|
ASUS: Insane prices, bizarre feature sets or design decisions, have released products with insufficient VRMs or insufficient coolers, questionable RMA practices. Will black list reviewers who are mean. Generally less lying in their PR/advertising than others, though? Asrock: What the gently caress is going on over there lately? Prices and products that make absolutely no sense. Probably last place in the BIOS wars, but they are all the same really. Have released boards with insufficient VRMs, questionable RMA practices. Will black list reviewers who are mean. Gigabyte: History of component downgrades after launch and/or releasing multiple versions of products in the channel to purposefully confuse buyers, history of releasing half baked products. History of PR that is straight up lying and aggressive to people who correctly point out problems. Have released boards with insufficient VRMs, very questionable RMA practices, and you know... EXPLODING PSUs. More competitive on price. MSI: Annoying product naming, ugly branding, history of fast but unstable BIOS releases, bizarre and shady corporate practices, have released (many) boards with insufficient VRMs or coolers including GPUs that would throttle on even open air test benches, very questionable RMA practices. More competitive on price. Others: EVGA: Have had multiple products that were underbaked or straight up flawed in design at launch. Support is exemplary though. Smaller manufacturer and it shows. Biostar/ECS: LOL Cygni fucked around with this message at 23:12 on Nov 21, 2021 |
# ? Nov 21, 2021 23:09 |
|
|
# ? May 22, 2024 13:52 |
|
The naming conventions are pretty funny. Basically every manufacturer has a "Pro" model, but how high up in their lineup it is seems completely random.
|
# ? Nov 21, 2021 23:18 |
|
Cost-cutting board revisions aren't a thing on Gigabyte products if you only buy the top-dollar super-overclocking limited-production models
|
# ? Nov 22, 2021 01:38 |
|
Anything specific about the motherboard I should look out for?
|
# ? Nov 22, 2021 01:45 |
|
The 5800X is back down to $300 at Microcenter.
|
# ? Nov 22, 2021 08:18 |
|
Cygni posted:ASUS: Insane prices, bizarre feature sets or design decisions, have released products with insufficient VRMs or insufficient coolers, questionable RMA practices. Will black list reviewers who are mean. Generally less lying in their PR/advertising than others, though? how can they all have insufficient vrms, how much are you juicing your cpus
|
# ? Nov 22, 2021 15:25 |
|
Surprise, surprise. Anecdotes are spotty and can color the perception of the brand’s entire product line
|
# ? Nov 22, 2021 15:48 |
|
Toalpaz posted:how can they all have insufficient vrms, how much are you juicing your cpus Every OEM makes super-cheap barebones mobos for the system integrator and office market, which are intended for pcs that run nothing more exciting than Word. A lot of those I would be uncomfortable putting in a gaming / enthusiast PC even if the CPU was never going to be overclocked. A cheap, no heatsinks, min-spec power system will run that CPU adequately enough. But it will be very hot and inefficient when doing so, increasing system temps and getting lower component lifetimes. OTOH some people just watch too many buildzoid videos.
|
# ? Nov 22, 2021 16:15 |
|
Would those low-end mobos still have problems if you're dropping in 65W chips like a 5600X or an eventual non-k i7-12700? I'm assuming that they'd do fine with lower TDP chips even running sustained load.
|
# ? Nov 22, 2021 17:07 |
|
an AMD CPU is going to regulate itself well enough that an underbuilt VRM on a cheap mobo isn't going to "kill" your computer, you'll just be leaving some performance on the table the more practical reason why you might not want to use an A320/A520 is due to very barebones I/O
|
# ? Nov 22, 2021 17:18 |
|
Twerk from Home posted:Would those low-end mobos still have problems if you're dropping in 65W chips like a 5600X or an eventual non-k i7-12700? I'm assuming that they'd do fine with lower TDP chips even running sustained load. That depends how long end you mean. The lowest end mobos may not let a "65W" go to its full potential (ie 100W) but anything marketed to DIYers will probably be fine there. For stuff like the 5800X with a higher TDP and then anything with 2 chiplets thats where the different VRMs actually matter. On the intel side you really need a good high quality motherboard to run the processors at 200W or whatever the gently caress for that last 5% performance gain you get at pl2 and max (potentially unlimited) tau. You probably don't actually want to do that though.
