Some entry level Z490 boards cannot handle an overclocked i5-10600K

[Mamonos]

Hardware Unboxed just released a video about VRM testing on some "entry level" Z490 motherboards.

 

The test features the i9-10900K in three configurations (stock, non power-limited, overclocked) and the i5-10600K (overclocked to 5.0 GHz).

 

The ASUS Prime Z490-P and the MSI Z490-A Pro performed very well in all scenarios, while the other boards (Gigabyte Z490 UD, ASRock Z490 Phantom Gaming 4 and ASRock Z490 Pro4) suffered from thermal throttling or even crashes in pretty much all scenarios.

 

Screens:

Spoiler

 

 

 

 

 

Personally is not a surprise for me that the ASUS board performed really well.

I am disappointed with the other boards, from a Z board I would expect it to be able to overclock at least the midrange chips, also a big fail considering that those are heavily marketed by their brands (see for example the screenshot from Asrock Z490 Pro4 below)

At the same time I think/wonder if it's another sign that Intel really went beyond the limits with the 14nm process.

 

 

[maartendc]

That is very disappointing, downright bad products.

 

For most products, bad OC'ing wouldn't necessarily be a problem.. But the MAJOR (really only) reason people buy the Z-chipset motherboards is to allow overclocking.

 

If the board cannot even handle a pretty standard overclock on a 6-core chip, the product is just not performing as you should expect it to.

 

Also considering the 'stock' turbo boost of a 10600K is 4.8 Ghz. So if it cannot handle a 5.0 Ghz overclock, perhaps (and this is speculation on my part) it also can't handle boosting for very long before running into thermal limitations of the motherboard.

[Jurrunio]

Pro4 and UD's VRM are just Z370 and Z390 versions with doublers, the number of mosfets powering the CPU cores hasnt changed. These results are no surprise to me.

[porina]

12 minutes ago, Jurrunio said:

Pro4 and UD's VRM are just Z370 and Z390 versions with doublers, the number of mosfets powering the CPU cores hasnt changed. These results are no surprise to me.

I'm running a 8086k on Z370 Pro4, so should be comparable to the 10600k on Z490 Pro4 then? I've had no problems with short term benching at beyond 5 GHz, however if I run a sustained compute load at stock for days, I sometimes get random crashes. They seem to go away if I turn off turbo and I kinda suspected the power delivery might not have been quite up to it.

[MageTank]

I am curious if ASRock ever fixed the LLC issues on their boards. When I was qualifying 10th gen boards a few months back, none of the LLC settings did anything across their entire product stack. You could manually change it from 1 to 5, all of the settings sagged. Leaving it on auto was the only way to achieve the vCore you dialed in, and even that was off by roughly 20mv (overshoot). The funny part is if you change voltage at all, it automatically selects LLC level 2, you have to manually go back and change it to auto.

 

I would hope they would have fixed it by now, but it would be interesting to see if that was part of the issue with their crashing. 

[5x5]

Leave it to board OEMs to make even a good product undesirable 

[Tan3l6]

ASrock has always been dubious. So no surprise there.

The highest of high end ASRock might be cool though.

[RejZoR]

5 hours ago, maartendc said:

That is very disappointing, downright bad products.

 

For most products, bad OC'ing wouldn't necessarily be a problem.. But the MAJOR (really only) reason people buy the Z-chipset motherboards is to allow overclocking.

 

If the board cannot even handle a pretty standard overclock on a 6-core chip, the product is just not performing as you should expect it to.

 

Also considering the 'stock' turbo boost of a 10600K is 4.8 Ghz. So if it cannot handle a 5.0 Ghz overclock, perhaps (and this is speculation on my part) it also can't handle boosting for very long before running into thermal limitations of the motherboard.

Problem with Intel turbo is that it's timed and it'll only run on certain number of cores. After that time, it'll drop to lower clock. If you do an all core turbo it'll run that turbo on ALL cores and that can kill crappy motherboards.

[Techstorm970]

Meh...sounds like a firmware issue.  Or something that a firmware update would fix, anyway.

 

This is why I don't rush to buy products right after they come out...

[Void Master]

7 hours ago, Mamonos said:

The ASUS Prime Z490-P and the MSI Z490-A Pro performed very well in all scenarios, while the other boards (Gigabyte Z490 UD, ASRock Z490 Phantom Gaming 4 and ASRock Z490 Pro4) suffered from thermal throttling or even crashes in pretty much all scenarios.

Reminds me of AM3 motherboards .... 

[PlayStation 2]

I was gonna make an AM3+ joke, but I just can't. Damn, that's horrifying.

[mr moose]

unless what techstorm said about firmware is the actual reason,  this highlights why Intel is not comfortable making their new processors work on old/cheap boards.

[GDRRiley]

4 hours ago, Tan3l6 said:

ASrock has always been dubious. So no surprise there.

The highest of high end ASRock might be cool though.

most of asrock boards are fine to great at the high end. they make one of the few x570 mATX boards.

 

[mr moose]

5 minutes ago, GDRRiley said:

most of asrock boards are fine to great at the high end. they make one of the few x570 mATX boards.

 

I have lots of Asrock boards, they have all been great.  this is the first time I have seen anything that suggests they skimped somewhere in the build. 

[Arika]

After watching a few GN videos about what motherboard manufacturers do to try and edge out the competition by breaking spec, this doesn't surprise me at all.

 

CPU manufacturers need to start raking these companies over coal to make them adhere to their specs set by the CPU, not just do what ever they want. 

