Low-Cost Meets Eight Cores: MSI MAG Z390 Tomahawk Review

Benchmark Results & Final Analysis

We started today’s test with a bunch of “errant” findings, having only assured that automatic overclocking had been disabled and CPU power-saving features had been enabled. After initially disregarding those results, we continuously retested the MAG Z390 Tomahawk until we got results that wouldn’t have MSI giving us the axe. Those results came by disabling the voltage regulator’s power saving mode, after which we feared a “China Syndrome” moment during our Prime95 max power draw test. So, we started with a new board…and included the results from both the performance-optimized “CPU defaults” and power-throttled “motherboard defaults.”

Synthetic Benchmarks

The MAG Z390 Tomahawk performed like any other board after being “optimized” by disabling its voltage regulator thermal throttling, but that throttling imparted a noticeable impact in its default configuration upon the most CPU-intensive 3DMark and PCMark tests.

It’s tough to imagine why Sandra wouldn’t have forced voltage regulator thermal throttling, except that its AVX load might have been in shorter bursts than would have caused the voltage regulator to overheat. Disabling that throttling helps the MAG Z390 Tomahawk stay at or near the top of our Compubench and Cinebench tests, but enabling tanks those benchmarks. In general, thermal throttling impact is proportional to loaded core time.

3D Games

Thermal throttling hurts the MAG Z390 Tomahawk’s results in every game, and the board performs expertly with that safety feature disabled.

Timed Applications

Our timed applications are either too short, using too few cores, or not AVX-optimized enough to force voltage-regulator thermal throttling on the MAG Z390 Tomahawk. That’ll push the default configuration up in our averages, which is a shame since many of the people who buy the Core i9-9900K need it for long-duration loads.

Power, Heat, & Efficiency

When fully loaded under Prime95 small-FFTs, the Z390 Tomahawk spikes to around 241W and then drops to around 140W. It also drops to around 3.40 GHz. And that’s not realistic since the CPU is only throttling to around 4.5GHz in several of our benchmarks. And so, we disabled thermal throttling and saw system power draw jump up to 330W.

Unfortunately, we forgot that temperatures over 115 °C were out of range for our thermistor. And so, temperatures kept climbing until the board shut down and refused to boot back up, instead showing a CPU error. No shooting flames, smoke geysers or welding arcs as we’ve seen in previous boards, just a quiet power down without recovery, and a burned thermistor.

Testing with MSI’s monitoring software allowed us to find the sweet-spot, where the voltage regulator would hover around 100 °C at full load at the CPU’s rated 4.70 GHz when running Prime95 small-FFTs.  But that’s when we found out that the voltage regulator heat sink temperature was around 40° lower than its MOSFET (transistor) temperature. Ouch.

Thus, we get efficiency scores that mean absolutely nothing. The board can’t power our CPU to full load without cranking down its core voltage, but doing so isn’t fair to any of the other products.

Overclocking

Having found the one voltage the lets us reach the standard 4.70 GHz Intel Turbo Boost frequency for eight loaded cores…without overheating the voltage regulator, there was no room left to push the CPU frequency upward. Our 1.20V setting is the be-all-end-all, you’ll get 4.70 GHz at the voltage regulator’s limit. Meanwhile, competing boards were fully capable of using 1.30V to push our CPU to 4.90 or 4.95GHz.

The MAG Z390 Tomahawk’s memory overclock was close to the top we’ve seen on this combination of CPU and DRAM, and its performance at that overclock was close to the average of other boards.

Final Thoughts

We tested the MAG Z390 Tomahawk with Intel’s Core i9-9900K and found that this processor is really outside its capabilities: When the power regulator reached its thermal threshold, it would drop the CPU frequency below its rated 3.60 GHz baseline. Lighter-load users wouldn’t notice that, but then again, why would they need a Core i9-9900K?

