|
loving what? why on earth would they remove the feature for their highest end product (wait, is it higher than a 5950x, well whatever highest or close to highest)? is it cooking VRMs or something.
|
# ? Mar 9, 2022 18:17 |
|
|
# ? May 15, 2024 19:42 |
|
It more likely would be something to do with blowing up the enormous L3 cache. When you stack things on top of each other you cant cool the lower parts very well so it may be more sensitive to mucking around with settings than other recent processors.
|
# ? Mar 9, 2022 18:24 |
|
CoolCab posted:loving what? why on earth would they remove the feature for their highest end product (wait, is it higher than a 5950x, well whatever highest or close to highest)? is it cooking VRMs or something. it may be cooking the cache die, or the bonding might not be able to handle that much heat/physical stress from thermal cycling/etc. This is one of the very first stacked-die consumer products and the effect of that on thermals has always been a giant question mark at best. We may be getting our answer here, that yes it's a problem and pouring a bunch of heat into the stacked die either imposes thermal limits that make OC difficult/impossible for stability, or it may actually be physically bad for the chip. in principle, just like MCM, the idea is supposed to be that this lets you get a bunch more silicon into play, so you can clock it all down and boost efficiency, but performance and economics dictate that in practice companies are still gonna clock it as high as they can get away with. And that poses thermal challenges. edit: someone pointed out this may all be a misreading/exaggeration and AMD just wanted the X3D controls to be hidden in beta BIOS until chips were actually released to prevent leaks/etc Paul MaudDib fucked around with this message at 19:00 on Mar 9, 2022 |
# ? Mar 9, 2022 18:28 |
|
Isn't the consensus now that overclocking is basically never worth it for this class of chip anyway and that the age of overclocking is over?
|
# ? Mar 9, 2022 19:04 |
|
Perhaps, but only because CPUs can boost themselves to near their limit automatically If they're being unusually paranoid about manual OCing they've probably set the boost parameters more conservatively than usual too
|
# ? Mar 9, 2022 19:07 |
|
AMDs website has a 400mhz bit to base clock and 200mhz to boost clock vs the normal 5800x so something is probably up power wise.
|
# ? Mar 9, 2022 19:18 |
|
Im gonna oc it anyways and you cant stop me, lisa su
|
# ? Mar 9, 2022 20:27 |
|
kliras posted:When the 5800X3D presumably comes out, I assume it's also going to run a little cooler because they're downclocking it a bit?
|
# ? Mar 9, 2022 20:51 |
|
Llamadeus posted:Isn't the consensus now that overclocking is basically never worth it for this class of chip anyway and that the age of overclocking is over? Zen 3 CPUs aren't worth overclocking, that's for sure. Their precision boost algorithm does the job for you, and PBO will also likely be a thing still (which is just PB but with power limits increased/removed). And that's assuming that this isn't a big misunderstanding somehow. edit: by "does the job for you" I don't mean it overclocks for you, but gets the most out of the chip for you. Pushing the clocks via traditional overclocking is extremely expensive thermally, and you're better off trying to push all-core clocks up by using the curve optimizer and PBO. Dr. Video Games 0031 fucked around with this message at 21:35 on Mar 9, 2022 |
# ? Mar 9, 2022 21:10 |
|
You can tweak the precision boost algorithm, but basically what it means is that AMD processors will automatically use whatever thermal headroom you give them. So, youre not overclocking per se but its still worth it to do all the OC tier cooling because itll get you more boost.
|
# ? Mar 9, 2022 21:15 |
|
Llamadeus posted:Isn't the consensus now that overclocking is basically never worth it for this class of chip anyway and that the age of overclocking is over? Dr. Video Games 0031 posted:Zen 3 CPUs aren't worth overclocking, that's for sure. The undervolt drops temps with opens up thermal headroom which lets the chips boost a little higher. It wasn't hard to get my 5800X to peak at 5.05GHz around 81C iirc, but it was not stable.
|
# ? Mar 9, 2022 23:12 |
|
denereal visease posted:Yeah, most people got more out of their Zen3 chips by undervolting via Curve Optimizer. I don't think temps play all that big of a role, to be honest. Most coolers give enough headroom as it is, at least on my 5600X (I understand the 5800X runs hotter). On my 5600X, I primarily run into power limits instead of thermal limits, and you need to undervolt in order to attain higher clocks at the same power level.
