Asus TUF Gaming B550M-Plus question regarding M.2 sockets and PCIe sockets

Dexter007

Limp Gawd
Joined
Mar 10, 2011
Messages
147
I'm looking at either the Asus TUF Gaming B550M-Plus or MSI MAG B550M Mortar motherboards for my next build, but I have a question regarding having 2 m.2 drives installed. The MSI board user manual says that if you have a m.2 drive installed in the second (PCIe 3.0) socket, it disables the bottom (full-length PCIe) socket. That's not ideal for me because I will need to have an audio card installed in one of the PCIe sockets, along with a GPU in the primary PCIe socket. I don't know that with the GPU I'll be able to install the audio card in anything but the bottom socket. If that's the case then I wouldn't be able to have 2 m.2 drives installed.

I can't find any information on the Asus board though with 2 m.2 drives installed. The user manual doesn't say anything about a PCIe socket being disabled when the second m.2 socket is being used. Should I assume nothing gets disabled when both m.2 sockets are occupied? Seemed like I read that it is common for B550 chipset that using both m.2 sockets disables something.

Basically, I'm looking to have a GPU (such as RTX 3080 or 6800XT) with an audio card and 2 m.2 drives on a B550M (micro-ATX board). The 2 boards listed above are my top choices at the moment. Can I make this work?
 
This particular ASUS board doesnt disable anything with the 2nd M.2 slot is used, since the motherboard specs page doesnt have any asterisks when discussing the PCIE/M2 slots. You'll see this when comparing ASUS boards that do disable stuff when the M2 is populated, see the below sample for the B550 Strix E/F:

*Share bandwidth with PCIe3.0 x1_1, PCIe3.0 x1_2
** When the M.2_2 Socket 3 is populated , SATA6G_5/6 ports will be disabled.

Edit: Didn't check your MSI board but if that's what you read, then that's what it is XD
 
Last edited:
Yeah. That’s why I skipped that one. Need a 10g card I. There. The x570s avoid that generally.
 
I'm looking at either the Asus TUF Gaming B550M-Plus or MSI MAG B550M Mortar motherboards for my next build, but I have a question regarding having 2 m.2 drives installed. The MSI board user manual says that if you have a m.2 drive installed in the second (PCIe 3.0) socket, it disables the bottom (full-length PCIe) socket. That's not ideal for me because I will need to have an audio card installed in one of the PCIe sockets, along with a GPU in the primary PCIe socket. I don't know that with the GPU I'll be able to install the audio card in anything but the bottom socket. If that's the case then I wouldn't be able to have 2 m.2 drives installed.

I can't find any information on the Asus board though with 2 m.2 drives installed. The user manual doesn't say anything about a PCIe socket being disabled when the second m.2 socket is being used. Should I assume nothing gets disabled when both m.2 sockets are occupied? Seemed like I read that it is common for B550 chipset that using both m.2 sockets disables something.

Basically, I'm looking to have a GPU (such as RTX 3080 or 6800XT) with an audio card and 2 m.2 drives on a B550M (micro-ATX board). The 2 boards listed above are my top choices at the moment. Can I make this work?

I know this post is getting old slowly but since both Mobos are still more than relevant, i joined to share my experience, since a couple of months ago i had the exact same dilemma and i cannot find anywhere on the net a relevant post or thread with regards to what i stumbled on.

Between the two i chose the TUF GAMING B550M PLUS that has no switching functionality of the type "for this to work you will not have the other". Since GPU prices were skyrocketing (and still are) i decided to keep my GTX 1080 and go a tad wilder on the other components. Went for RYZEN 3700X, GSKILL TRIDENT Z NEO 3600 GTZNC, SAMSUNG 980 PRO 500 GB on M2_1 slot and SAMSUNG 970 EVO 1TB on M2_2 slot, Kept my SOUNDBLASTER ZX on PCIEXx16_2 and also a SAMSUNG 870 EVO PLUS 1Tb on SATA6G_1. Running on windows 10 64b v1909. On BIOS 2006. All working beautifully together without single glitch or blue screen since day 1. Or do they? ...

