Is SVI3 The Best Way To Measure AMD CPU Voltage?

Gordon had a chance to chat with Aris from @HardwareBusters at Computex 2023 about SVI3 and whether it's the best way to measure AMD CPU voltage.

Follow PCWorld for all things PC!
------------------------------­----

#amd #cpu #computex2023

[louisfriend]

Aris is the Greek God of PSU testing. Nobody does this better than him.

[techiegk]

Great talk. Thank you for the information and thanks to Aris for all of his testing

[JusticeGamingChannel]

This was a very informative video, it clears up what SVI3 is and why it matters when reporting on this specific issue. Thanks.

[kelownatechkid]

More content with Aris please!! His content is so underrated

[ForzaE]

I learned this info from another engineer and he confirmed him in this video. Thank you so much. I feel more safe now.

[Loppyu]

Cheers, didn't know diddly-squat about this SV12/ SV13 for AMD in HWiNFO. Dropping some knowledge gents.

[Starscreamious]

9:00 The one big problem with LGA is that after a few years the cost/value of the motherboards far exceeds the cost/value of the CPU.
ie after 5 years or so the motherboards increase in value while the cpu's decrease in value....LGA makes that a bit worse as people are prone to breaking those sockets.

[Starscreamious]

7:00 lol I think that's Aris' partner sneaking behind them there....pulling an Adam.

[VideogamesAsArt]

I actually find that the CPU being able to turn off VRMs is incredibly useful! More power efficiency like this please!

[TheGameBench]

Your own video on the MSi PSU with the yellow connector showed the issue that just making sure it's connected isn't enough. There's too much play on the connector and it needs to be revised with a better latching mechanism. Seeing this doesn't give me a lot of trust on 12VHPWR especially at the PSU side where it could easily become unseated, even though it's fully latched and burn up on the PSU side. I guess what MSi is doing is great on the GPU side where it's easy to see, but it's just a band aid. This clearly is NOT user error. It's just poorly designed and need a revision with better latching accounting for at least half to two thirds the total length of the connector, or latches on both sides, to prevent the connector from being able to move once it's latched.

Outside that... great info on the voltage side of things with AM5.

[themusesquad]

Not true it is the CONNECTOR If you can plug it out partly which cause it to melt while th lock is still locked that means there is a problem with the lock itself and it doesn't lock the connector at the end only halfway.

[retrosean]

Hmm this is quite interesting, nobody mentioned this before in any of the videos I saw about the exploding CPUs lol

[tacticalcenter]

The nvidia connectors are still melting regardless of being fully plugged in and also not being bent.

[farmertrueVR]

In HWInfo64, SVI3 is listed on many different items. There is CPU VDDCR_VDD Voltage (SVI3), CPU VDDCR_SOC Voltage (SVI3), CPU VDD_MISC Voltage (SVI3), CPU Core Current (SVI3), and 8 or so more. So which in HWInfo64 is the one we should be keeping an eye on for this? I don't see just a SVI3 anywhere.

I boot up HWInfo64 first on my pc. Windows loads, and then I'm opening HWInfo64. It is an amazing tool that all pc users should use IMO to monitor their pc and hardware, but it also has an overwhelming amount of info for the average joe.

As always, great content and thank you for sharing!

[marcmathes]

this guy is preaching what i tell people oc depends on wall power how clean a then psu then the vrm ...

[christopherjackson]

I'm not sure have as much trouble with accepting amds voltage measurement when the target voltage is given by amd.

In a sense what we're concerned with in this situation is a relative voltage measurement not an absolute one.

At least that's how I understand it, if I'm wrong someone pls correct me. 😊

[andrewleno]

Thanks for info guys. Can you please clarify SVI3 is relevant for all Ryzen CPU?
I have Ryzen R9 5950X + Asus X470 Pro with the latest BIOS update and latest drivers. I use HWiNFO64 v7.46-5110 and there is no SVI3 at all! Only SVI2 TFN.

[creatorart]

Siv system information viewer, is great app show every sensor on pc, it is been for years why not many knowing about it? very easy to read. the hwinfo is great but difficult to read all this readnings.

[azjeep]

I have asked is my 3900xt safe the x570 asus tuff motherboard goes to 1.4+ all the time i have tried to cap it at 1.35 and it wont

[johndoh]

Too much talking. I got through about 4 minutes without hearing very much, and for a person who learned how voltage delivery works to do my job and have understood it for 40 years now I got bored. Most everything said is trivial, at least up to the point I stopped watching. Too much about needing a good PSU. Sure, that's true. It has nothing to do with the topic of measuring CPU voltage.

It doesn't really matter than much where you measure CPU power consumption as long as it's ACTUAL power going to the CPU. It certainly doesn't matter enough to have Youtuber wars over it. Steve at GN does a perfectly fine job measuring at the power rail. The fact is very little power is lost through the regulators or VRMs, unless you have a really bad board with bad components and Steve doesn't use bad boards for a testing platform. His measurements are very close to Techpowerup The issue is being consistent so a comparison can be made.

A BAD way to measure would be to open up a PSU and measure the inputs to the transformer, which is the step down from one level of AC to lower level of AC which then feeds into a rectifier that does the initial conversion to DC. Once you eliminate the whole PSU from the power consumption values, you're close enough.

Power consumption is determined by the load, in this case the CPU. A load pulls the amount of power it needs from a source. A source can only deliver what's being pulled unless there's a problem in the circuit, so normal ops of power delivery Yes, having a good PSU matters, but it's not going to affect the readings for power going to a CPU except in a VERY trivial way, because once again it's up to the load, not the source on how much power is being delivered. So wherever this conversation was going I don't care. One set of VRMs will consume a SLIGHTLY different amount of power than another set, which is why testers want to use a quality MB when testing, and most MBs have a good quality set of VRMs. Being ABSOLUTELY PERFECT is just about impossible. Being within a couple or three watts (more when the consumption is higher, much closer to actual when the consumption is small unless once again you have cheap VRMs) is good enough, and including the small level of VRM power consumption is trivial.