MSI Confirms X670 Motherboard Pricing, including MEG X670E GODLIKE for $1,299.99

[Tsing]

MSI has updated its website with the MSRPs of its AMD X670 motherboards for Ryzen 7000 Series processors.

Go to post

 

[Grimlakin]

I could see MAYBE going with the 479 model but even that hurts.

 

[Zarathustra]

Yeah, this is starting to be a bit much.

It doesn't seem THAT long ago that a top end motherboard was just north of $200...

Heck, a few years before that, the legendary Abit BP6 dual celery motherboard cost $145 brand new.

I love to blame all the useless crap the kids want these days (on board everything, RGB lighting and controllers, fancy looking cosmetic heatsinks, fancy boxes, paint schemes, etc, but as others have pointed out, chips are more expensive and traces are much more difficult for modern PCIe standards.

Still, I can't help but think that if there were a market for a clean all go, no show motherboard, and it could be sold in large enough numbers, it could be cheaper.

 

[Tempest]

Not even the ridiculously overpriced motherboards from any of the manufacturers seem likely to support certain basic capabilities of the CPU that I desire, which are inexpensive to implement (e.g., proper support for ECC memory).

I guess what I'm looking for is an entry-level "workstation", as opposed to a "gaming PC". I consider it a given that a general-purpose computer is capable of playing games whether or not it is has "gaming" explicitly mentioned in the marketing. Weird, isn't it?

Was considering a Ryzen 9 7950X a while back, but I'll probably sit this generation out as DDR5 and especially PCIe 5.0 are too immature at this stage.

 

[LazyGamer]

Not even the ridiculously overpriced motherboards from any of the manufacturers seem likely to support certain basic capabilities of the CPU that I desire, which are inexpensive to implement (e.g., proper support for ECC memory).

That's more up to Intel and AMD.

I guess what I'm looking for is an entry-level "workstation", as opposed to a "gaming PC". I consider it a given that a general-purpose computer is capable of playing games whether or not it is has "gaming" explicitly mentioned in the marketing.

So, a computer?

Aside from ECC - well, there's not much else that doesn't push you immediately into HEDT territory or Xeons / Epycs. If you need more RAM or more connectivity for storage or compute than the consumer parts offer, HEDT is what you're looking at regardless.

Was considering a Ryzen 9 7950X a while back, but I'll probably sit this generation out as DDR5 and especially PCIe 5.0 are too immature at this stage.

I'd tell you to look at Intel instead; Raptor Lake (13th-gen) is their second go at both DDR5 and PCIe 5.0, and looks to be turning out quite nicely.

Still, I can't help but think that if there were a market for a clean all go, no show motherboard, and it could be sold in large enough numbers, it could be cheaper.

That's your 'Pro' board in MSI's lineup. Main problem I have with it all is that you have to step up to a Carbon just to get a post-code readout, and an Ace if you want dual-BIOS capabilities, and on a first-gen release, you generally do want those if you're doing any tuning.

 

[Zarathustra]

That's more up to Intel and AMD.

It is, but most consumer Intel and AMD CPU's (at least the higher end ones) DO support ECC. It's just that most motherboards don't implement the feature in BIOS, and those that do, rarely document it properly, so it's a crap shoot to find one that works, and once you do, no guarantee it won't magically disappear after a BIOS update.

Same with IOMMU/VT-d support, and sometimes even PCIe bifurcation.

 

[LazyGamer]

It is, but most consumer Intel and AMD CPU's (at least the higher end ones) DO support ECC. It's just that most motherboards don't implement the feature in BIOS, and those that do, rarely document it properly, so it's a crap shoot to find one that works, and once you do, no guarantee it won't magically disappear after a BIOS update.

I'll give you an example - Intel released an i3 that had ECC, meant to be used in a consumer socket but with a Xeon chipset. Another, AMD only enabled ECC on their 'Pro' APUs, which they only sold to integrators.

Not having the feature enabled in CPUs means that there is essentially zero incentive for the feature to be supported on consumer boards.

Same with IOMMU/VT-d support, and sometimes even PCIe bifurcation.

I'm with you there; though from what I understand the limits are prevalent once you move to the mid-range products.

 

[Brian_B]

I would just like to point out, that even in this price range, the psychology of $0.99 still works, for some ungodly reason

 

[Tempest]

That's more up to Intel and AMD.

I don't expect motherboard manufacturers to support CPU features that aren't present. ECC was just an example, but support for such is present in the Ryzen CPUs in question. AMD has even publicly stated that they support ECC (albeit "unofficially") for previous generations, and some motherboards do implement the feature so that errors are properly reported to the OS.

