ASRock TRX40 Taichi

[damric]

Thanks for the links, they are much appreciated.

First, check my post above... I've reached out to confirm the process as well.

Second, the only relevant portion (knew the rest... ) was this (please note the terms in bold):


This is a GSkill person speculating about a mobo vendor's process. "probably" and "some" isn't as confident as your post.

......and that was from nearly 8 years ago.

So, anyhoo, I will report back as I get the responses.

Don't be surprised if they dodge the question and talk in circles it's got electrolytes that's what the plants crave

 

[dicktracy]

I’m surprised there’s no 4 DIMMs design for TRX40 like they did with x299 Apex and EVGA x299 Dark. Those have one of the best RAM tweakability!

I am really interested in TRX40 but I cannot think of a use case that would make me jump from X399 to TRX40 especially the huge gap in costs. Not the motherboards but the CPUs.

Everything has jacked up in prices. Flagship boards that used to be around $500 now goes up to $850 and CPU up to $4000. Irony is that this is coming from AMD and not Intel.

 

[kapone32]

Everything has jacked up in prices. Flagship boards that used to be around $500 now goes up to $850 and CPU up to $4000. Irony is that this is coming from AMD and not Intel.
[/QUOTE]

Well to be honest the pricing is not too stupid. I mean a $2000 24 core CPU that can OC to 4.4 GHZ on all cores is nothing to sneeze at vs the 10 core Intel chip that was and still is $2000 on some sites. $4000 may seem like a lot of money but @ 64 cores it's not bad if you have a use case for it. There is also the size of the cache on TRX40 (so far) CPUs that is quite eye-opening. What I laugh at though is how the board vendors are going to say that it is because of the VRM, PCI_E 4.0 and WIFI 6. I saw today that the As Rock X570 Pro is on sale at Newegg for $179.99 so is PCIE_4.0 really that expensive to implement? WIFI 6 adapters are $30 on Amazon. I could see the VRM but a lot of the VRMs on TRX40 are overkill in terms of the quality of the components (Maybe not for the 3990X) . The funny thing though is that some X399 (Asus Zenith Extreme Alpha) boards are more expensive than their TRX40 cousins. I cannot believe the GIgabyte X399 Designaire is $1999 on Newegg.ca.

 

[EarthDog]

so is PCIE_4.0 really that expensive to implement? WIFI 6 adapters are $30 on Amazon.

It all adds up man... that plus larger chipset heatsinks and fans...as you mentioned more robust VRMs and heatsinks, etc.

But yeah, cost of entry on the platforms when comparing like for like (mobo to mobo) tends to chip away at any price advantage AMD has (mainstream).

 

[Sashleycat]

That VRM cooling solution looks like it might actually dissipate heat! (and looking at the graphs, it does indeed)

Looks like a solid board, I've switched to Asrock in recent history and not been disappointed, the tech support is also on point.

It all adds up man... that plus larger chipset heatsinks and fans...as you mentioned more robust VRMs and heatsinks, etc.

But yeah, cost of entry on the platforms when comparing like for like (mobo to mobo) tends to chip away at any price advantage AMD has (mainstream).

To be fair, 99% of consumers buying AM4 should be looking at a B450 or upcoming B550, even I find little use for the Gen4 interfaces right now (that will likely change over the next couple years). I mainly went X570 for the appeal of having Gen4, lol.

 

[EarthDog]

To be fair, 99% of consumers buying AM4 should be looking at a B450 or upcoming B550, even I find little use for the Gen4 interfaces right now (that will likely change over the next couple years). I mainly went X570 for the appeal of having Gen4, lol.

True. But you(royal you, lol) can't have it both ways. On one hand people list its appeal, and it does have some and will have more in the coming years and practically shame intel for not having it. I liken it to RTX with Nvidia, honestly. While we can see the use in some cases (games), for most it is still a wait and see technology and not worth the additional cost of entry. PCIe4 and the additional bandwidth can be helpful for those who want ultra fast storage and a lot of it.

I feel PCIe 4.0 bandwidth is a generation away as well. Meaning, next generation, talking INtel 7nm (lol) and Zen 3 is where more will feel the ceiling of PCIe 3.0 bandwidth.

 

[Sashleycat]

True. But you(royal you, lol) can't have it both ways. On one hand people list its appeal, and it does have some and will have more in the coming years and practically shame intel for not having it. I liken it to RTX with Nvidia, honestly. While we can see the use in some cases (games), for most it is still a wait and see technology and not worth the additional cost of entry. PCIe4 and the additional bandwidth can be helpful for those who want ultra fast storage and a lot of it.