As advertised ASUS TUF GAMING B550M PLUS gives all lanes on all PCIEX & M2 ports (PCIEX16_1 16 lanes 4.0 spec from CPU , PCIEX16_2 4 lanes 3.0 spec From Chipset, PCIEX1_1 1 lane 3.0 spec from Chipset and M.2_1 4 lanes 4.0 spec from CPU, M.2_2 4 lanes 3.0 spec from Chipset. Each SUPOSEDLY giving the full bandwidth of its respective generational specification without turning off anything or sharing bandwidth! (if if a port shares bandwidth with another it must ALSO be mentioned in the MB specifications in the manual, nothing was mentioned on the site or the manual!).

I started benchmarking after the build and all was well overall and i was very happy. Wen i benchmarked the M.2 SSDs the 980 pro gave 6850 read / 5015 write, PERFECT! And the 970 Evo Plus gave 3251 read / 2552 write ... what? 980 right on the spot and 970 evo plus acting as 970 evo speed wise. To make a long story short after a lot of troubleshooting and a lot of wise guys been heavily opinionated about m.2 ssd speed and how full it is and it does not work the same for all builds, my mind set as a buyer was I DONT CARE! i want what i bought (with a small margin of error due to the variety of builds) and not this piece of s**t results based on what was advertised. I may never actually utilize those speeds nonetheless my money bought your advertisement and i want to have what i bought. Back to the problem now (sorry for the rant) after a lot of troubleshooting and testing i finally found out that when i removed the soundcard from PCIEX16_2 (also tested with Sound card on PCIEX1_1) leaving vacant both PCIEX16_2 & PCIEX1_1 slots vacant the 970 evo plus speeds were getting back to a normal 3450 read / 3310 write. When populating either PCIEX slot the M2_2 970 Evo Plus drive was falling again on 3250/2500.

Whos is to blame for this? All slots are supposed to have the bandwidth of their respective lanes and generational Connection. ASUS omitted to mention that there is bandwidth sharing somewhere between slots? AMD has screwed up something in the chipset? Samsung has somehow messed the SSD firmware? Do you have any ideas for this situation? Cause someone has screwed up somewhere somehow and i don't know who is to blame or how to fix it !

Thank you in advance for reading, any ideas welcome!
 
Limitations of the chipset and CPU. If you want enough lanes to run everything at full speed all the time, buy Threadripper.

Real world you’re never going to know.
 
I started benchmarking after the build and all was well overall and i was very happy. Wen i benchmarked the M.2 SSDs the 980 pro gave 6850 read / 5015 write, PERFECT! And the 970 Evo Plus gave 3251 read / 2552 write ... what? 980 right on the spot and 970 evo plus acting as 970 evo speed wise. To make a long story short after a lot of troubleshooting and a lot of wise guys been heavily opinionated about m.2 ssd speed and how full it is and it does not work the same for all builds, my mind set as a buyer was I DONT CARE! i want what i bought (with a small margin of error due to the variety of builds) and not this piece of s**t results based on what was advertised. I may never actually utilize those speeds nonetheless my money bought your advertisement and i want to have what i bought. Back to the problem now (sorry for the rant) after a lot of troubleshooting and testing i finally found out that when i removed the soundcard from PCIEX16_2 (also tested with Sound card on PCIEX1_1) leaving vacant both PCIEX16_2 & PCIEX1_1 slots vacant the 970 evo plus speeds were getting back to a normal 3450 read / 3310 write. When populating either PCIEX slot the M2_2 970 Evo Plus drive was falling again on 3250/2500.

I have also noticed this on my X570 build. Not so much for the drive benchmarks, but for my Adobe Premiere benchmark testing. It turned out that most of the mainstream motherboards simply piled on so many onboard features that adding even two PCIe x1 cards into their corresponding slots would result in bandwidth sharing within the chipset. This is why my results with the Ryzen 7 3800X in my ATX X570 build were actually worse than that of my Ryzen 7 3700X in my current mini ITX B550 build (which has since been upgraded to a Ryzen 9 5900X). There simply isn't enough available bandwidth within the chipset to handle the expansion card load.
 