Speaking of memory, I wonder how many people are aware that enabling the XMP profile (or whatever AMD's counterpart is called) in their factory-overclocked gaming memory voids their processor warranty. That is true for both Intel and AMD, and even applies to "AMD Certified" memory. Whether doing so is "safe" or advisable is another matter and a personal choice, but in my opinion the companies involved do a poor job of informing their customers of that fact.

So, a computer?

A stable, non-buggy computer, yes. It needn't be equipped with enough RGB LEDs to light up Times Square on Christmas, either. (There I go again, bringing up RGBs like a broken record...)

Aside from ECC - well, there's not much else that doesn't push you immediately into HEDT territory or Xeons / Epycs. If you need more RAM or more connectivity for storage or compute than the consumer parts offer, HEDT is what you're looking at regardless.

Zarathustra mentioned virtualization, another feature-set for which non-buggy support is often found wanting on consumer boards. Many of the problems boil down to firmware implementations. Intel and AMD provide proprietary binary blobs to firmware vendors, but UEFI is sufficiently complex that I couldn't say offhand who is responsible for what. By now the specification probably exceeds the length of OED. Regardless on who the blame falls, at some of the price points we've been seeing, I have zero tolerance for buggy firmware.

While I would certainly like to have more PCIe lanes, that's an example of something I obviously cannot expect from motherboard vendors.

I'd tell you to look at Intel instead; Raptor Lake (13th-gen) is their second go at both DDR5 and PCIe 5.0, and looks to be turning out quite nicely.

I plan on doing so, though I was kinda hoping to escape Intel this time around. I recall being mildly interested when they announced the W680 "chipset", but that never seemed to takeoff. Does Intel even make entry-level Xeon processors anymore? What happened to the Xeon-W processors for Alder Lake-S? I haven't paid much attention lately.

 

[LazyGamer]

I don't expect motherboard manufacturers to support CPU features that aren't present. ECC was just an example, but support for such is present in the Ryzen CPUs in question. AMD has even publicly stated that they support ECC (albeit "unofficially") for previous generations, and some motherboards do implement the feature so that errors are properly reported to the OS.

Present, but disabled by Intel and AMD themselves on specific SKUs. It'd be pretty nice if they just left it on and it became a standard feature, but since the implementation has a cost on the boards and the DIMMs, I don't expect much traction.

Speaking of memory, I wonder how many people are aware that enabling the XMP profile (or whatever AMD's counterpart is called) in their factory-overclocked gaming memory voids their processor warranty. That is true for both Intel and AMD, and even applies to "AMD Certified" memory. Whether doing so is "safe" or advisable is another matter and a personal choice, but in my opinion the companies involved do a poor job of informing their customers of that fact.

It's communicated, but generally not much of an issue. It's extremely rare to 'break' a CPU through normal enthusiast use, in my experience, and XMP / DOCP / EXPO aren't a primary contributor to CPUs actually breaking.

So yeah, it's an issue, but not a big one.

A stable, non-buggy computer, yes. It needn't be equipped with enough RGB LEDs to light up Times Square on Christmas, either. (There I go again, bringing up RGBs like a broken record...)

They can be turned off, and for the most part, the LEDs are on other things. GPUs, RAM, fans, etc. MSI has gone so far as to add a switch to turn off LEDs on and connected to some of their boards.

As for bugs / stability - this is always relative. Whatever cost there is for an unstable system should be represented in component selection and settings. Enthusiast parts are still useful for system stability as they tend to be overspec'd from a power delivery and stability standpoint, as well as from a cooling standpoint. Especially if you want stability and low noise!

Zarathustra mentioned virtualization, another feature-set for which non-buggy support is often found wanting on consumer boards. Many of the problems boil down to firmware implementations. Intel and AMD provide proprietary binary blobs to firmware vendors, but UEFI is sufficiently complex that I couldn't say offhand who is responsible for what. By now the specification probably exceeds the length of OED. Regardless on who the blame falls, at some of the price points we've been seeing, I have zero tolerance for buggy firmware.

Well, I use virtualization regularly, and again on mid-level and higher configurations, I haven't seen any real shortcomings here. Perhaps my experience is more recent though.

I plan on doing so, though I was kinda hoping to escape Intel this time around. I recall being mildly interested when they announced the W680 "chipset", but that never seemed to takeoff. Does Intel even make entry-level Xeon processors anymore? What happened to the Xeon-W processors for Alder Lake-S? I haven't paid much attention lately.

