Trying to choose a motherboard (for 7800X3D, almost certainly)

The 13600K thermally throttles almost immediately with default MSI BIOS settings,
but underVolting to address that allowed modest overclocking with air cooling.
Perhaps some unlucky chips require stock settings...

An 7800X3D seemingly always runs hot.
"Hot" is a relative term. If a CPU runs at 110 degrees but is safe to run until 150 degrees then 110 degrees isn't hot. It's then a meaningless number. These CPUs are meant to run at 90-ish degrees all day.

7800X3D handily beats a 13600K in most games. It usually trades blows (sometimes slower, tied, faster) with the 13700K and 13900K/13900KS/14900K.

P.S. I own a 12900KS and haven't used an AMD CPU since the FX-57 (I think the fastest single-core consumer CPU ever! $1000 back in 2005! Haha!!!) so I'm no fanboy of AMD nor any company. I just tell it like it is.
 
Last edited:
"Hot" is a relative term. If a CPU runs at 110 degrees but is safe to run until 150 degrees then 110 degrees isn't hot. It's then a meaningless number. These CPUs are meant to run at 90-ish degrees all day.
Yes, but you still have the smell of hot dust/air together with the heat in your room :p
7800X3D handily beats a 13600K in most games. It usually trades blows (sometimes slower, tied, faster) with the 13700K and 13900K/13900KS/14900K.
Little addition: in games, that actually make use of more than 8 cores, the fps are high enough anyway or the graphics card will be the limit in widely used resolutions (thinking of titles like the F1 games or Assassin's Creed, that show a lot of threads in process explorer).

Example: my 7600x can get maxed out (full 100% on all cores) in AC Valhalla if I disable vsync & limiters.
But it goes together with my 3080 hitting its limit at 60 fps in 1920x1080, so I only see this in lower resolutions, which really doesn't make sense outside of benchmarks.

Only use case would be to play with a 4070 ti or better on a high Hz 1080p monitor.
 
At Neil: maybe an extension card would make more sense for you?
If it's only about sata ports, which I totally understand, having quite a few older SSDs laying around, then it might be better to get B650 board + extension slots.
For some reason I mostly read about issues with the X()70 chipsets from AMD, not so much with the B()50 chipsets.

I'm quickly finding 8x SATA extension cards for the small pci-e slots.
If you're planning a fancy build with a glass side panel etc., forget about this. But if you're like me, using a case without any windows... You might save some money!
My mobo is 190€, the Taichi is 530€... I'd rather buy an extension card and get a B850 mobo with, I don't know, DDR5 9800 support and 6x USB 40 gbps than to splash out 340€ for some maybe useful pci-e 5.0 and 2 sata ports.
 
Last edited:
Results I found for AC had 13600K ahead, but overall not much difference.
Ignore that site. It's what I like to call a "bot site". There are tons of "bot sites" like that around whether it be CPU comparisons, kitchen appliance comparisons / reviews / top-10 lists, or for just about any conceivable type of product.

Go to actual reviews. Real benchmarks by real human beings.
dsgdfgdfgdfgfd.PNG

gfhfghfghfghh.PNG
 
Last edited:
Results I found for AC had 13600K ahead
If we can find someone with a 13600k, I'd happily do some cpu limit benching in AC with my 7600x.

The 5800x3D matches the 7600x in AC, there's a little benchmark thread about this here on RD.
Would be interesting!
AMD recommends at least 280mm radiator for 7800X3D.
That is relatively greater cooling than for 13600K.
Yeah but I don't give much about such recommendations. It's similar to PSU recommendations, which are mostly 50% over the top.

The 7800X3D doesn't use more than about 70W while gaming. It also has so much performance, that you can simple set a power limit at 90W and a thermal limit at 85°C and just let it hit the limits during peaks or rendering stuff etc.

I tested this and I gained about 3 seconds per minute of rendering in davinci without any limits, all fans at maximum, just barely not throttling at slightly above 90°.
I gladly wait 3 seconds longer per minute and have a cpu that only uses 75W and doesn't go above 80°C.

The throttling works very well, no hard frame time spikes when it happens ingame. Just slightly going down fps.
 
Glad to hear it's nicely cool during gaming :thumbsup:

Ta, nice video, I learned quite a lot about AMD chipsets. However, given that I want more than 4 SATA ports, I am one of the unusual people who genuinely has something to gain from the X670 :)

I imagine the 4 sata port limit also applies to a lot of X670 boards below £300. It might be worth looking on pcpartpicker to see what boards cater for 6 sata ports on both platforms. The MSI MAG B650 TOMAHAWK looks promising... There is a youtube review of 35 B650 boards by Hardware Unboxed.
 