lopoetve & E4g1e I agree with you both except that B550 together with Ryzen 3700x have enough lanes to work those components up to full specs. Knowing this situation i described on my previous post from the get go i would probably again chose the TUF. Now it feels to me that MSI with its Mortar was far more straightforward to the consumers by having less USB ports and disabling 16_2 slot when populating m.2_2 seems to me now an honest try to deliver what was advertised. I can not know for sure (i can not test it so i only presume) but i believe i would get much better results for my ssd's with the same build using the Mortar at the expense of some connectivity, the difference is that the expense of that connectivity was already made known by the MSI ! ASUS having advertised full connectivity without noting any compromises in its specs and advertisements looks to me now they hoped that for this mid to low range board none would go really high end on any other components that would require full bandwidth from each port, so they did not annotate anything anywhere and thought they would get away with it. It feels to me like one of this new age Marketing "strategies" so to say, devised by 30 year old fresh out of college kids who presume to know everything and end up discrediting the company they work for. So to wrap this up, knowing by ASUS making this information available to me from the get go , i would probably again buy the ASUS between the two and i would cheap out on the second M.2 ssd buying a slower one with more capacity, Now that i got the feeling i was a bit "mislead" ... for my next build ASUS is going to get ... far back in my que.
 
Last edited:
Your later slots and later NVMe spots split the x4 link from the chipset back to the processor - and shares that with a bunch of other things too (likely sata controllers, possibly USB controllers, etc). That's even true on X570 - on Ryzen, of the 24 lanes, 4 go to the first NVMe, 4 to the chipset, and the remaining 16 are split between (hopefully) the top two slots (odd things do happen there). Board makers can choose to disable slots or get creative (PLX chips), but you're still limited by that x4 link from chipset to processor that is shared with many other things (the chipset has its own set of PCIE lanes out of it; I want to say it's 20 or 24 total). All consumer boards are this way right now - my Z490 is the same way... Hell, my X99 disables/enables certain slots based on what CPU you have!

Your benchmark results are fine - you're not going to see a limitation in the real world unless your workload is literally copying large files from one NVMe drive to another - small files even will have more overhead from IOPS/fileops than the bandwidth limitation.

Basically, relax - it doesn't really matter, which is why most people ~don't~ buy Threadripper/x299. If you really want every slot to run at full power, you need HEDT - and that comes at a cost.

As for them misleading you; I strongly suspect it's in the manual or the detailed tech specs on the webpage. You even said they advertised it as splitting with the chipset - there's a lot of stuff going over that link, and on B550, that's PCIE3 instead of PCIE4.
 
Your later slots and later NVMe spots split the x4 link from the chipset back to the processor - and shares that with a bunch of other things too (likely sata controllers, possibly USB controllers, etc). That's even true on X570 - on Ryzen, of the 24 lanes, 4 go to the first NVMe, 4 to the chipset, and the remaining 16 are split between (hopefully) the top two slots (odd things do happen there). Board makers can choose to disable slots or get creative (PLX chips), but you're still limited by that x4 link from chipset to processor that is shared with many other things (the chipset has its own set of PCIE lanes out of it; I want to say it's 20 or 24 total). All consumer boards are this way right now - my Z490 is the same way... Hell, my X99 disables/enables certain slots based on what CPU you have!

Yes i understand all you say and i agree. Actually ASUS did a good Job managing all lanes & links together added with bunch of features (for a B550 chipset i mean). Having unpopulated PCE1_1, PCIE16_2 actually gives full bandwidth (or very close to it) to both m.2 slots and usb & sata slots. I tested a lot to verify.

Your benchmark results are fine - you're not going to see a limitation in the real world unless your workload is literally copying large files from one NVMe drive to another - small files even will have more overhead from IOPS/fileops than the bandwidth limitation.

Basically, relax - it doesn't really matter, which is why most people ~don't~ buy Threadripper/x299. If you really want every slot to run at full power, you need HEDT - and that comes at a cost.

Yes i mentioned that already, benches are fine either way and i will probably never need all that bandwidth, it works since day 1 without a glitch or blue screen. Its the principle of the thing that bothers me as i have described. Generally i am happy.

As for them misleading you; I strongly suspect it's in the manual or the detailed tech specs on the webpage. You even said they advertised it as splitting with the chipset - there's a lot of stuff going over that link, and on B550, that's PCIE3 instead of PCIE4.