|
# ? Nov 22, 2021 17:33 |
|
hobbesmaster posted:On the intel side you really need a good high quality motherboard to run the processors at 200W or whatever the gently caress for that last 5% performance gain you get at pl2 and max (potentially unlimited) tau. You probably don't actually want to do that though. This was the case when I was looking at B560 motherboards for a 11600K (ended up with a Z590). Some boards would begin throttling with the power limit disabled.
|
# ? Nov 22, 2021 17:47 |
|
Twerk from Home posted:Would those low-end mobos still have problems if you're dropping in 65W chips like a 5600X or an eventual non-k i7-12700? I'm assuming that they'd do fine with lower TDP chips even running sustained load. It depends what your definition of "fine" is. Operates at designed socket power? Yes. Even with a higher TPD processor, though that gets sketchy with current Intel since the "designed power" is at this point an elaborate piece of fiction. Boosts to the full performance potential of the CPU? Maybe, maybe not. Works for 5 or 10 years and is still good and 100% stable? I dunno, how many hours a week are you running at high load? And how hard is the CPU actually working -- even a lot of games don't call for max power from the CPU. Keep in mind that this is not just "low end", but the lowest of the "sort by price: lowest" end. Here's an example I can give off the top of my head: during the AMD 400-chipset era, MSI had the B450 Tomahawk, the Pro-VDH, and the M-A Pro. IIRC all three used basically the same 4-phase VRM component setup. The tomahawk was one of the most frequently recommended boards for average mid-range builds, great for 6 and 8 core ryzens. Tons of goons in the pc build thread bought those. The Pro-VDH had less features than the Tomahawk, but at least it had a VRM heatsink. It got some nods for people on limited budgets, and it had a MATX variant while the Tomahawk's MATX equivalent was impossible to get outside of the EU and was more expensive than the ATX for some reason. And the M-A Pro had no heatsink, and was a stay-away other than someone building a mom-puter. The price differences with all three was about $30 at the time. gradenko_2000 posted:an AMD CPU is going to regulate itself well enough that an underbuilt VRM on a cheap mobo isn't going to "kill" your computer, you'll just be leaving some performance on the table It's not gonna hurt the CPU or other parts, but the mobo components like capacitors and mosfets have lifespans that are strongly affected by operating temperature. VRMs without heatsinks dissipate most of their heat through the PCB, where it raises the temperature of all parts. And a crap budget VRM makes more waste heat per watt of CPU. This is some of the same issue with the "does overvolting harm your CPU?" question. It depends a lot on what your idea of a reasonable lifetime for components is. Some people say no because their horizon for life is much shorter than the damage from overvolting does harm. Other people say yes because anything that reduces lifespan qualifies as harm.
|
# ? Nov 22, 2021 18:08 |
|
Klyith posted:This is some of the same issue with the "does overvolting harm your CPU?" question. It depends a lot on what your idea of a reasonable lifetime for components is. Some people say no because their horizon for life is much shorter than the damage from overvolting does harm. Other people say yes because anything that reduces lifespan qualifies as harm. Does this apply to memory as well? Using XMP profiles, not necessarily custom OCing.
|
# ? Nov 22, 2021 18:46 |
|
Rinkles posted:Does this apply to memory as well? Using XMP profiles, not necessarily custom OCing. xmp profiles are how you get the manufacturer's advertised speeds out of your memory. so if you like paying a 40% premium for OC you won't use on a 3600 speed ram, and running it at 2200 to add some durability despite it's lifetime warranty, underclocking it makes sense.