And so the low price that looks like a great deal in our price-to-performance table is a terrible deal for anyone who would like to use the Core i9-9900K at its full potential. We wish we were able to recommend such an inexpensive board to Core i5-9600K users, but we simply didn’t have the spare CPU to verify that the board is even capable of pushing that processor’s performance potential. Doing so wouldn’t have been fair to the competing boards anyway…and so, we’ll set aside the MAG Z390 Tomahawk sample that we didn’t burn while we consider options that include a roundup of inexpensive boards that won’t properly support our Core i9-9900K.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Create a new thread in the Reviews comments forum about this subject
20 comments
Comment from the forums
    Your comment
  • Rogue Leader
    Absolutely shocked (not) that MSI has yet again released a fancy board that looks the part but can't walk the walk. This is what they are known for. I literally will not buy or recommend their boards for this fact, because for any good boards they release, they trick unwitting consumers with junk like this.
  • Onus
    This is the sort of board you might find in a Cyberpower or iBuypower "Gamer." It's a shame, because some MSI boards are really nice. I wish Asus were more willing to submit boards for review, they seem consistently good. With ASrock, the boards are generally good but typically have anomalies, Biostar is mostly good but sadly unfinished, I have a hard time now trusting Gigabyte, but this looks like a real "gotcha" as far as roping in suckers.
  • g-unit1111
    Hey it's the new 78LMT for 2018. Remember how that motherboard combined with an 8350 or a 9590 would be instant death for that PC? I could see it happening here easily.

    And sure, a $160 board can support a $500 CPU. Also, hold my beer! :lol:
  • karolbe
    I have to say that this review was based on a broken sample or is simply -------. I have i9-9900k and this mainboard and I have no such issues at all. I have been stress testing the system for couple of days with OCCT and Prime95 and it is rock stable and performs as expected.

    mod edit: language
  • Onus
    A bad sample is certainly possible. Hopefully Thomas will see this, and get an answer from MSI.
  • Crashman
    Anonymous said:
    I have to say that this review was based on a broken sample or is simply -------. I have i9-9900k and this mainboard and I have no such issues at all. I have been stress testing the system for couple of days with OCCT and Prime95 and it is rock stable and performs as expected.

    mod edit: language

    I find your words incredible. There are only six phases feeding the CPU cores, and holding 4.70 GHz on the Core i9-9900K at the Core i9-9900K's stock voltages is simply impossible for the board. With our CPU sample it throttles down to 3.4 GHz, which is lower than stock. But that's an "edge case": With a less-leaky CPU sample, you might get 3.60 GHz and have less room to complain. Of course you can always try reducing the core voltage, which is how we got it to run at 4.70 GHz in the O/C test.

    BTW, you can also get better results by running a lower-stress test in tandem, than by running Prime95 small-FFTs on its own. Prime95 assumes secondary priority, so the lower-stress "primary" task will have dominance over power load.

    That's also why, in order to test maximum power consumption for a gaming and encoding PC, we used to leave the FIRST thread idle in Prime95: The graphics test needed that thread, so that a system that consumed 550W with 7 P95 threads and a game test running might only need 400W with 8 P95 threads and the same game test running.

    If you want to draw max CPU power, run Prime95 small-FFTs with all other tasks idle.

    Anonymous said:
    A bad sample is certainly possible. Hopefully Thomas will see this, and get an answer from MSI.

    It was confirmed with MSI in a most...unflattering way. I have another MSI board with the same voltage regulator sitting here, not getting reviewed yet. I've offered to try a "Best Z390 for Core i7" article to retest all the boards that couldn't hold the Core i9-9900K at 4.70 GHz under Prime95 small-FFTs, but haven't received a response. Maybe that's because of the language I used...I might have said something about a loser board retest roundup...

    BTW, I'd like to apologize for not returning to this thread sooner: I'm no longer getting email notifications of these responses for some strange reason.
  • Peter Martin
    Stay away from msi everything
  • karolbe
    https://imgur.com/a/XYuQs71 there you go.

    I have been running it only for 5 minutes, I don't have more time now. But believe me, I have been stress testing the system for hours continuously using Prime95 and OCCT because I had problems with stability and wasn't sure what is faulty, turned out that 2 out of 4 HyperX modules died. However, with the new memory modules, there were no issues with stability. As you can see from the screenshot CPU was at 4.7GHz for the 5 minutes test period.

    This is on the default BIOS settings, I have only enabled XMP for the memory, and that's it.
  • Crashman
    Anonymous said:
    https://imgur.com/a/XYuQs71 there you go.

    I have been running it only for 5 minutes, I don't have more time now. But believe me, I have been stress testing the system for hours continuously using Prime95 and OCCT because I had problems with stability and wasn't sure what is faulty, turned out that 2 out of 4 HyperX modules died. However, with the new memory modules, there were no issues with stability. As you can see from the screenshot CPU was at 4.7GHz for the 5 minutes test period.