NO its NOT in the manual & its NOT in the detailed specs of their website. AND THAT ITS WERE MY PROBLEM IS !, i am not one of them bozos buying 1tb drive and then go back to the store and make a fuzz because they are only able to see and use 931gb. But i strongly believe that when you showcase a product you must describe everything and all functionalities about it in you specs even "hidden" ones, you don't expect from the consumer to derive functionalities based on the general design of a particular technology used in your product, because as you said very correctly "a lot of going on" between components and the way each manufacturer handles connections in his design.

This time ASUS haven't been straight forward about this situation, please fell free to direct me if i missed something in their site or manual, i would really appreciate it, because all of this would then be my bad missing it, and actually it would make me feel more happy with my build than i feel right now (I do feel happy though).
 
Last edited:
In my comparison, bandwidth is not the same as throughput. They are different things.
 
Yes i understand all you say and i agree. Actually ASUS did a good Job managing all lanes & links together added with bunch of features (for a B550 chipset i mean). Having unpopulated PCE1_1, PCIE16_2 actually gives full bandwidth (or very close to it) to both m.2 slots and usb & sata slots. I tested a lot to verify.



Yes i mentioned that already, benches are fine either way and i will probably never need all that bandwidth, it works since day 1 without a glitch or blue screen. Its the principle of the thing that bothers me as i have described. Generally i am happy.



NO its NOT in the manual & its NOT in the detailed specs of their website. AND THAT ITS WERE MY PROBLEM IS !, i am not one of them bozos buying 1tb drive and then go back to the store and make a fuzz because they are only able to see and use 931gb. But i strongly believe that when you showcase a product you must describe everything and all functionalities about it in you specs even "hidden" ones, you don't expect from the consumer to derive functionalities based on the general design of a particular technology used in your product, because as you said very correctly "a lot of going on" between components and the way each manufacturer handles connections in his design.

This time ASUS haven't been straight forward about this situation, please fell free to direct me if i missed something in their site or manual, i would really appreciate it, because all of this would then be my bad missing it, and actually it would make me feel more happy with my build than i feel right now (I do feel happy though).
https://dlcdnets.asus.com/pub/ASUS/...US/E17265_TUF_GAMING_B550M-PLUS_UM_v2_WEB.pdf -
Page 1-2 details what happens with different M2 form factors, which is based on the PCIE limitations, although they don't directly call it out super clear.

https://www.asus.com/Motherboards-Components/Motherboards/All-series/TUF-GAMING-B550M-PLUS/techspec/
Note the PCIE expansion section. It's kinda annoying how they do it, but one slot from CPU, two from chipset (x16 at x4 and x1).

Under storage, the second M2 slot is clearly from the chipset as well (as are all SATA ports). Same design, which is what MSI and Gigabyte do too (in terms of webpage layout).

Asus does NOT publish a block diagram that I've found, but for this level of board, that's not unexpected - Gigabyte is better about that, as is MSI (sometimes) - but mostly for high-end boards.

To put it simply, there is not a consumer motherboard that allows all slots at full speed anymore. Sucks, but it's true - PLX chips are expensive, and no one uses all the slots anymore unless you're odd (or me, but that's a different story). No one really uses sound cards now, or NICs, or anything that would go in an expansion slot - except for a GPU. So they stopped caring. If you care, you end up buying HEDT - which is why I have two different Threadripper setups and two different Intel Extreme Edition builds too, for some of the odd stuff I do.
 
In my comparison, bandwidth is not the same as throughput. They are different things.
My case concerns bandwidth and not actual throughoutput., PCIEX 3.0 with x4 lanes has a max bandwidth of about 3938 mb/s. 970 evo plus has a max read/write bandwidth & measured throughoutput at a very specific scenario of 3500/3300 mb/s. Maximum bandwidth is a constant number, throughoutput (measured speed during transfers) varies and is dependent on a lot of things and transfer/usage scenarios and has as cap the bandwidth. My problem is that m.2_2 slot losses Bandwidth and that means that the 970 throughoutput cap is at a much less max bandwidth than its rated, and that can only mean that at my certain setup scenario, m2_2 loses to or shares lanes with the populated pciex1_1 & pciex16_2 (unpopulate both and m2_2 bandwidth goes back capping to the Evo Plus normal max bandwidth). THAT should have been mentioned in the manual clearly with something like this: *M2_2 Slot Shares Bandwidth With PCIEX16_2 & PCIEX1_1. (even that would not be entirely accurate/correct but would serve its purpose).
 