|
# ? Nov 22, 2021 19:08 |
|
Klyith posted:Works for 5 or 10 years and is still good and 100% stable? I dunno, how many hours a week are you running at high load? And how hard is the CPU actually working -- even a lot of games don't call for max power from the CPU. this point is worth emphasizing, even in the HUB example that's no-power-limits Blender (so, all-core AVX2 workload) which is practically speaking one of the heaviest "realistic" loads possible - video encoding being the other one imo, as you can potentially do that for hours and hours and it's the same thing, very heavy all-core AVX2 workload. if you're gaming, that same board likely never throttles even without a power limit, and games won't be bumping into the power limit in the first place. So for the average home user, the solution is to keep a "realistic" power limit (say, 100W or 125W) which should never be bumped into while gaming, and then if you do actually end up doing blender sometime then you just deal with 5% less performance due to the power limit cap. if you know you're gonna be doing a ton of 3d rendering/video encoding then yeah, you gotta optimize your system around it, but a lot of users won't ever, or would only occasionally do it. Remember that the 12900K has 90% performance at half the power - Intel just set an insane power limit to try and push it into competition against the 5900X and 5950X, despite it having way fewer P-cores. It's impressive that 8 P-cores and 4 E-cores can even compete with 12 P-cores at all, and if the price were right that would be an obvious win vs (say) the 5800X - which is why you've seen prices plummet on the 5800X, as AMD brings their prices back down to compensate. AMD needs to position the 5900X against 12700K, and the 5800X against the 12600K, if it's 5800X against the 12700K then the 12700K can be strongly argued to be the better buy, even if it needs a more expensive board. but generally speaking you don't want to be running the $100 tier of motherboard anyway, they're poo poo. The difference between a $100 and a $150 motherboard is pretty stark and $200 buys you a really nice one, at some point it does just become false economy to go with a really really lovely motherboard just to win a build comparison, someone would have to live with that hardware. Paul MaudDib fucked around with this message at 19:51 on Nov 22, 2021 |
# ? Nov 22, 2021 19:45 |
|
I ended up spending a little bit more than I intended to get better wifi and more usb ports. Plus it comes with a free copy of Outriders!
|
# ? Nov 22, 2021 19:48 |
|
Toalpaz posted:how can they all have insufficient vrms, how much are you juicing your cpus they have all tried to cut costs on a product farther than they should have over the years and released models that were flatly unacceptable. most of the time it’s the low end boards others have talked about, but there have also been GPUs and even “high end” Gamer boards that struggled, throttled, or were unstable at even stock settings i probably should have also noted that all 4 manufacturers are also guilty of “auto-overvolt if you dare touch XMP” dogshit at one point or other, too
|
# ? Nov 22, 2021 20:46 |
|
Bring back ABit imo
|
# ? Nov 22, 2021 20:55 |
|
PC "Probably Counterfeit" Chips
|
# ? Nov 22, 2021 21:05 |
|
redeyes posted:Got a Gigabyte X570 Aorus pro wifi. No issues at all. Love it. quote:Asrock: [/b]What the gently caress is going on over there lately? Prices and products that make absolutely no sense. Probably last place in the BIOS wars, but they are all the same really. Have released boards with insufficient VRMs, questionable RMA practices. Will black list reviewers who are mean. The "insufficient" VRM thing is really overblown if it's anything on AM4. Not as up to date on any dumb moves they've made on the intel side lately. My issue with ASRock x570 was stock at release & port choices that blew for my use case. Gigabyte went "we're going to design one, maybe two motherboards, and then just gate certain features up and down the stack" which meant their $200-$260 boards were rocking similar components to everyone else's $300 boards at release. They also made lots of really practical choices that other vendors didn't. MSI made zero sense in their x570 decision making process outside of their $180-$210 boards which were competitive. Asus got murdered on price at release. ASRock had one or two good boards but at $300 and $lowend. In general, I like ASRock & MSI. ASRock tries to do cool poo poo and usually cuts the right corners at lower price points (the x370/x470 taichi boards, asrock's am4 server boards, the b450 pro4 at $60 and similar are all standard asrock). MSI usually makes a leader at some price point in any given generation's stack (b450 tomahawk, x470 pro carbon are peak "what MSI does" examples -- no bullshit boards that hit what a decent portion of people want & are completely peerless at their price point), Asus is the most consistent company at making reasonable boards, and Gigabyte does the same thing MSI does sometimes (x570 elite / elite wifi, x570 pro wifi slayed at their price points in the release lineup.) Khorne fucked around with this message at 22:00 on Nov 22, 2021 |
# ? Nov 22, 2021 21:37 |
|
priznat posted:Bring back ABit imo That Abit Nforce board was sooooo good. I think it was the Abit K7
|
# ? Nov 22, 2021 21:42 |
|
Stanley Pain posted:That Abit Nforce board was sooooo good. I think it was the Abit K7 They had some monster boards back in the day. The one I had to OC the Celeron 300A was a rock solid beast.