    This is on the default BIOS settings, I have only enabled XMP for the memory, and that's it.
    I tried to look for the Prime95 threads in your screenshot, but after a couple minutes of looking I got an auto-redirect adware that attempted to hijack my browser...

    OK, looked at it in a new browser, is that monitoring app showing the CPU bouncing between 1.5 and 4.7 GHz?
  • karolbe
    Anonymous said:

    OK, looked at it in a new browser, is that monitoring app showing the CPU bouncing between 1.5 and 4.7 GHz?


    Well, yes but it was before I started the test which is perfectly normal and expected. When the test started it was solid 4.7GHz on all cores (so there is just one line obviously at 4.7GHz).

    Better screenshot (this time on dropbox, didn't know that imgur is so bad nowadays)

    https://www.dropbox.com/s/mswid3kg7k84gas/Selection_021.png?dl=0
  • Crashman
    Anonymous said:
    Anonymous said:

    OK, looked at it in a new browser, is that monitoring app showing the CPU bouncing between 1.5 and 4.7 GHz?


    Well, yes but it was before I started the test which is perfectly normal and expected. When the test started it was solid 4.7GHz on all cores (so there is just one line obviously at 4.7GHz).

    Better screenshot (this time on dropbox, didn't know that imgur is so bad nowadays)

    https://www.dropbox.com/s/mswid3kg7k84gas/Selection_021.png?dl=0

    Ah, and that's Prime95 small-FFTs holding the Turbo Boost multiplier up?
  • karolbe
    It was OCCT, all is visible on the screenshot. Prime95 test I did a few days ago was similar, 4.7GHz on all cores, it was not the small FFTs test though, I was testing memory and not CPU back then and the small FFTs test is more CPU bound. I expect that AVX instructions can lower frequency a bit but for sure not to the 3.4GHz level as the review states. I can do this test when time allows (I am Linux user so I can not always reboot to Windows to run Prime95).

    Also, looking for other reviews (not many sadly) of the m/b I found this one. Have a look at the screenshots there, 5.1GHz on all cores on the very same mainboard. How do you explain this?

    https://thinkcomputers.org/msi-mag-z390-tomahawk-motherboard-review/5/

    I find this review much more believable than the one here on tom's hardware, it is simply more true to what I can see with my own eyes.

    I would personally never post a review bashing a product without first understanding the root cause, which I am rather convinced is far from bad m/b design but rather is related to a bad sample or testing error.
  • Crashman
    Anonymous said:
    It was OCCT, all is visible on the screenshot. Prime95 test I did a few days ago was similar, 4.7GHz on all cores, it was not the small FFTs test though, I was testing memory and not CPU back then and the small FFTs test is more CPU bound. I expect that AVX instructions can lower frequency a bit but for sure not to the 3.4GHz level as the review states. I can do this test when time allows (I am Linux user so I can not always reboot to Windows to run Prime95).

    Also, looking for other reviews (not many sadly) of the m/b I found this one. Have a look at the screenshots there, 5.1GHz on all cores on the very same mainboard. How do you explain this?

    https://thinkcomputers.org/msi-mag-z390-tomahawk-motherboard-review/5/

    I find this review much more believable than the one here on tom's hardware, it is simply more true to what I can see with my own eyes.

    I would personally never post a review bashing a product without first understanding the root cause, which I am rather convinced is far from bad m/b design but rather is related to a bad sample or testing error.
    You have to first understand that this is an edge case: Prime95 small-FFTs was the only workload we could find to make the board do that at default voltage settings and maximum power settings. We're using maximum power settings because most boards will throttle back "more gently" when using small-FFTs, which prevents testing consistency for maximum power and heat. So if a board doesn't hold 4.70 GHz under that excessive load, we increase the power limits and CPU thermal thresshold to allow it.

    If I recall correctly, the difference between "stays at 4.70 GHz" and "eventually fries the voltage regulator" is around 40W. It's understandable that MSI was a little miffed about the excessive load testing, because boards at this price point aren't designed for that type of abuse. And certainly, the thermal protection that prevents the voltage regulator from frying should NOT be disabled for normal use. But like I said, we were looking for the worst-case-sceenario in power and heat measurements.