Last edited:
My case concerns bandwidth and not actual throughoutput., PCIEX 3.0 with x4 lanes has a max bandwidth of about 3938 mb/s. 970 evo plus has a max read/write bandwidth & measured throughoutput at a very specific scenario of 3500/3300 mb/s. Maximum bandwidth is a constant number, throughoutput (measured speed during transfers) varies and is dependent on a lot of things and transfer/usage scenarios and has as cap the bandwidth. My problem is that m.2_2 slot losses Bandwidth and that means that the 970 throughoutput cap is at a much less max bandwidth than its rated, and that can only mean that at my certain setup scenario, m2_2 loses to or shares lanes with the populated pciex1_1 & pciex16_2 (unpopulate both and m2_2 bandwidth goes back capping to the Evo Plus normal max bandwidth). THAT should have been mentioned in the manual clearly with something like this: *M2_2 Slot Shares Bandwidth With PCIEX16_2 & PCIEX1_1. (even that would not be entirely accurate/correct but would serve its purpose).
That’s every single motherboard that is made now for the consumer market: B550, z490, z590, x570, etc. it’s somewhat assumed that is common knowledge now.
The only time that isn’t true is HEDT; and that’s not a given there either. That’s why they say those slots come from the chipset- anything powered by that is going to be sharing bandwidth with something.
 
https://dlcdnets.asus.com/pub/ASUS/...US/E17265_TUF_GAMING_B550M-PLUS_UM_v2_WEB.pdf -
Page 1-2 details what happens with different M2 form factors, which is based on the PCIE limitations, although they don't directly call it out super clear.

Says nothing there, what described with the usage of the table is when PCIEX16_1 is used with bifurcation mode on and an multiple m.2 ssd adapter card they call "hyper m2" and all this is off the CPU lanes anyway.

https://www.asus.com/Motherboards-Components/Motherboards/All-series/TUF-GAMING-B550M-PLUS/techspec/
Note the PCIE expansion section. It's kinda annoying how they do it, but one slot from CPU, two from chipset (x16 at x4 and x1).

This configuration fit my needs for what i do

Under storage, the second M2 slot is clearly from the chipset as well (as are all SATA ports). Same design, which is what MSI and Gigabyte do too (in terms of webpage layout).

Asus does NOT publish a block diagram that I've found, but for this level of board, that's not unexpected - Gigabyte is better about that, as is MSI (sometimes) - but mostly for high-end boards.

That was all known and clear from the beginning but it does not address what i "complain" about.

To put it simply, there is not a consumer motherboard that allows all slots at full speed anymore. Sucks, but it's true - PLX chips are expensive, and no one uses all the slots anymore unless you're odd (or me, but that's a different story). No one really uses sound cards now, or NICs, or anything that would go in an expansion slot - except for a GPU. So they stopped caring. If you care, you end up buying HEDT - which is why I have two different Threadripper setups and two different Intel Extreme Edition builds too, for some of the odd stuff I do.

In between our conversation i dig a bit around regarding available lanes of the CPU and B550 chipset and various setups from manufacturers. From what i can make out. The configuration ASUS used falls short 1 or 2 lanes (since there is no block diagram i made a few presumptions here i may very well be wrong) from the chipset to work everything together simultaneously at rated bandwidths (meaning m2_2, PCIEX1_1, PCIEX16_2) . So i am somewhat confident that, for the reasons you describe above, ASUS decided to not mention anything and let it work to their benefit that they have a board advertised in witch all works simultaneously without loosing something. That been the case then lane sharing exists and they should have added an asterisk of the type *M2_2 Slot Shares Bandwidth with PCIEX16_2 and/or PCIEX1_1 when those are populated. and then they would be fair gentlemen. Probably let it on the hands of Junior Marketing squad and ... there you have it.
 
That’s every single motherboard that is made now for the consumer market: B550, z490, z590, x570, etc. it’s somewhat assumed that is common knowledge now.
The only time that isn’t true is HEDT; and that’s not a given there either. That’s why they say those slots come from the chipset- anything powered by that is going to be sharing bandwidth with something.