|
# ? Nov 22, 2021 21:58 |
|
priznat posted:They had some monster boards back in the day. The one I had to OC the Celeron 300A was a rock solid beast. BX6/BH6/BX6 Rev 2 are legendary, baby. Found this old AT roundup, rip to like 90% of these brands in the consumer space. https://www.anandtech.com/show/294/7
|
# ? Nov 22, 2021 22:08 |
|
Rinkles posted:Plus it comes with a free copy of Outriders! I'm so sorry
|
# ? Nov 22, 2021 22:51 |
|
Khorne posted:Same. Great motherboard. I haven't had any issues beyond self-inflicted issues with a fairly tight memory OC. i will in defense of my choice say that on paper the gigabyte board was the best bang for the buck at the time I was buying it, and except for the fan I had no issues - and the fan was a potential issue on every x570 board sold at the time. however, since then their RMA policy and general shadiness is just... ugh.
|
# ? Nov 22, 2021 23:29 |
|
Khorne posted:The "insufficient" VRM thing is really overblown if it's anything on AM4. Not as up to date on any dumb moves they've made on the intel side lately. The B350 HDV is one I would qualify as having an insufficient VRM from Asrock. Its official CPU support list includes stuff like the 3950X and 3800XT, but the VRM cannot handle them at stock settings without direct airflow, as users have repeatedly confirmed in real world situations. Frankly, they never should have been put on the support list and its not really an acceptable situation to market them as they do. And like we talked about, literally every vendor has pulled something similar... and most of em did it relatively recently with B350/B450/A320. And yeah, not many people would pair a bargain basement B350 or A320 board with a 3950X in a 1 case fan setup, but thats kinda not the point. The board vendors market the boards as supporting the 3950X, and they need to back that up or remove it from the list. Or hell, at least put an asterisk so consumers know what they are getting in to. The vendors know this, all 4 have A320 boards that specifically don't include stuff over 65w on the support list for just this reason. So yeah, i dont give em a pass.
|
# ? Nov 23, 2021 00:07 |
|
Keeping VRM temperatures down can be achieved by using more of them. Waste heat is a square function of current.
|
# ? Nov 23, 2021 01:20 |
|
Cygni posted:BX6/BH6/BX6 Rev 2 are legendary, baby. Ohhh yeah Shuttle was another favourite of mine, they had some great SFF stuff iirc.
|
# ? Nov 23, 2021 05:02 |
|
Khorne posted:ASRock has the best bios in some ways. They allow ECC on pretty much anything. They also allow headless boot on pretty much everything while with most other vendors it's hit or miss or you have to enable fast boot and pray it stays on. ...I don't think I've seen a system -- DIY or otherwise -- since the introduction of USB (the final death of KEYBOARD FAULT boot messages) that wouldn't boot headless. Where are you finding things that still care if they have something attached to a video-out?
