Rumor Possibility of AMD on STX form factor

There are so many frieds discussed about AMD solution on STX form factor.
I told Crew ASRock had one AMD FM2+ 5x5 board....
Let me show off one board I found on my boss's table.

We did a very very quick research during CNY.
AMD 2200G box fan is so high that can't be installed into our chassis.
BUT... how about install CPU fan and heatsink separately? HAHAHA...

Only for fun

 

[Ross Siggers]

Changing the DC barrel to an internal 4-pin for the HDPlex 160W AC-DC should be a simple de-solder of the existing part & re-solder the new part...?

Also curious to know this... >_>

 

[Windfall]

Also curious to know this... >_>

Even simpler, just wire two cables, solder to the underside of the board, and boom.

 

[Ross Siggers]

Even simpler, just wire two cables, solder to the underside of the board, and boom.

For serviceability, I'd like to still put a plug on, if I knew how. Like a 4-pin EPS

 

[Kmpkt]

The only B-Die SO-DIMMs I'm aware of are the 3466 kits from late 2017 (3466 MHZ, 17-17-17-37) and the slightly newer 3200/3600/3800 MHz kits (16-16-16-36 ot 18-18-18-38) G.Skill released specifically for the AsRock X299 board. While the kit in question may very well be B-Die I wasn't able to find any information from any manufacturer that specified any other kit as being so. I would have preferred to have bought the 16-16-16-36 BDie kit because of price, but couldn't find it nor the 3600 kit anywhere.

 

[peak_mu]

Here's the Thaiphoon Burner for F4-3200C16D-16GRS.

It also struck me as odd that the manufacturing date is so new. If that's accurate, this particular kit was born only a couple weeks ago!

 

[peak_mu]

On a related note, the Dell power supply from my Precision 5510 (similar to XPS 15) looks like it would be a prime candidate for integration into a custom brickless case. It's 130W, slim, and matches the depth and width of the Deskmini very well. Approx dimensions 5.63" x 2.63" x 0.88"

I searched but couldn't find anything already on the forums...all the hits I got were referencing Dell's other rectangular type bricks. I know it's got a 3 pin connector so the laptop communicates with it, so I'm not sure if it'll provide power to a non-Dell system. I just ordered the XPS dc jack, so I guess I'll find out shortly. Also on order is a power meter. I'll provide some power figures when it comes in unless someone else comes through.

 

[Jello]

For anyone who's running this board, have you done any benchmarks to see what kind of performances increases there are when increasing the GPU ram allocation, or does someone have a link to some results?

 

[Zackmd1]

So question for those with the A300 in hand... Can anyone confim / know if the top m.2 port (the one over the m.2 wifi) is a direct pcie x4 to the cpu or does it go through the chipset? I might be tweaking my plans and going with this STX board (now that I see it is in stock at newegg) and connecting a GPU through the m.2 port. I would prefer an m.2 port with a direct pcie connection to the cpu and not the chipset for this function.

 

[Boil]

So question for those with the A300 in hand... Can anyone confim / know if the top m.2 port (the one over the m.2 wifi) is a direct pcie x4 to the cpu or does it go through the chipset? I might be tweaking my plans and going with this STX board (now that I see it is in stock at newegg) and connecting a GPU through the m.2 port. I would prefer an m.2 port with a direct pcie connection to the cpu and not the chipset for this function.

I believe there is no real chipset, it is the I/O built into the APU itself...?

But since there is no GPU slot, I also want to say both of the M.2 SSD slots are NVMe PCIe x4 straight to the APU (when using a Ryzen APU)...?

 

[Zackmd1]

I believe there is no real chipset, it is the I/O built into the APU itself...?

But since there is no GPU slot, I also want to say both of the M.2 SSD slots are NVMe PCIe x4 straight to the APU (when using a Ryzen APU)...?

That is what I am not sure about... I know a few intel boards have PCIE lanes that are routed through the chipset instead of directly to the CPU. I am not sure if AMD does the same or if all PCIE lanes are direct to the CPU.