Yes i get that now but i still think it should have been mentioned Cleary for everyone's perusal when reviewing parts for a new build (To let you understand why i insist on this i have to tell you that my last build was done with a 4960x on an x79 board back on 2012, very different era, with very different priorities on behalf of companies towards the consumers)

Thank you for the very nice conversation!
 
Yes i get that now but i still think it should have been mentioned Cleary for everyone's perusal when reviewing parts for a new build (To let you understand why i insist on this i have to tell you that my last build was done with a 4960x on an x79 board back on 2012, very different era, with very different priorities on behalf of companies towards the consumers)

Thank you for the very nice conversation!
Those boards tended to use PLX chips; they were more limited in some ways, but did PCIE switching differently. That tech is owned by Broadcom now and much more expensive than it used to be. Super micro is rolling out a board with one here soon; it’ll be interesting to see if that changes things again - but we just don’t use add in cards anymore.
 
Says nothing there, what described with the usage of the table is when PCIEX16_1 is used with bifurcation mode on and an multiple m.2 ssd adapter card they call "hyper m2" and all this is off the CPU lanes anyway.



This configuration fit my needs for what i do



That was all known and clear from the beginning but it does not address what i "complain" about.



In between our conversation i dig a bit around regarding available lanes of the CPU and B550 chipset and various setups from manufacturers. From what i can make out. The configuration ASUS used falls short 1 or 2 lanes (since there is no block diagram i made a few presumptions here i may very well be wrong) from the chipset to work everything together simultaneously at rated bandwidths (meaning m2_2, PCIEX1_1, PCIEX16_2) . So i am somewhat confident that, for the reasons you describe above, ASUS decided to not mention anything and let it work to their benefit that they have a board advertised in witch all works simultaneously without loosing something. That been the case then lane sharing exists and they should have added an asterisk of the type *M2_2 Slot Shares Bandwidth with PCIEX16_2 and/or PCIEX1_1 when those are populated. and then they would be fair gentlemen. Probably let it on the hands of Junior Marketing squad and ... there you have it.
That’s the thing. They don’t say that because EVERY motherboard does that. It’s the same as saying “we used silicon to make this”. Since there is no PLX chip, the chipset has to send everything back to the CPU- can’t get creative with what runs off of the cpu, it’s hard coded.
You can’t even run two PCIE slots off the cpu at full speed- they drop to 8x.
 
Oh maybe I misunderstanding you. Are you complaining that the first slot is powered by the CPU, and they are not saying that the second slot is powered by the chipset?
 
lopoetve for msg#17: I used the X79 chipset era from 10 years ago because back then the MB manuals and specifications were far more detailed and every aspect was covered. I get now that regarding current chipsets somethings are considered as "common knowledge" so some functionalities are not mentioned in detail or not at all. I still think its not fair. In some countries on the side mirrors of a car is printed "things are closer than they appear to be" for driver that is an obvious statement but its there any way. I wanted something similar from the ASUS mobo manual in specifications.

lopoetve for msg#18: Yes i get it that now this is "common knowledge" but i still think it should have been mentioned.

lopoetve for msg#19: No that is clear from the manual and web side specifications. What was not clear to me and i would like to see it mentioned in the manual is that Chipset lanes with regards to m2_2 slot and PCIEX1_1 and 16_2 are shared when all slots are populated.
 
lopoetve for msg#17: I used the X79 chipset era from 10 years ago because back then the MB manuals and specifications were far more detailed and every aspect was covered. I get now that regarding current chipsets somethings are considered as "common knowledge" so some functionalities are not mentioned in detail or not at all. I still think its not fair. In some countries on the side mirrors of a car is printed "things are closer than they appear to be" for driver that is an obvious statement but its there any way. I wanted something similar from the ASUS mobo manual in specifications.

lopoetve for msg#18: Yes i get it that now this is "common knowledge" but i still think it should have been mentioned.

lopoetve for msg#19: No that is clear from the manual and web side specifications. What was not clear to me and i would like to see it mentioned in the manual is that Chipset lanes with regards to m2_2 slot and PCIEX1_1 and 16_2 are shared when all slots are populated.
But chipset lanes are always shared. They only have 4 lanes back to the cpu...? It’s always been that way. And I mean forever. Hell, the early NEAT chipsets had bandwidth limitations back to the CPU, pre DMA days.

as for your slots, the first one is CPU fed and doesn’t share with anything. Everything else is chipset fed. And the web page clearly says that the second x16 slot is only x4...?
 