Well, AMD has abandoned their 'non-Pro' Threadrippers, while Intel hasn't updated their Xeons to the Alder Lake architecture yet, so we're more or less 'between' HEDT generations with no real indication where either company is going.

 

[THUMPer]

Luckily, we dont HAVE to buy this. I'm looking for a B650 or B650E MATX.

 

[Burticus]

Yeah, this is starting to be a bit much.

It doesn't seem THAT long ago that a top end motherboard was just north of $200...

Yeah $200 is my top end limit for motherboards and that would be a high end one. I usually aim for the 100-150 range for mainstream and even cheaper than that for bargain builds (ie - b350/450 boards etc)

 

[Tempest]

Present, but disabled by Intel and AMD themselves on specific SKUs. It'd be pretty nice if they just left it on and it became a standard feature, but since the implementation has a cost on the boards and the DIMMs, I don't expect much traction.

Well, I'm not giving motherboards a pass on that one, at least not the higher-priced ones and when they can afford to incorporate all that bling. The exception would be boards on which the feature is unable to be used regardless of CPU selection. Lack of demand is the main reason ECC UDIMMs are so expensive. Perhaps we should be encouraging the adoption of ECC?

It's communicated, but generally not much of an issue. It's extremely rare to 'break' a CPU through normal enthusiast use, in my experience, and XMP / DOCP / EXPO aren't a primary contributor to CPUs actually breaking.

I don't mind the extended memory profiles per se, but what bothers me is that the modules are advertised according to their factory-overclocked settings, and it requires some amount of detective work to find out the native speed of the memory chips, if that information is obtainable at all.

They can be turned off, and for the most part, the LEDs are on other things. GPUs, RAM, fans, etc. MSI has gone so far as to add a switch to turn off LEDs on and connected to some of their boards.

It would be nice to see a greater selection of RGB-free devices. To me they are signal that form has taken precedence to function. That doesn't seem unfounded in light of the exploit-ridden software device makers ship to control the lighting.

My complaints aren't really directed at MSI, but rather the PC hardware market as a whole. There doesn't seem to be much diversity left (pricing notwithstanding).

Well, AMD has abandoned their 'non-Pro' Threadrippers, while Intel hasn't updated their Xeons to the Alder Lake architecture yet, so we're more or less 'between' HEDT generations with no real indication where either company is going.

Thanks for the update on the Xeons. That explains why I haven't seen anything concrete on them except for old rumors.

Luckily, we dont HAVE to buy this. I'm looking for a B650 or B650E MATX.

Not ITX or mini-ITX? ;-) From what I've seen so far, the B650E seems like it might have the most appeal to me. It looks cleaner without all the superfluous crap on the dual "Promontory 21" variants, but I also want PCIe 5.0 wherever possible because I keep my machines in service for a very long time.

I'm probably going to skip this generation though.

 

[LazyGamer]

Well, I'm not giving motherboards a pass on that one, at least not the higher-priced ones and when they can afford to incorporate all that bling. The exception would be boards on which the feature is unable to be used regardless of CPU selection. Lack of demand is the main reason ECC UDIMMs are so expensive. Perhaps we should be encouraging the adoption of ECC?

I wouldn't mind it, but realistically - consumers don't need ECC on the desktop. That's where the lack of demand stems from in my opinion, because the real need for ECC really only exists in professional spaces. That's where HEDT should be, yet that space is currently underserved. You can pay up for a Threadripper Pro if that's what you really want.

I don't mind the extended memory profiles per se, but what bothers me is that the modules are advertised according to their factory-overclocked settings, and it requires some amount of detective work to find out the native speed of the memory chips, if that information is obtainable at all.

There's no 'native speed' for memory chips. There's the JEDEC speeds, which are included as profiles on DIMMs, there's what the CPU manufacturers state that is 'stock' for their memory controllers, and there's overclocked profiles beyond JEDEC, which are validated on a range of configurations but not guaranteed.

Another thing to consider is that JEDEC speeds are typically extremely conservative, but over time tend to advance with memory technology.

It would be nice to see a greater selection of RGB-free devices. To me they are signal that form has taken precedence to function. That doesn't seem unfounded in light of the exploit-ridden software device makers ship to control the lighting.

Check out MSI's Unify / Unify-X and ITX Unify boards. These exist!

For anything else - well, the best fans lack lighting, but I'll say that anything that doesn't have an easy way to turn the lighting off is a foul.