|
# ? Mar 9, 2022 23:57 |
|
Dr. Video Games 0031 posted:On my 5600X, I primarily run into power limits instead of thermal limits, and you need to undervolt in order to attain higher clocks at the same power level. Do you know if that's made a difference in any games?
|
# ? Mar 10, 2022 00:03 |
|
Rinkles posted:Do you know if that's made a difference in any games? Not a very big one. You do get about 10% higher multi-core cinebench scores on a 5600X by doing a modest undervolt via the curve optimizer and lifting the power limits (so it runs at ~95 - 100 watts instead of ~75). In games though, either they're lightly threaded and the CPU can reach its full boost clock anyway, or they're heavily multi-threaded and you don't really need that extra power. The undervolt likely does more work than lifting the power limit does.
|
# ? Mar 10, 2022 00:24 |
|
Dr. Video Games 0031 posted:I don't think temps play all that big of a role, to be honest.
|
# ? Mar 10, 2022 00:27 |
|
Dr. Video Games 0031 posted:Zen 3 CPUs aren't worth overclocking, that's for sure. Their precision boost algorithm does the job for you, and PBO will also likely be a thing still (which is just PB but with power limits increased/removed). And that's assuming that this isn't a big misunderstanding somehow. Raichu, the same leaker, just responded that PBO and basically all other controls are also disabled for the 5800X3D.
|
# ? Mar 10, 2022 06:01 |
|
Oh man theyre juicing every last mhz out of this thing
|
# ? Mar 10, 2022 07:49 |
|
If the rumors are true, it feels like Broadwell all over again. Broadwell wasn't a slouch compared to Skylake, but the optimization crew at Intel knew how to keep iterating on Skylake, while the Broadwell experiment went back to collect dust on the shelf. With Zen 4 already in production, why continue to bother innovating with a Hail Mary one-off design when tons of people will buy the normal thing that just works, anyways?
|
# ? Mar 10, 2022 08:05 |
|
Sidesaddle Cavalry posted:If the rumors are true, it feels like Broadwell all over again. I personally know two people that put off upgrading to 12th gen to say "I'm waiting to see Zen3D". It's a powerful thing to have "the best gaming CPU" looming on the horizon.
|
# ? Mar 10, 2022 08:10 |
|
Definitely gonna wait for reviews on this one, too many caveats not to
|
# ? Mar 10, 2022 10:57 |
|
Got a 5600G that keeps crashing on heavy AVX processing. It's running in an Asus B-550 PLUS motherboard (BIOS 2423), and the RAM (F4-3600C16D-16GTZN) is on the qual list. Memtest86+ shows no errors even with XMP (DOCP) on, but the crashing happens even with RAM on default speed. No PBO or other changes enabled. I'll try upgrading the BIOS to 2604 or failing that downgrading to 1804 and see if it fixes it.
|
# ? Mar 10, 2022 12:04 |
|
Trolling Thunder posted:Oh man theyre juicing every last mhz out of this thing Or it's really easy to fry that 3D cache if you futz with the voltage
|
# ? Mar 10, 2022 14:00 |
|
https://twitter.com/1usmus/status/1501889826848755714?s=21 https://twitter.com/1usmus/status/1501890684206432257?s=21
|
# ? Mar 10, 2022 17:11 |
|
Vir posted:I'll try upgrading the BIOS to 2604 or failing that downgrading to 1804 and see if it fixes it.
|
# ? Mar 10, 2022 18:54 |
|
Whatever, I don't care about overclocking. Show me some benchmarks.
|
# ? Mar 10, 2022 21:41 |
|
Cygni posted:https://twitter.com/1usmus/status/1501889826848755714?s=21 yep compute die is cooking the cache so they have to turn one or the other down a bit (hence clock rate reductions and no OC). Probably limited by how fast they can move heat through the IHS, so no point messing with it. Everyone's always acknowledged it as being the big open question about die stacking, what do you do about heat, and imo this suggests that yes it's a problem. I would actually say this bodes poorly for the future of 600W GPUs because the same thermal density things apply there too. At some point physics will determine that you are gonna have to clock that compute die down a bit. (I also wonder if with the cache so exposed on top like that, could you heat/cryo glitch it in some way? I don't think that's the reason here but if it's crashing, can you get it to crash in an exploitable way?)
