GPU-Z misreading Vega M bus width

[Brendan9140]

Hello, just a small issue with the detection of the bus width of the Vega M GL's HBM2, where for some reason GPU-Z reads a 256-bit interface for the 4GB HBM2 buffer, instead of the expected 1024-bit that the online database shows. I have the 19.8.1 drivers installed, but has also done it with older drivers.
Not sure if it's just me with this problem and/or if it does this with the GM variant of the chip.

 

[Keullo-e]

Have you tried other GPU-Z versions?

 

[Bill_Bright]

What does HWiNFO64 say?

 

[R-T-B]

What does HWiNFO64 say?

This site is the homepage for GPU-Z, thus if it is a problem with our product, reccomending another should not be considered an acceptable fix. I assume he wants this treated as a bug report.

It is in the wrong forum for that, though.

Pinging @W1zzard

 

[W1zzard]

Isn't Vega M actually Polaris? So 256-bit might be correct?

 

[Keullo-e]

Isn't Vega M actually Polaris? So 256-bit might be correct?

AMD Radeon RX Vega M GL Specs

AMD Polaris 22, 1011 MHz, 1280 Cores, 80 TMUs, 32 ROPs, 4096 MB HBM2, 700 MHz, 1024 bit

www.techpowerup.com

The bus is still 1024-bit

 

[W1zzard]

Yeah confirmed on ARK too https://ark.intel.com/content/www/u...ga-m-gl-graphics-8m-cache-up-to-4-10-ghz.html

 

[Bill_Bright]

This site is the homepage for GPU-Z, thus if it is a problem with our product, reccomending another should not be considered an acceptable fix. I

No one said anything about HWiNFO being "a fix". If your volt meter said your "working" 12V car battery was putting out 3V, would checking with a different multimeter to verify if that reading was correct be considered "a fix"?

And "our" product? I don't care whose product it is. There is no harm in verifying the results with a 2nd product.

As W1 has pointed out, it should be 1024. So is the OP's card reporting the wrong information? Or is GPU-z? At this point, we don't know. If a different HW monitor reports 256 as well, then we can assume the problem is with the card or GPU - not the monitoring software.

Verifying a reading through another means is just plain good troubleshooting.

 

[R-T-B]

No one said anything about HWiNFO being "a fix". If your volt meter said your "working" 12V car battery was putting out 3V, would checking with a different multimeter to verify if that reading was correct be considered "a fix"?

And "our" product? I don't care whose product it is. There is no harm in verifying the results with a 2nd product.

As W1 has pointed out, it should be 1024. So is the OP's card reporting the wrong information? Or is GPU-z? At this point, we don't know. If a different HW monitor reports 256 as well, then we can assume the problem is with the card or GPU - not the monitoring software.

Verifying a reading through another means is just plain good troubleshooting.

Then I misunderstood your intent with linking that. No need to to get worked up.

If your volt meter said your "working" 12V car battery was putting out 3V, would checking with a different multimeter to verify if that reading was correct be considered "a fix"?

Does nothing but verify the voltage, which we can do with ARK docs. It certainly does not fix the multimeter (GPU-Z) which is the point of a bug report.

I did note that it is in the wrong forum for a bug report, in your defense he may not even be asking it be treated as one.

You need to work a bit on assuming innocence in intentions. Most people here are not here to attack you or your credibility. Just saying what I observe. Feel free to take it or leave it.

 

[W1zzard]

As W1 has pointed out, it should be 1024. So is the OP's card reporting the wrong information? Or is GPU-z? At this point, we don't know. If a different HW monitor reports 256 as well, then we can assume the problem is with the card or GPU - not the monitoring software.

"w1z" or "wiz" please

i think the reason is that they hacked a different memory interface onto polaris, and the bus width is stored in a different place than with the old mc

 

[Bill_Bright]

"w1z" or "wiz" please

Will do! Thanks for making your preference known! That said, you are the #1 "W" I know (and Bill is short for "W"illiam! ).

No need to to get worked up.

I was not the one who got worked up. It seems it was you for me suggesting the OP "try" a different program. In fact, it seems you are still worked up about it.

Does nothing but verify the voltage, which we can do with ARK docs.

No you can't. Ark docs only states what Intel says it should be - which is important to know. However, an incorrect reading by one HW monitor program indicates something is wrong either with the processor or with the HW monitor program but does NOT show which is at fault. Checking with a 2nd HW program does (unless you have a second identical processor handy).

It is the same with graphics problems. Swapping in a 2nd monitor quickly shows if the problem is with the graphics solution (computer or card) or the original monitor.

 

[R-T-B]

What part of my post appears worked up bill? Believe me I'm not. Getting worked up makes my Wolff Parkinson's White do bad things to me.

Please pm me if you feel the need to continue, honestly this should've ended three posts ago. I apologize for my misinterpretation of your intent.

 

[Bill_Bright]

What part of my post appears worked up bill?

The mere fact you felt it necessary to actually join a thread to make a post just to chastise another for suggesting the OP try another program (something other than "our" program) that would quickly isolate the discrepancy to either the processor or GPU-z.

I apologize for my misinterpretation of your intent.

Then we are good to go. Thanks.

 

[R-T-B]

I mainly joined to ping w1zzard.

Glad it's settled.