But chipset lanes are always shared. They only have 4 lanes back to the cpu...? It’s always been that way. And I mean forever. Hell, the early NEAT chipsets had bandwidth limitations back to the CPU, pre DMA days.

Yes, but chipset to cpu lanes have nothing to do with what i argued. Maybe you have a lot in your mind because somewhere there seems like you missed my point. What is going now days with Chipset lanes (not chipset to cpu lanes) is clear for me now and what was happening then and now used to be mentioned in the manuals, apparently not any more. You helped to clear that. Thanks.

as for your slots, the first one is CPU fed and doesn’t share with anything. Everything else is chipset fed. And the web page clearly says that the second x16 slot is only x4...?

Yes 16_2 is 4x lanes of the chipset, i have not disputed that and i do not understand why you mention it.
 
Ok, color me confused - you're looking for them to put "not all PCIE slots/NVMe slots have full bandwidth" in the manual?
lopoetve for msg#17: I used the X79 chipset era from 10 years ago because back then the MB manuals and specifications were far more detailed and every aspect was covered. I get now that regarding current chipsets somethings are considered as "common knowledge" so some functionalities are not mentioned in detail or not at all. I still think its not fair. In some countries on the side mirrors of a car is printed "things are closer than they appear to be" for driver that is an obvious statement but its there any way. I wanted something similar from the ASUS mobo manual in specifications.
x79 is HEDT; it had to spell that out because slot availability was tied partially to CPU (I believe, it certainly is on X99) - different CPUs had different lane capabilities, so every slot was wired for more than a base CPU would provide. You HAD to spell that out, or there was no way to know what CPU would enable what slots and at what speed - and you had more than the base 3 slots. I'm still confused why you would need that (or want that) on a consumer grade board, especially a mid-range one like B550... :confused:

I think I understand you now, but honestly, I figured it was just because everyone ~knows~ that - there's not a single consumer board where it matters, because they're all fixed slots now (minus ATX where it's either x16/x0 or x8/x8, but those are your only two options) - again, x299/x399/TRX40 are your exception, because those are HEDT and have all that spelled out in their manuals since it depends on CPU installed (and a dozen other options, in fact - I can pic SATA ports or NVME slots or a lot of other things on my TRX40 board)

I checked both my Z490 board manual, my x570 manual, and my x470 manual - none of them say anything about it because it's a given on that chipset... I think you're asking for them to put the CPU/Chipset spec in the manual?
 
Ok, color me confused - you're looking for them to put "not all PCIE slots/NVMe slots have full bandwidth" in the manual?

x79 is HEDT; it had to spell that out because slot availability was tied partially to CPU (I believe, it certainly is on X99) - different CPUs had different lane capabilities, so every slot was wired for more than a base CPU would provide. You HAD to spell that out, or there was no way to know what CPU would enable what slots and at what speed - and you had more than the base 3 slots. I'm still confused why you would need that (or want that) on a consumer grade board, especially a mid-range one like B550... :confused:

I think I understand you now, but honestly, I figured it was just because everyone ~knows~ that - there's not a single consumer board where it matters, because they're all fixed slots now (minus ATX where it's either x16/x0 or x8/x8, but those are your only two options) - again, x299/x399/TRX40 are your exception, because those are HEDT and have all that spelled out in their manuals since it depends on CPU installed (and a dozen other options, in fact - I can pic SATA ports or NVME slots or a lot of other things on my TRX40 board)

I checked both my Z490 board manual, my x570 manual, and my x470 manual - none of them say anything about it because it's a given on that chipset... I think you're asking for them to put the CPU/Chipset spec in the manual?

Look, this thread have taken a few odd turns here and there. I will go back to the beginning just to try and make it clear one last time ... with images.

Look the TUF and the MSI side by side and tell me witch of the two manufacturers handled specification memorialization in the manual better and who of the two left it a bit vague? (as my bench results in my message #5 clearly show) .

MSI.JPG


TUF2.JPG
 
I guess I never thought about it because 1, it's a B550 board, and 2, it's a consumer board - what would I ever put in those slots anyway? I see what you're saying now though.

edit: asus does tag that info in their X570/etc boards, at least the ATX ones, and on their HEDT ones where it tends to matter more too.
 