|
# ? Mar 11, 2022 03:41 |
|
Paul MaudDib posted:I would actually say this bodes poorly for the future of 600W GPUs because the same thermal density things apply there too. At some point physics will determine that you are gonna have to clock that compute die down a bit. IIRC the die size of the 3090 is ~6x larger than the 5800X3D, I'd imagine that makes things a little easier
|
# ? Mar 11, 2022 07:44 |
|
New Zealand can eat me posted:IIRC the die size of the 3090 is ~6x larger than the 5800X3D, I'd imagine that makes things a little easier Isn't that just more transistors putting out heat that need cooling?
|
# ? Mar 11, 2022 17:59 |
|
Danhenge posted:Isn't that just more transistors putting out heat that need cooling? Yeah but the thermal density is lower. Regardless of it being possible though, at some point heat and power consumption will hit a wall because your average person doesn't want their office to feel miserably hot in the summer.
|
# ? Mar 11, 2022 19:42 |
|
Danhenge posted:Isn't that just more transistors putting out heat that need cooling? Its more of a transistor density issue. When it is too dense thermal energy cannot dissipate into the heat spreader fast enough. When things are spread out on a larger die this isn't a issue.
|
# ? Mar 11, 2022 19:48 |
|
Inept posted:Yeah but the thermal density is lower. Regardless of it being possible though, at some point heat and power consumption will hit a wall because your average person doesn't want their office to feel miserably hot in the summer. It's a galaxy-brain move by AMD and Nvidia since they'll be able to market their GPUs as being a way to heat your home without Russian gas.
|
# ? Mar 11, 2022 20:44 |
|
The new Threadrippers have shown up on Passmark. The 64 core PRO 5995WX hits a new record multi-core score of 108,822. The old record was a 64 core epyc at 88,338, while the highest non-AMD score is an Intel 40 core at 62,317. https://www.cpubenchmark.net/high_end_cpus.html For Passmark, AMD still need a 20% single thread performance boost to match the best Intel, or even just 10% to match the M1. Hopefully they've put a lot of focus on ST for Zen4 & 5; while Epyc might still benefit from more core density, I think consumer Ryzen has hit the point where more cores isn't that useful right now both in terms of workflows that can make use of them or a useful performance uplift. The 5950X always seems to be an underwhelming increase over the 5900X, failing to scale close to the 33% extra core count. There's the drop in Ghz but I also suspect that often there's a memory bandwidth bottleneck.
|
# ? Mar 12, 2022 11:40 |
|
someone selling a 3900X and cooler on facebook market for 120 i could ebay my 3600 for at least 100. can an MPG X570 GAMING PLUS comfortably run that chip? i also have a 650w PSU and a 3070 i won't have power issues will i? very very tempting
|
# ? Mar 15, 2022 14:15 |
|
CoolCab posted:can an MPG X570 GAMING PLUS comfortably run that chip? CoolCab posted:i also have a 650w PSU and a 3070 i won't have power issues will i? very very tempting
|
# ? Mar 15, 2022 14:28 |
|
loving sweet! aight, 12c/24t here i come lol
|
# ? Mar 15, 2022 16:01 |
|
beep beep here comes the AMD-mobile
|
# ? Mar 15, 2022 16:04 |
|
https://twitter.com/VideoCardz/status/1503720246900932611 tbh the prices are a little bizarre. Comparing the current Intel non-microcenter street prices, 5500 is $15 more than a 12100F, 5600 is $20 more than a 12400F, 5700X is bang on the 12700F. I guess you could say that's basically the price adjustment for the motherboards, but B550 and B660 are really closely priced in the market right now. And it seems likely that the Intel options will be (a little) more performant and have a better feature set. Guess they aren't quite feeling the pinch yet enough to get deeper into the price war, so instead just met prices 1 to 1. The 5800X3D at $449, i sorta feel like they dont want our expect to move many. Guess we should wait for reviews, cause I have no idea how that thing is gonna perform.
|
# ? Mar 15, 2022 20:03 |
|
Since theyre within like $20 I think well have to wait to see street prices.
|
# ? Mar 15, 2022 20:07 |
|
300 series support is surprising. I was at one point hoping to get a 3900x to upgrade my 1700 but prices never dropped and decided to wait till Zen4 and make the DDR5 jump because I want to upgrade to 64gb of ram as well. May decide to stick it out if MSI releases new bios.
|
# ? Mar 15, 2022 23:16 |
|
|
# ? May 15, 2024 19:42 |
|
Lol wow my B350 might be able to use Zen 3? I've been playing the long game all this time and I didn't even know it.
|
# ? Mar 16, 2022 01:09 |