I feel PCIe 4.0 bandwidth is a generation away as well. Meaning, next generation, talking INtel 7nm (lol) and Zen 3 is where more will feel the ceiling of PCIe 3.0 bandwidth.

Fair comment. That said, if I were to defend myself a bit here: I've been a fan of the 20-series since launch, I'm very much a proponent of future-looking technology.

As for 4.0 bandwidth: you're most likely correct. I think RTX 3080 Ti will use a Gen4.0 16X interface, but IDK how much it will gain from it, perhaps 5% in some niche gaming scenario; not enough to be a game-changer in any capacity.

In all honesty, I think perhaps the best selling point of X570 is the Gen4 back-end: 8GB/s between SB and CPU is pretty nice if more board makers would actually wire up more than 1 NVME to the chipset, lol.

 

[EarthDog]

Sorry I wasnt clear.. I was trying to compare ray tracing on rtx cards and it's current usefulness to pcie 4.0. I think both will have its place, but next gen and beyond.

..and agreed with what you said.

 

[Sashleycat]

Sorry I wasnt clear.. I was trying to compare ray tracing on rtx cards and it's current usefulness to pcie 4.0. I think both will have its place, but next gen and beyond.

..and agreed with what you said.

I thought that was what you meant: and it's a fair comparison. But since I also have a positive outlook on the RTX cards, I figured I'd have free reign to nag at Intel for lack of gen4.0 on their current platforms.

It actually doesn't bother me too much on Z390, except for the DMI as I said, but Z490 as mentioned in another article? That might be an issue. IDK.

 

[lukart]

Fair comment. That said, if I were to defend myself a bit here: I've been a fan of the 20-series since launch, I'm very much a proponent of future-looking technology.

As for 4.0 bandwidth: you're most likely correct. I think RTX 3080 Ti will use a Gen4.0 16X interface, but IDK how much it will gain from it, perhaps 5% in some niche gaming scenario; not enough to be a game-changer in any capacity.

In all honesty, I think perhaps the best selling point of X570 is the Gen4 back-end: 8GB/s between SB and CPU is pretty nice if more board makers would actually wire up more than 1 NVME to the chipset, lol.

With Pice16x or 8x the gaming performance decrease is small, I doubt we will see any real use of 4.0 in GFXs...
Storage, on the other hand is a different story.

Anyways, solid board.

 

[LocutusH]

One thing that grinds my gears... if they can place nice vertical connectors for all the usb and sata on the front of the motherboard, why not the atx 24pin too? Would be much nicer for cable management...

 

[thebluebumblebee]

That VRM cooling solution looks like it might actually dissipate heat! (and looking at the graphs, it does indeed)

Looks like a solid board,

Yep, it looks like a good board for WCG!

 

[Sashleycat]

Yep, it looks like a good board for WCG!

Exactly my thinking healthy VRM is essential for 24/7, since I doubt 99% of users are going to be pumping 200W+ through the system all day long. (This is why I have a 3000 RPM top-down 140mm cooler/fan on my 3950X: blows air onto the VRM & also the solid caps that aren't touching the HS: these have significantly reduced life over a certain temp).

Oh man, what I wouldn't give for a 3990X.

Santa, please?

:3

 

[gamefoo21]

That bottom plate triggers me. It's an absolute nightmare engine if you watercool your GPU.

It has to come all off so you can get at the M2 slots. Can't just ditch it either because it's part of keeping that twin fin chipset cooler from revving the fan.

Sucks too because if you could hinge it or remove the half that goes to the M2... You could leave the nifty looking gear assembly. Sadly it's all or nothing....

This is on my X570 TaiChi... I wish I would have caught that before I ordered it, now I get to make my own passive cooler.

At least it looks like Asrock improved the VRM cooler design... The X570 version is basically two blocks of aluminum and a heatpipe with hot melt glue shoved in... Heat soaks not sinks... LoL

I am so jealous of that chipset cooler... It actually has fins and the plate sorta splits...

They also added thermal pads to the back plate to make it do something...

Man this really feels like they figured out where they made mistakes on the X570 version.

*Cries*

 

[EarthDog]

he QVL is tested in AUTO, not X.M.P. Ask G.Skill Tech, or go read his posts on the G.Skill forums.

So, anyhoo, I will report back as I get the responses

@damric - just an fyi... heard back from a memory vendor (gskill) and he confirmed the list is created by testing xmp/docp/amp speeds by both them and mobo vendors.

 

[damric]

@damric - just an fyi... heard back from a memory vendor (gskill) and he confirmed the list is created by testing xmp/docp/amp speeds by both them and mobo vendors.

Good, but I bet they didn't share exactly what voltage settings, whether they tested on a golden IMC, ect. If that's all they said then that was a generic reply from marketing, not an engineer.