And I'll echo the concern about their software. Security is a big one, but beyond that, the massive fracturing of these software suites is also concerning. I've been running MSI's software for my board and it's not been terrible - and that because you can install only their RGB software. And it works.

Thanks for the update on the Xeons. That explains why I haven't seen anything concrete on them except for old rumors.

Xeons on new architectures tend to lag significantly behind consumer releases, so they're probably coming, if they're coming.

 

[Tempest]

I wouldn't mind it, but realistically - consumers don't need ECC on the desktop. That's where the lack of demand stems from in my opinion, because the real need for ECC really only exists in professional spaces. That's where HEDT should be, yet that space is currently underserved.

Well, I couldn't disagree more, and there are many who share that opinion, but we're obviously in the minority. There is of course Linus' (now infamous) rant on the subject with which I strongly agree, and I'm one to generally dismiss his public tirades. And even if one doesn't care about any of the traditionally argued benefits of ECC, there's also Rowhammer to consider. I don't think consumers should need to pay for $5000+ "workstations" to receive a basic feature such as a error detection that used to be standard.

You can pay up for a Threadripper Pro if that's what you really want.

I can buy an AM4 Ryzen system right now with working ECC support. You aren't going to change my mind on this matter.

There's no 'native speed' for memory chips. There's the JEDEC speeds, which are included as profiles on DIMMs, there's what the CPU manufacturers state that is 'stock' for their memory controllers, and there's overclocked profiles beyond JEDEC, which are validated on a range of configurations but not guaranteed.

"Native speed" isn't a technical term. I meant whatever rating the memory manufacturer assigns to the modules, which as far as I know corresponds to one of the JEDEC memory standards. Gaming memory often doesn't even provide the JEDEC profiles in the product information, which can be a major headache where compatibility with different systems matters.

And I'll echo the concern about their software. Security is a big one, but beyond that, the massive fracturing of these software suites is also concerning. I've been running MSI's software for my board and it's not been terrible - and that because you can install only their RGB software. And it works.

The day that I install any software to control RGB lighting...

 

[Riccochet]

I'll gladly pay $250-350 for a motherboard....but $1300, and it doesn't even have a full cover waterblock on it? GTFO. This thing should be set up to water cool everything on it, right out of the box.

Meth is a helluva drug. MSI smokes it.

 

[MadMummy76]

Member when that was a whole gaming PC?

 

[MadMummy76]

The day that I install any software to control RGB lighting...

Well, I install the software to neuter the lighting.

 

[LazyGamer]

"Native speed" isn't a technical term. I meant whatever rating the memory manufacturer assigns to the modules

That's the thing - they don't.

I can buy an AM4 Ryzen system right now with working ECC support. You aren't going to change my mind on this matter.

Have to be careful with that - I eventually just stopped looking because it was difficult to verify if ECC was actually 'working'. And if AMD had released an APU for consumers with guaranteed ECC support (and a suitable board was identified), I'd already be using it for a personal server.

I'll gladly pay $250-350 for a motherboard....but $1300, and it doesn't even have a full cover waterblock on it? GTFO. This thing should be set up to water cool everything on it, right out of the box.

Meth is a helluva drug. MSI smokes it.

EK is their partner for water-blocked boards, and they've done versions with the Torpedo and the Carbon for Z690. Both are pretty affordable considering what you're getting.

The 'Godlike' is in a different class altogether, and really targeted at a) those with too much money, b) those that will actually use most of the features (including those unavailable on other boards), and c), those trying to set benchmarking records.

But if you're looking for a true higher-end 'tuner's' board, look at MSI's Unify-X and ASUS' Apex lines, which are ATX boards with two DIMM slots.

 

[Tempest]

That's the thing - they don't.

I'm not quite sure what you mean. The settings programmed in my Samsung brand memory had to come from somewhere. Where do they come from if not Samsung? Micron also publishes guides to their module-numbering systems, which I've referenced in the past for Crucial/Micron brand memory. They include details such as speed bins and grades. I believe they can be found under "DDRx Module Part Numbering System": https://www.micron.com/numbering

Have to be careful with that - I eventually just stopped looking because it was difficult to verify if ECC was actually 'working'. And if AMD had released an APU for consumers with guaranteed ECC support (and a suitable board was identified), I'd already be using it for a personal server.

It's pretty easy to verify under Linux. There are kernel modules and tools to assist with that. But as you know, the pickings are slim in the "suitable board" department, which was part of my original complaint.