    Now I'd like to send you to the Asus Maximus XI Hero follow-up thread, where people are arguing that I "must not have" used such a rigorous test, because that board DIDN'T fail ;)
  • Crashman
    BTW, I asked in the lead "Can a $160 motherboard support a $500 processor? Should it?" because in reality most users hoping to push a Core i9-9900K to its limit are looking at nothing less than the $200 MPG Gaming Pro Carbon. And that board passed.

    MSI thinks that the MAG Z390 Tomahawk is the perfect money-saving board for Core i5 and Core i7 users, and that belief is probably correct.
  • karolbe
    Anonymous said:
    BTW, I asked in the lead "Can a $160 motherboard support a $500 processor? Should it?" because in reality most users hoping to push a Core i9-9900K to its limit are looking at nothing less than the $200 MPG Gaming Pro Carbon. And that board passed.

    MSI thinks that the MAG Z390 Tomahawk is the perfect money-saving board for Core i5 and Core i7 users, and that belief is probably correct.


    I disagree and I showed a proof that the MAG can easily power i9-9900k fully loaded. If I were you I would be looking for answer why your mainboard sample behaves so strangely under AVX load.

    BTW that MPG Gaming Pro Carbon, what is the frequency in Prime95 in Small FFTs test? Does it run on 4.7GHz? Or slightly less, e.g. 4.5GHz? (I am thinking here about the default setup, no bios changes etc.).
  • Crashman
    Anonymous said:
    Anonymous said:
    BTW, I asked in the lead "Can a $160 motherboard support a $500 processor? Should it?" because in reality most users hoping to push a Core i9-9900K to its limit are looking at nothing less than the $200 MPG Gaming Pro Carbon. And that board passed.

    MSI thinks that the MAG Z390 Tomahawk is the perfect money-saving board for Core i5 and Core i7 users, and that belief is probably correct.


    I disagree and I showed a proof that the MAG can easily power i9-9900k fully loaded. If I were you I would be looking for answer why your mainboard sample behaves so strangely under AVX load.

    BTW that MPG Gaming Pro Carbon, what is the frequency in Prime95 in Small FFTs test? Does it run on 4.7GHz? Or slightly less, e.g. 4.5GHz? (I am thinking here about the default setup, no bios changes etc.).

    The problem is that Prime95 small-FFTs are the edge case that I claimed. Default behavior for VRM thermal protection was to kick it back to 3.4 GHz. We tried adding a fan but found that the thermal protection was overly aggressive. And that's what lead to disabling it completely (the board lacking any "in between" setting), which exposed that the thermal interface for its heat sink was also insufficient.

    Now you could be assuming that we just got a bad board, but we burned the first one. So you'd then have to assume that we got two bad boards several weeks apart.

    With Prime95 small-FFTs, the only way to approach the limit without frying the board was to disable VRM thermal throttling, manually monitor the MSI thermal report, and kill the process when it reached 112 °C. And so it was with the second board, after the first died from not being monitored while exhibiting the same behavior.

    I had experimented a great deal with the second board and found that merely disabling Hyperthreading on the Core i9 was enough to prevent the problem under Prime95 small-FFTs, which takes me back to MSI's assertion that this is a great board for Core i7's.
  • William_X89
    The low price point and other factors convinced me to buy this board for use with an i7-8700k before this review had come out. Should I expect problems down the line if I ever attempt to overclock it?
  • Crashman
    Anonymous said:
    The low price point and other factors convinced me to board for the i7-8700k before this review had come out. Should I expect problems down the line if I ever attempt to overclock it?
    We did quite a bit of experimentation to find ways to get the CPU under the voltage regulator's limit using the Core i9-9900K. One thing that worked was disabling hyperthreading, another was to reduce voltage levels. Either way, cutting 40W from its maximum load power gets you into the safe zone, so the Core i7-8700K should already be there. As for overclocking the 8700K, your guess is almost as good as mine: I believe my 8700K ran around 270W at peak load when overclocked to highest stable frequency using 1.30V core, so you're "probably OK".

    BTW, all those numbers (including the ones in the charts) are at the wall, so there's the 10% power supply inefficiency, the ~20W of idle components during the test, power losses at the CPU voltage regulator, etc etc etc to consider. 270W is closer to 200W.
  • liam.fermo
    i have this mobo, will i7 9700k be good?
  • Crashman
    Anonymous said:
    i have this mobo, will i7 9700k be good?
    We got power under control by disabling HyperThreading on the 9900K, so you should be fine.