Last edited:
I guess I never thought about it because 1, it's a B550 board, and 2, it's a consumer board - what would I ever put in those slots anyway? I see what you're saying now though.

edit: asus does tag that info in their X570/etc boards, at least the ATX ones, and on their HEDT ones where it tends to matter more too.

When i chose between the MSI and the ASUS my thought was that on the ASUS i can have my SB Zx on one slot and a PCIx 1x NIC for my very old SAS on the other, and have the option of removing the NIC in the future and add a 3rd M.2 with adapter on the PCIEX16_2 4x. The MSI was leaving only 2 pcie1x free after m2_2 been occupied so the 3rd m.2 ssd with adapter was out of the question with the MSI. So I went with ASUS.

ASUS m2_2 benchmark without any of the PCIEX slots occupied (970 Evo Plus):

Capture2.JPG


ASUS m2_2 benchmark with either of the PCIEX slots occupied (970 Evo Plus):

Capture1.JPG


There is something going on there that is not mentioned in the manual and it should. I would prefer PCEX16_2 to work on x1 after m2_2 slot been used and have the full bandwidth of the m2_2 slot than the "crippling" of the m2_2 slot after PCIEX slots been occupied.
 
When i chose between the MSI and the ASUS my thought was that on the ASUS i can have my SB Zx on one slot and a PCIx 1x NIC for my very old SAS on the other, and have the option of removing the NIC in the future and add a 3rd M.2 with adapter on the PCIEX16_2 4x. The MSI was leaving only 2 pcie1x free after m2_2 been occupied so the 3rd m.2 ssd with adapter was out of the question with the MSI. So I went with ASUS.
Um... B550 only has 10 lanes. You were never going to get that, without a board that could effectively totally shut off the SATA controllers/etc. So you should have been looking at x570 to begin with, which has 20 (4 for CPU connection, 16 for other things). Sure, that should have been in the manual, but this is an entry-level board and chipset. If you're looking at 3x NVMe drives, you should be looking at X570 or something higher end. I get why you wanted it in the manual, but I ~also~ understand why they didn't say anything about it.

ASUS m2_2 benchmark without any of the PCIEX slots occupied (970 Evo Plus):

View attachment 357526

ASUS m2_2 benchmark with either of the PCIEX slots occupied (970 Evo Plus):

View attachment 357527

There is something going on there that is not mentioned in the manual and it should. I would prefer PCEX16_2 to work on x1 after m2_2 slot been used and have the full bandwidth of the m2_2 slot than the "crippling" of the m2_2 slot after PCIEX slots been occupied.
Look at the second set of numbers; you’re bandwidth limited because you have more crossing the chipset. The IOPS numbers are fine. that’s because there’s more going on with that card in there- commands are still working fine, you just don’t have all the bandwidth anymore... but that's not a difference you'll realize in the real world either, unless you're just copying data - and only large files at that. x570 might have helped since PCIE4 is used for the connection back to the CPU, but I've not really tested that, and I'm not cracking my x570 system open to tinker either.

Look, I get it - my argument here is that you started this out the wrong way. If you were planning multiple add-in cards and/or multiple NVMe drives, B550 was never the platform for you to start with - arguably it should be in the manual, but there was never a B550 board that was going to do what you wanted either. That's why they're supposed to be cheaper platforms.
 
Look, I get it - my argument here is that you started this out the wrong way. If you were planning multiple add-in cards and/or multiple NVMe drives, B550 was never the platform for you to start with - arguably it should be in the manual, but there was never a B550 board that was going to do what you wanted either. That's why they're supposed to be cheaper platforms.

Bottom line for me is that i can still have what i want from the ASUS than the MSI (i can still have the 3 m.2 together with my dedicated Sound card) just at lower speeds. I agree that anyway i would not use the full bandwidth that often so i would not justify going high end, and i also understand why they did not write anything in the manual. I still believe that they should and must mention these things no mater if its for low, middle or hi end products. Its all about letting the consumer (especially the not very well versed in components or the half knowledgeable as i am, know exactly what they are buying, even if it is considered common knowledge.

EDIT: If that info was written in the manual, what i would do differently, would be buying a lower speed, higher capacity M.2 for the M2_2 slot.
 
Back
Top