Last edited:
"Hot" is a relative term.
Agreed...
Yes, but you still have the smell of hot dust/air together with the heat in your room :p
Yes, though I think the impact on the room is mostly down to the dissipated power rather than the absolute temperature the CPU reaches ;)
And since the X3D chips have to stay a little cooler than the others (given the thermal resistance of the cache layer), that probably translates to slightly less of the "hot electronics" smell along with the warm air...? :p
At Neil: maybe an extension card would make more sense for you?
Yes possibly but I've previously looked and to my great surprise failed to find any promising SATA cards, so have yet to buy anything.
I'm quickly finding 8x SATA extension cards for the small pci-e slots.
Yeah I have found cards in the past that looked superficially OK but I found that all had one or more of these failings: stupidly expensive, no-name manufacturer, terrible reviews/feedback from users. If you have found a decent candidate please share :) (I can no longer remember all of the details of my searches, I confess, and maybe the market has just plain improved since I looked.)
AMD recommends at least 280mm radiator for 7800X3D.
That is relatively greater cooling than for 13600K.
Was unaware of this. Given that the max dissipation is lower for the X3D chip (and typical dissipation probably a lot lower), I wonder if this is a reflection of the lower Tjmax (harder to dissipate same amount of heat with lower temperature differential to ambient)...
Results I found for AC had 13600K ahead, but overall not much difference.
You mention AC, but the link seems to go to a generic benchmark which doesn't mention any games. Wrong link pasted in by accident?
Ignore that site. It's what I like to call a "bot site".
Userbenchmark has a lot to answer for but it's not one of those bot sites I think you're talking about (mindless junk). It has utterly nonsensical editorials about AMD chips (both CPU and GPU) written by some very salty people who clearly hate AMD, but the actual benchmark data - derived from users - generally looks sound to me.
 
I am on amd now. Last time was the athlon thunderbird or what it was called I believe? Or maybe I had some x2 or something later I can´t really recall lol

7800x3d is supposed to use less watt then the intel and run cooler in general? Mine idles at 37 degrees. But that is a new mount so thermal paste hasnt quite settled yet and my water cooler is optimized more against low noise then maximum cooling so it´s hardly empiric data.

As for sata card I got the asrock x670e pg lightning and it only got 4 sata slots so I ordered this https://www.amazon.se/dp/B08F56WKW7?th=1.

Delock is supposed to make some decent then there is lsi and other alternatives. But compability etc is always a gamble so who knows if it will work out.

I am not having a smoth sailing though. Memory seem to have to run 1-3 slot 0 never works? Half the upstarts the boot manager is not found. I had a bluescreen and then had to fight hard to clear cmos one time. Now doing a clean win 11 install and hope things will be smooth from now on...

bios is unfamiliar terrain. Only thing I tried is using my memorys expo settings. Currently only using one stick at stock settings to be on the safe side though.
 
I was looking at a typical board mentioned above. It has 16x 90A power stages, for a total of 1,440 Amps. At 1.1 or so volts, that's at least 1500W - just for the CPU! That corresponds to about 2 horsepower, by the way.

So to use your board at its full potential (which actually seems impossible unless your CPU chip is in the process of melting into slag), you also need a 1500 W power supply. That is about the maximum that a typical U.S. household plug can safely provide. That's also before adding a 4090, etc. Crazy! And it's not close to the highest power boards!

No wonder my room gets hot after a half hour of racing.
 
Userbenchmark has a lot to answer for but it's not one of those bot sites I think you're talking about (mindless junk). It has utterly nonsensical editorials about AMD chips (both CPU and GPU) written by some very salty people who clearly hate AMD, but the actual benchmark data - derived from users - generally looks sound to me.
Userbenchmark is clearly one of the countless bot sites around. No one is manually testing all those CPUs and games and manually entering in their results for the thousands and thousands of PC parts and software combinations on that site. It's all auto-entered/generated. Even the write-ups and descriptions in comparisons often make little sense and are clearly auto-generated from certain variables and values.
 
Last edited:
Yes this stuff bothers me somewhat. In fact, one of the (several) reasons I have blacklisted Asus is because of what my Z170-A did when I used the XMP profile for a 3200 kit. If the AMD ecosystem is currently a worse place for that kind of crap then I might back away from it... :-S
Based on the research I did, and just having built a new system about a month ago.
(See my recent thread if you want) In it, I shared many videos on products and why I chose what I did...

Any original issues on release with early boards (Asus X670e) seem to of been sorted. This is also highlighted as its a motherboard used by many system builders, including high end systems/packages from companies like ocUK.