 

[EarthDog]

Good, but I bet they didn't share exactly what voltage settings, whether they tested on a golden IMC, ect. If that's all they said then that was a generic reply from marketing, not an engineer.

LOL, you found this.. great, now the same discussion in two places, lol....


I talked to an engineer (through my PR rep) for both. They do not share CPU settings (SOC, etc) as that will of course vary by CPU.. you know this and that is an argument without merit. The point and context here, damric, is all about the sticks. And they test the sticks at the XMP settings - this includes voltage for the sticks. You do not need to go checking multiple QVL lists. The reason why some sticks do not work can be many... and you also know this... an IMC not capable... voltage on the CPU needs to be changed as it is 'overclocking' the IMC... bad bios, etc.

Please, let's update your knowledge data base as it doesn't appear to be accurate so far. As I said, I will report back if I hear back from other mobo vendors.

 

[damric]

LOL, you found this.. great, now the same discussion in two places, lol....


I talked to an engineer (through my PR rep) for both. They do not share CPU settings (SOC, etc) as that will of course vary by CPU.. you know this and that is an argument without merit. The point and context here, damric, is all about the sticks. And they test the sticks at the XMP settings - this includes voltage for the sticks. You do not need to go checking multiple QVL lists. The reason why some sticks do not work can be many... and you also know this... an IMC not capable... voltage on the CPU needs to be changed as it is 'overclocking' the IMC... bad bios, etc.

Please, let's update your knowledge data base as it doesn't appear to be accurate so far. As I said, I will report back if I hear back from other mobo vendors.

Sorry. I didn't know you posted this until you mentioned in other thread. For some reason it did not give me the alert. Back to other thread.

Edit: also it appears that you are mostly in agreement with me anyway so it's pointless to continue this further.

 

[EarthDog]

Edit: also it appears that you are mostly in agreement with me anyway so it's pointless to continue this further.

lol wat?

My dude....... you said QVL lists are tested on auto... so far (according to a memory vendor who says they and all mobo vendors test at XMP speeds - along with one mobo vendor confirming the same) that is BS.

The point has been made (with proof as you asked of me), making it pointless to continue. Not that we agree on much here.

I digress here as well. The posts speak for themselves.

 

[damric]

lol wat?

My dude....... you said QVL lists are tested on auto... so far (according to a memory vendor who says they and all mobo vendors test at XMP speeds - along with one mobo vendor confirming the same) that is BS.

The point has been made (with proof as you asked of me), making it pointless to continue. Not that we agree on much here.

I digress here as well. The posts speak for themselves.

Since we are posting sources. Would you mind posting a screenshot of the email from the engineer? I bet I know him, just curious if he is the one they ended up hiring when I wanted way too much money.

 

[EarthDog]

Since we are posting sources. Would you mind posting a screenshot of the email from the engineer? I bet I know him, just curious if he is the one they ended up hiring when I wanted way too much money.

This was off the record through my PR reps, so no. He asked an engineer to confirm, note. It wasn't directly with one.

Again, lets move along. If you want to continue this in PM for whatever reason, my box is open.

 

[gamefoo21]

Motherboard QVL lists are as far as I know set by what the XMP or JEDEC or whichever call for.

Like my one G.Skill kit is 4x 8GB DIMM kit running 3466 at 16-18-18-38 at 1.35v

G.Skill specifically says and only has QVL'd the kit for Intel platforms Z170 and up, but it shows up under QVL lists for Asrock, Gigabyte and MSI's X570 boards, usually under Zen 2, because 4 sticks at 3,466 is pushing your luck on Zen+

But the bios has a bunch of adjusted sub-timings to make it work at the XMP speeds. The QVL lists from the mobo makers list the stock speed 2,666(JEDEC) and they list the functioning speed 3,466(XMP).

According to Gigabyte for example on certain X570 boards, QVL'd kits are much more likely to crack 4000, the default sub timings for compatibility break high speed overclocking.

Which is why once Asrock ran the kit through the QVL process it started running. Of course once you exceed the stated supported memory speeds of the memory controller, you are not guaranteed anything. If I get a dud cpu that only does 2,666 on 4 dimms, I have to suspect the IMC.

When the board vendors QVL the halo kits, you have to expect that they are doing it on golden binned samples under better than stock cooling.

This is exemplified by the fact that a BIOS update with an updated QVL listing, is the only thing that changed between the kit not doing it's XMP speed and running XMP.

That's my understanding of it all.

 

[Deleted member 158293]

With limited PCIe slots, 1 slot used for additional nvme drives (ok with that), but no onboard 10 gigabit networking means needing another card slot. Taichi boards are my first choice, but as a TR workstation this board falls short for me... . Have to keep looking...