|
# ? Nov 23, 2021 06:15 |
|
Rinkles posted:Does this apply to memory as well? Using XMP profiles, not necessarily custom OCing. XMP is interesting because the overvoltage isn't where you might think. Most ram can actually take super high vDIMM, especially if it's the Bdie or DJR that has xmp profiles at 1.45-1.5v. Increasing your vDIMM isn't going to harm your CPU, as those high voltages aren't actually touching CPU silicon. The exception to this is past 1.65v vDIMM where the ratio of it that the CPU gets hits Intel spec, but you're very unlikely to be going that high on a daily overclock. The issue is with VCCSA/VCCIO voltages, which are needed to be increased to run higher ram speeds and tighter voltages. These are actually on the CPU and cause degradation at very high values. Because these are usually at "AUTO" in the BIOS and no-one checks after boot, motherboard makers will juice them like crazy when you turn on XMP to absolutely ensure its stable. For example, on my board, XMP 4000 sets both VCCSA/VCCIO to 1.36+ which is super high. So if you have a high end kit of memory and enable XMP you want to manually set those values to sensible amounts. I personally run 3900c14 with insanely tightened subtimings, which requires 1.52v vDIMM, 1.2v VCCIO, 1.3v VCCSA. Despite being way faster, this is a safer overrall voltage setup than the default XMP 4000c17 1.35v vDIMM 1.36v/1.40v VCCIO/VCCSA. Plus I'm looking at 34.5ns vs 52ns memory latency.
|
# ? Nov 23, 2021 08:56 |
|
mdxi posted:...I don't think I've seen a system -- DIY or otherwise -- since the introduction of USB (the final death of KEYBOARD FAULT boot messages) that wouldn't boot headless. Where are you finding things that still care if they have something attached to a video-out? MSI might have updated bios to allow it recently. ASRock lets you boot with no gpu at all. Asus supposedly lets you do it too on most boards.
|
# ? Nov 23, 2021 15:10 |
|
Ant Online through ebay has the 5800x for $315 https://www.ebay.com/itm/124628149567
|
# ? Nov 23, 2021 16:14 |
|
Khorne posted:Gigabyte & MSI AM4 boards generally won't boot without a GPU on AM4. You don't have to have anything attached to video out. You have to turn on fast boot with a GPU inserted to get them to do it. Ah, I misunderstood what you meant by headless. Gotcha.
|
# ? Nov 23, 2021 16:53 |
|
BurritoJustice posted:
Setting Vref above the absolute maximum rating including overshoot is probably not doing good things to the life of those modules though. https://www.samsung.com/semiconductor/global.semi/file/resource/2017/11/DS_K4A8G165WB_BC_I_Rev1_6-1.pdf Page 11 for max ratings, 14 for overshoot calc.
|
# ? Nov 23, 2021 17:07 |
|
mdxi posted:Ah, I misunderstood what you meant by headless. Gotcha.
|
# ? Nov 23, 2021 18:34 |
|
dex_sda posted:The goddamn chipset fan on my x570 gigabyte elite is wearing down, loud noises all around. I can't believe such a tiny thing is the problem. What do? I got really unlucky with my X570 mobo and have a dead chipset fan and bizarre memory issues (which I thought were the fault of the CPU until I got Zen 3 and a different mobo). I actually set the curve to turn the fan off entirely and never have had any issues.
|
# ? Nov 23, 2021 21:58 |
|
Confirmation from a Linux kernel patch that 12 CCD CPUs are on the way: https://www.phoronix.com/scan.php?page=news_item&px=AMD-Zen-12-CCDs-hwmon
|
# ? Nov 24, 2021 05:56 |
|
mdxi posted:Confirmation from a Linux kernel patch that 12 CCD CPUs are on the way: Makes sense, should be 96 core Genoa
|
# ? Nov 24, 2021 06:07 |
|
|
# ? May 22, 2024 13:52 |
|
looks like the Ryzen 6900 is going to be nice
|
# ? Nov 24, 2021 06:18 |