What I can tell you is that while I spent more than I needed on the Crossfire Hero X670e but in fairness, I wanted a good "performance" motherboard, with decent onboard audio, lots of connectivity. Also with the benefits with this we have the potential for CPU upgrades for upto 2 years. On that point I am not sure if all boards are but I choose a board that will also support the bandwidth for PCIe 5.0 for GPU as well.

I would highlight that already M.2 PCI 5.0 is showing advantages with recent drives, but again in a year or so. Then we are only going to see prices continue to fall and performances increase much more for new improved drives. Personally I went with one of the current best performance M.2 drives available in 2TB for my game drive and then a 990 Pro for system drive. I dont need or will use SATA as will just transfer anything form other systems to this new rig.

Will say I am very happy with the performance I have seen compared to what I had. No issues, simple firmware update for the motherboard and even with the very latest Corsair memory (EXPO) it has worked perfectly.

Overclocking, Im not that fussed if I gain 5-10 frames, as its not that big of an issue to drop down a resolution or reduce some settings, but in certain scenarios or for some then it may be wanted additional performance.

Hope you end up happy with what you get...

See end of video as Dan also emulates the 7800X3D to compare to intel.

 
Last edited:
I am not having a smoth sailing though. Memory seem to have to run 1-3 slot 0 never works?
One odd thing I noticed with one (or more?) of the AM5 boards I looked at: while the RAM slots were numbered A1/A2/B1/B2, you are told to populate the A2 & B2 slots first, which is just bloody stupid IMHO. Why not number them "correctly"?! No idea if this is part of what's going on for you though.
I was looking at a typical board mentioned above. It has 16x 90A power stages, for a total of 1,440 Amps. At 1.1 or so volts, that's at least 1500W - just for the CPU! That corresponds to about 2 horsepower, by the way.
I certainly don't fully understand the power phase stuff for CPUs but for sure, it's not delivering all of that current to the CPU. 1500 W would unequivocally destroy any desktop CPU with air/water cooling (and likely even if cooled by liquid nitrogen).
In my head, it makes some kind of sense if the separate stages/phases are sharing the current, with different phasing to keep the voltage very stable.
Switching speed is limited, and every switching transient dissipates a bit of energy in the MOSFET, so if you spread the switching across many phases you can increase the effective switching speed (= more stability I guess?) and more effective power handling (because many MOSFETs are taking the heat). But yeah, this is partly guessed from what I know of switch-mode power supplies.

Userbenchmark is clearly one of the countless bot sites around. No one is manually testing all those CPUs and games and manually entering in their results for the thousands and thousands of PC parts and software combinations on that site.
Nope, I have personally downloaded their benchmark tool and tested my own PCs, uploading the data for analysis (I think the upload was intrinsically part of the test sequence, but it has been years since I did it).
What I can tell you is that while I spent more than I needed on the Crossfire Hero X670e but in fairness, I wanted a good "performance" motherboard, with decent onboard audio, lots of connectivity.
Yeah, I'm inclined to err on the generous side myself. One thing I didn't say explicitly in the OP (but people may have inferred from the 7700K! ;)) is that I don't upgrade often - 5+ years would be totally normal for me between major upgrades.
 
I also went from the 7700k. my 850w psu seem to be plenty enough though only using a 3080. But I know those that run 4090 on same psu and cpu and don´t have any issues.

x670e is probably not necessary at all for 7800x3D power delivery wise.
 
Yeah I have found cards in the past that looked superficially OK but I found that all had one or more of these failings: stupidly expensive, no-name manufacturer, terrible reviews/feedback from users. If you have found a decent candidate please share :) (I can no longer remember all of the details of my searches, I confess, and maybe the market has just plain improved since I looked.)
Only gonna reply to this one quickly:
I agree that the market doesn't seem great.. It's becoming more and more a thing that the Chinese manufacturers of the stuff that is normally branded for known companies sell more and more directly under seemingly not trustworthy names.
I don't like this, but from my experience that stuff usually works very well, you just don't get the "bloatware" you normally get.

On German Amazon, there's a 29€ 4-Port card with 4.4 stars and 1.372 reviews. That's the one I'd buy:
"Ziyituod SATA Karte, PCIE 3.0, 4 Ports mit 4 SATA-Kabeln, SATA-Controller-Erweiterungskarte mit Low-Profile-Halterung, Non-Raid, Booten als Systemfestplatte https://amzn.eu/d/73kZ7GP"

And then there's Startech at 89€. They are the ones that also build true USB controllers into their cards to actually get more endpoints.
Sadly extremely overpriced imo...

"StarTech.com 4 Port PCI Express 2.0 SATA III 6Gbit/s RAID Controller Karte mit HyperDuo SSD Tiering - PCIe SATA 3 Controller Adapter (PEXSAT34RH) https://amzn.eu/d/4Z6ObXl"
 
And then there's Startech at 89€. They are the ones that also build true USB controllers into their cards to actually get more endpoints.
Sadly extremely overpriced imo...
Yeah I dimly recall a promising Startech card that turned out to be out of stock and the alternatives were OP in both ways (overpowered compared to my needs, and overpriced :roflmao:).
 