 

[peak_mu]

Here's the PCI bus layout from HWinfo64. If I'm reading this correctly, it looks like it's going directly to the CPU. It's interesting to see that they're being reported as hot-plug capable.

 

[Zackmd1]

Am I reading that correctly where it says 8x? M.2 is only gen 3.0 4x right? Is that just what the CPU has allocated to the port even though the port can only support 4x?

 

[peak_mu]

I would think both the top and back side M.2 ports are on the same bus, therefore 8x.

 

[W4RR10R]

I would think both the top and back side M.2 ports are on the same bus, therefore 8x.

Does this mean that you could theoretically adapt both m.2 ports into a single pcie gen 3 x8 slot? You would lose nvme drive support but an effectively unrestricted gpu and a sata ssd may be worth it. Also if 3rd gen ryzen allows gen 4 speed on non chipset pcie slots (ie all the pcie on this board) would pcie gen 4 x4 be ~ pcie gen 3 x8? Does pcie scale up, I know it scales back.

Also, from some digging I've done, everything suggests that even high end gpus (not RXT I have found data) would only lose ~10% performance on a pcie gen 3 x4 connection, and slightly less as you increase resolution. So would attempting to adapt both m.2s even be worth it?

 

[Zackmd1]

Not quite sure if that would be possible. With that said, even at 4x a GPU does not take a massive performance hit at all. In many cases its only around a 2-5fps hit. Take a look at this link.

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1080_PCI_Express_Scaling/3.html

 

[W4RR10R]

That's one of the articles I pulled my data from. I used that one a the one on the r9 290 or fury (can't remember which) I then used the userbenchmark ratings of those cards and vega 56 to interpolate an estimated performance loss. I intend to use this board a 2600 and a vega 56 in my build. I can test everything tomorrow. My deskmini is sitting in a post office 8 miles from my house to be delivered tomorrow.

 

[Zackmd1]

That's one of the articles I pulled my data from. I used that one a the one on the r9 290 or fury (can't remember which) I then used the userbenchmark ratings of those cards and vega 56 to interpolate an estimated performance loss. I intend to use this board a 2600 and a vega 56 in my build. I can test everything tomorrow. My deskmini is sitting in a post office 8 miles from my house to be delivered tomorrow.

Good luck on the build! If I were to go this route I would be paring a Ryzen 3 3300 (6c/12t processor rumored) with a GTX 1660Ti.

 

[W4RR10R]

Well at the moment I don't believe there is a 1660 ti that fits the incredibly small GC shell (~160 mm external depth) The vega nano pcb just squeezes in, I'm currently working on a custom air cooler for it. And for 3xxx processor I would likely go for the 3600 8c/16t, I can always upgrade later.

 

[Zackmd1]

Well at the moment I don't believe there is a 1660 ti that fits the incredibly small GC shell (~160 mm external depth) The vega nano pcb just squeezes in, I'm currently working on a custom air cooler for it. And for 3xxx processor I would likely go for the 3600 8c/16t, I can always upgrade later.

Sorry, should have worded that a bit differently! I am planning on building a new SFF and since this board is in stock finally I will likely be paring it with a ryzen 3 and 1660.

In terms of your build, the nano is likely your only option on the AMD side (although power hungry for sure). Only option on Nvidia side at this point is last gen. That is of course assuming that the non Ti 1660 will not have a smaller board design. That should be arriving by the end of this month and will likely consume much less power and produce much less heat then the vega nano with similar gaming performance.

 

[Kmpkt]

I think this is how the PCIe lanes on Ryzen are supposed to part out. Where AM4 differs from Intel is it gives 4 lanes to M.2 straight from the CPU as well as 4 lanes to the "Chipset". Intel on the otherhand has all M.2 drives serviced from the 4 lanes provided to the chipset. What this means (hypothetically) for M.2 GPUs is that on AMD platform you should be able to have a clear and dedicated x4 connection to the CPU for a GPU while on an Intel platform you're always going to have to give up some overhead for SATA, a second M.2 or your USB devices.