[CarlBar]

27 minutes ago, Arika S said:

After watching a few GN videos about what motherboard manufacturers do to try and edge out the competition by breaking spec, this doesn't surprise me at all.

 

CPU manufacturers need to start raking these companies over coal to make them adhere to their specs set by the CPU, not just do what ever they want. 

 

Whilst i only watch the majority rather than all of their MB teardowns i don't recall running into an example of a board before that is outright out of spec, they like to shave the VRM as close as they can get it, but they don't normally outright undercut it.

 

I have to wonder if MB manufacturers got bad power info from Intel in early design and weren't able to adjust in time? because this is very unusual.

[GDRRiley]

1 hour ago, mr moose said:

I have lots of Asrock boards, they have all been great.  this is the first time I have seen anything that suggests they skimped somewhere in the build. 

I'm using one and have built 3 ryzen PC with one of their B450 boards. They also tend to break specs with external clock generators or OC on non Z chipsets for intel. Their rack division also makes some cool servers

 

58 minutes ago, Arika S said:

CPU manufacturers need to start raking these companies over coal to make them adhere to their specs set by the CPU, not just do what ever they want. 

its more on intel than anyone else, maybe if your turbo wasn't 2-3x your TDP you wouldn't have so many issues around boards.

[Arika]

23 minutes ago, GDRRiley said:

its more on intel than anyone else, maybe if your turbo wasn't 2-3x your TDP you wouldn't have so many issues around boards.

TDP is not power draw. If MB manufacturers don't know this, then they need to quit out of the game. If the same CPU causes motherboard VRMs to thermal throttle on different boards and not on others, than that's on the board manufacturer

[Fnige]

4 hours ago, handymanshandle said:

I was gonna make an AM3+ joke

Wait... what happened with AM3+ to relate to this?

[CarlBar]

9 hours ago, Arika S said:

TDP is not power draw.

 

Sort of yes and sort of no. Unless your playing games with your TDP Definition or have really inefficient silicon the average power draw should be within a very small percentage of it. Of course intel does play games. But it's not like intel dosen';t normally tell the Mb manufacturer's the values they actually need. Which is why i'm so surprised at this. MB manufacturers aren't stupid, if they knew the CPU's where going to draw this much they'd have built with it in mind. That suggests somwhere along the way the MB manufacturers, (or their VRM design people anyway), got bad info. How and why that happened though we don't know.

[5x5]

17 minutes ago, CarlBar said:

 

Sort of yes and sort of no. Unless your playing games with your TDP Definition or have really inefficient silicon the average power draw should be within a very small percentage of it. Of course intel does play games. But it's not like intel dosen';t normally tell the Mb manufacturer's the values they actually need. Which is why i'm so surprised at this. MB manufacturers aren't stupid, if they knew the CPU's where going to draw this much they'd have built with it in mind. That suggests somwhere along the way the MB manufacturers, (or their VRM design people anyway), got bad info. How and why that happened though we don't know.

Or they were expecting the usual 150-180W not the behemoth 250-300W that 10th gen CPUs draw. One would think pushing clocks past the forms of sanity would be a bad design idea and canned by management. Not like we have FX to serve as an example of pushing beyond safe limits...

[CarlBar]

5 hours ago, 5x5 said:

Or they were expecting the usual 150-180W not the behemoth 250-300W that 10th gen CPUs draw. One would think pushing clocks past the forms of sanity would be a bad design idea and canned by management. Not like we have FX to serve as an example of pushing beyond safe limits...

 

They were expecting what intel told them it would use. They don't design VRM's blind. Intel gives them detailed specifications ahead of time and once they get past the early prototyping stage they actually receive CPU's to put into the boards to validate they work. Thats why this is so weird. They should have seen these result in their own internal testing well before the boards where shown to the public. So why didn't they, what happened. If it was one board from one manufacturer i could dismiss it as one manufacturer screwing up, but 3+ boards from 3 different manufacturer's. Thats way past the statistical likelihood of it happening without some other factor at play.

[5x5]

23 minutes ago, CarlBar said:

 

They were expecting what intel told them it would use. They don't design VRM's blind. Intel gives them detailed specifications ahead of time and once they get past the early prototyping stage they actually receive CPU's to put into the boards to validate they work. Thats why this is so weird. They should have seen these result in their own internal testing well before the boards where shown to the public. So why didn't they, what happened. If it was one board from one manufacturer i could dismiss it as one manufacturer screwing up, but 3+ boards from 3 different manufacturer's. Thats way past the statistical likelihood of it happening without some other factor at play.

It's a pretty common occurrence on low-end boards. They simply get power throttled rather than therm. We've tested a 9900K on B-series board - it cannot even maintain 4GHz

[maartendc]

16 hours ago, 5x5 said:

It's a pretty common occurrence on low-end boards. They simply get power throttled rather than therm. We've tested a 9900K on B-series board - it cannot even maintain 4GHz

Z series board should not be low-end by definition. It is the high-end chipset. With high-end prices to match.

 

First thing you should be able to expect from a Z chipset board is that it doesn't limit your performance on a basic OC

If you go up in price on the Z-boards, you should expect more features, more ports, more fancy dual BIOS etc. etc. But to my mind, even the cheaper Z-boards should deliver on at least reasonable OC performance.

[5x5]

1 minute ago, maartendc said:

Z series board should not be low-end by definition. It is the high end chipset. With high end prices to match.

300W i9 go Brrrrr

Memes aside, it's a sad reality but true. With Intel unlocked CPUs, you can't get a value board if you want performance to match even stock levels.