Yeah I dimly recall a promising Startech card that turned out to be out of stock and the alternatives were OP in both ways (overpowered compared to my needs, and overpriced :roflmao:).
I believe both cards are pci-e gen 2. The StarTech states it & the Ziyituod card uses an ASMedia chip (1062). Compared to Gen 3, Gen 2 uses a less efficient encoding scheme, and from what I've read, there should be a noticeable performance difference. So it's more under-powered & overpriced :roflmao:. That said, most of the B650 boards (MSI) offering 6 sata ports use ASMedia chips (mostly 1061, Pro uses 1064), including the B650 Tomahawk board that I suggested, to add two additional sata ports (4 native + 2 ASMedia). MSI seem to use ASMedia (1061) chips in some of their X670 boards too (MSI Pro X670-P WiFi...). I've only looked @ MSI, others might support 6 native sata ports or give an either/or choice (i.e., a pcie slot vs 4/5 sata ports ).

Here's a cheap Gen 3 card (I think :cautious::roflmao:).
 
I believe both cards are pci-e gen 2. The StarTech states it & the Ziyituod card uses an ASMedia chip (1062). Compared to Gen 3, Gen 2 uses a less efficient encoding scheme, and from what I've read, there should be a noticeable performance difference. So it's more under-powered & overpriced :roflmao:. That said, most of the B650 boards (MSI) offering 6 sata ports use ASMedia chips (mostly 1061, Pro uses 1064), including the B650 Tomahawk board that I suggested, to add two additional sata ports (4 native + 2 ASMedia). MSI seem to use ASMedia (1061) chips in some of their X670 boards too (MSI Pro X670-P WiFi...). I've only looked @ MSI, others might support 6 native sata ports or give an either/or choice (i.e., a pcie slot vs 4/5 sata ports ).

Here's a cheap Gen 3 card (I think :cautious::roflmao:).
Nice info!
Do you know more about the potential performance difference?
From what I've read Gen 2 offers 5 gbps per lane, SATA has a maximum of 6 gbps.
But I'm yet to see a SATA SSD pushing data with the theoretical 750 MB/s that equal 6 gbps.

Your Gen3 card is cheaper anyway, but I'm not sure if the performance isn't neglectable for normal usecases, where you never reach the full throughput of the drives anyway due to the controller being the bottleneck when files aren't very big.
 
I believe both cards are pci-e gen 2.
Oooh, hadn't spotted that. It would be potentially OK if it was a 4-lane card but it's a 1x so... yeah, bottleneck if you want to read more than one drive at a time, and given that that Startech card is a RAID card, I'm pretty stunned. I see that the non-RAID Startech cards are less stupidly priced, but still not great (and some of them are PCIe 3.0, and/or more than 1 lane).
Here's a cheap Gen 3 card (I think :cautious::roflmao:).
Gosh that is cheap, relative to the competition. Feels like a fair price though, tbh, given how little is on the card. I like that it's also 2 lanes, so ballpark 2 GB/s total throughput.
From what I've read Gen 2 offers 5 gbps per lane, SATA has a maximum of 6 gbps.
But I'm yet to see a SATA SSD pushing data with the theoretical 750 MB/s that equal 6 gbps.
Yeah I tend to mentally convert SATA 3's 6 Gbps to 600 MB/s, allowing for I/O overhead, and then most of the SATA SSD drives that I've tested seem to more or less reach that kind of level (IIRC, has been a while).
Your Gen3 card is cheaper anyway, but I'm not sure if the performance isn't neglectable for normal usecases, where you never reach the full throughput of the drives anyway due to the controller being the bottleneck when files aren't very big.
Will probably be relevant mainly when talking to multiple disks in parallel I guess (e.g. via s/w or h/w RAID).

This reminds me that I currently use the Intel h/w RAID on my Z170-A Asus board - RAID 1 for system volume - and plan to keep doing that on whatever I buy next.
I love that I can just rotate a pool of drives to get a full backup of everything I care about. (Game installations mostly sit on a separate non-RAID volume.)

After some light googling, I am feeling a little cautious about RAID on AMD motherboards - anyone tried it?
 

Latest News

What's needed for simracing in 2024?

  • More games, period

  • Better graphics/visuals

  • Advanced physics and handling

  • More cars and tracks

  • AI improvements

  • AI engineering

  • Cross-platform play

  • New game Modes

  • Other, post your idea


Results are only viewable after voting.
Back
Top