letting DDR4 motherboards for it launch if DDR5 is still in too bad a place Mr.Radar posted:If they went ahead with DDR4 AM4 I wonder if they would launch a DDR4 AM5 platform or if they would launch Zen 4 simultaneously on AM4 and AM5.
|
|
#
?
Jul 28, 2022 19:01
|
|
|
Gwaihir posted:Yeah that's kinda the thing that gets forgotten when most people bring up M1 comparisons- Apple doesn't give a poo poo (well, they do, a little, but not like intel and amd do) about the individual CPU unit cost because they're never selling them on their own, so they are absolutely titanic size wise. Sure but also the profit margin, die size, transistor count, etc are not really the end customers problem. On battery power, the M2 macs absolutely destroy the i7 1260p and Ryzen 6800U that it competes against. Apple will sell you an M2 laptop right now for $1100 that on battery power will outscore an i7 1260P in both single and multi core in something like Geekbench. While using 1/3rd of the power and having huge battery life. With no fans. And it can still do that even with the performance cores disabled and only using the efficiency cores, lol. (apologies for the annoying rear end Tuber vid, but i havent seen anyone else do that particular test yet) I personally don't see a way for AMD to stay competitive in the non-gaming laptop world without investing in small cores. But I'm not an expert! It's possible that AMD stays the all big core course, but it sure seems like "the ARM problem" is especially difficult in laptop world.
|
|
#
?
Jul 28, 2022 19:52
|
|
|
Maybe I'm valuing gaming too much but if Phoenix Point can finally deliver competitive numbers to midrange laptop dGPUs, that's a whole segment that neither Intel nor Apple can really touch.
|
|
#
?
Jul 28, 2022 21:08
|
|
|
Just updated to a 5600x with a msi tomahawk. Booted up fine afyer fitting it all, then i did a bios update using m-flash. Now when i power on i just get a black screen. No bios booting. Have tried pulling the cmos and trippong the cmos reset. Pulled all components out except the cpu. Also tried individual ram sticks. Anyone had this?
|
|
#
?
Jul 29, 2022 16:18
|
|
|
did it lose power when trying to do the m flash? that sounds like it's not posting, which might be indicative of an issue with the bios usually. if it lost power while bios updating that could be very bad.
|
|
#
?
Jul 29, 2022 16:21
|
|
|
https://twitter.com/pouiee903357/status/1553039099761672193?s=21&t=L_x7nz2AFUovVaNvXDyyag V nice single core number there
|
|
#
?
Jul 29, 2022 16:40
|
|
|
CoolCab posted:did it lose power when trying to do the m flash? that sounds like it's not posting, which might be indicative of an issue with the bios usually. if it lost power while bios updating that could be very bad. Nope. Ran through the process fine with mflash then when it rebooted after completion it did this.
|
|
#
?
Jul 29, 2022 16:59
|
|
|
Do you have boot status LEDs or indicators on the motherboard? I would think a Tomahawk would have something. Check the motherboard manual and then see what they indicate.
|
|
#
?
Jul 29, 2022 17:03
|
|
|
Cygni posted:https://twitter.com/pouiee903357/status/1553039099761672193?s=21&t=L_x7nz2AFUovVaNvXDyyag too bad userbenchmarks is a trash company.
|
|
#
?
Jul 29, 2022 17:20
|
|
|
v1ld posted:Do you have boot status LEDs or indicators on the motherboard? I would think a Tomahawk would have something. Check the motherboard manual and then see what they indicate. All fine. Cpu led lights up briefly when you turn power on but goes off after half a second. There is talk online about it causing issues with display ports and so on. Can hear my physical hard drive doing its thing but keyboard and mouse dont seem to get power. I tried using the flash button on the back whuch you can use to flash wuth no cpu in. It starts the process, accesses the usb then stops. Obviosly nothing on screem to tell you its going on. Currently habe unplugged the cmos battery and leavung it a few hours to see if that works.
|
|
#
?
Jul 29, 2022 17:39
|
|
|
Hmmmm. Seems for flashback i need to change the usb to MBR instead of GPT. Will give that a try.
|
|
#
?
Jul 29, 2022 18:03
|
|
|
Cygni posted:Sure but also the profit margin, die size, transistor count, etc are not really the end customers problem. On battery power, the M2 macs absolutely destroy the i7 1260p and Ryzen 6800U that it competes against. Apple will sell you an M2 laptop right now for $1100 that on battery power will outscore an i7 1260P in both single and multi core in something like Geekbench. While using 1/3rd of the power and having huge battery life. With no fans. And it can still do that even with the performance cores disabled and only using the efficiency cores, lol. Oh, for sure, I think it's a given that AMD is going to diversify their core designs. Modern computing being such a mix of low intensity "just display some text/play a youtube" next to "Run full 3d rendering/play games/scientific models" that it's impossible to make something power competitive without specializing the hardware for each type of workload. They're already pretty much doing that for the next generation, although we don't really know how much to the same degree as the arm big core/little core chips are.
|
|
#
?
Jul 29, 2022 18:08
|
|
|
wargames posted:too bad userbenchmarks is a trash company. Unfortunately, we don’t get to choose which benchmark the Asrock tech accidentally uploads because they forgot to turn off wifi, lol
|
|
#
?
Jul 29, 2022 18:11
|
|
|
Bigsteve posted:There is talk online about it causing issues with display ports and so on. You have a separate GPU right? The 5600x does not have an integrated GPU so any video output on the mobo is useless. Bigsteve posted:Hmmmm. Seems for flashback i need to change the usb to MBR instead of GPT. Will give that a try. Yeah it'll need to be MBR and basic FAT32. While using flashback be absolutely sure that it's finished before you pull the stick / reset power.
|
|
#
?
Jul 29, 2022 19:09
|
|
|
Yeah, seperate gpu. My plan us to start the flashback and just leave it till i wake up.
|
|
#
?
Jul 29, 2022 19:11
|
|
|
Klyith posted:While using flashback be absolutely sure that it's finished before you pull the stick / reset power. Quaint Quail Quilt fucked around with this message at 19:24 on Jul 29, 2022 |
|
#
?
Jul 29, 2022 19:22
|
|
|
Ok well at least the dude at Asrock has a sense of humor. This is the same leaked 7600X result as before, with the name changed:
|
|
#
?
Jul 30, 2022 04:35
|
|
|
lmao, at least they're owning it
|
|
#
?
Jul 30, 2022 04:42
|
|
|
|
| # ? Jun 5, 2024 19:32 |
|
|
How do the numbers work with AMD, why were (iinm) the 4000 and 6000 Ryzen chips mobile only?
|
|
#
?
Jul 30, 2022 04:53
|
|
