So the dude is blind and cant see the pins are various pretty much all way around....

Sweeeper:

So the dude is blind and cant see the pins are various pretty much all way around....

The 'dude' you are referring to is Der8auer... He is pretty much one of the best CPU overclockers in the world. I think he knows pretty much all there is to know about how these chips function.

Well, EPYC requires ECC Memory so that might be the reason why it didn't get past the memory check 🙂

Sweeeper:

So the dude is blind and cant see the pins are various pretty much all way around....

I find it funny how you can NOT see the similarities... Did AMD say that later on people would be able to run the Epyc on X399? Or did he just try out of looking at both CPU packages and noticing they'd fit into the slot?

Toetje583:

Well, EPYC requires ECC Memory so that might be the reason why it didn't get past the memory check 🙂

Hmmmz, I think EPYC can be equipped with ECC DIMMs, but it's not a mandatory requirement much like Threadripper can run ECC. EDIT: Nope I might have the wrong idea, all server configs do list ECC rather specifically.

While EPYC is for server loads...Would be nice to see a workstation mobo out there at least.

Apart from the need for ECC ram there's also a few differences with the resistor layout on both packages aswell

Though this is interesting, aside from "just for giggles" standpoint, I don't really see the point of trying this. Best case scenario, you get half the CCXs (and in turn, half the core count), half the memory channels, and half the PCIe lanes. In other words, you'd be getting a Threadripper for a hefty price. For example, if you get one of the 8-core Epycs, you're likely going to get a 4 core. Kind of weird to have an Epyc that would be degraded to be worse than a Threadripper. I think a much better use of Der8auer's time would be figuring out a way to get Coffee Lake CPUs to work on <200 series boards, or, >300 series boards on pre-CL CPUs.

Athlonite:

Apart from the need for ECC ram there's also a few differences with the resistor layout on both packages aswell

If anything, most likely a pin swap mod like was done with 771 cpu's on 775.

schmidtbag:

Though this is interesting, aside from "just for giggles" standpoint, I don't really see the point of trying this. Best case scenario, you get half the CCXs (and in turn, half the core count), half the memory channels, and half the PCIe lanes. In other words, you'd be getting a Threadripper for a hefty price. For example, if you get one of the 8-core Epycs, you're likely going to get a 4 core. Kind of weird to have an Epyc that would be degraded to be worse than a Threadripper.

you very well could get a full 32 cores on thread ripper since the dies are connected on the substrate and zepplin( ryzen die) is essentially a soc, you would lose 4 memory channels and half the pcie , but the cores would still work, the ccx's just have to access memory via the infinity fabric , which normal threadripper does when running in single channel mode.

I think a much better use of Der8auer's time would be figuring out a way to get Coffee Lake CPUs to work on <200 series boards, or, >300 series boards on pre-CL CPUs.

thats probably a dead end atm , the me firmware is the limiting factor, unless someone manages to patch it ( basically impossible without intel's keys) or trick newer z370 me firmware to run on z270 /z170, not gonna happen with the 6 core chips anytime soon. the 4 core coffee lake chips do boot on z170 with microcode , since it appears they are just rebranded kabylake chips, they will even boot in ddr3 mobos.

user1:

you very well could get a full 32 cores on thread ripper since the dies are connected on the substrate and zepplin( ryzen die) is essentially a soc, you would lose 4 memory channels and half the pcie , but the cores would still work, the ccx's just have to access memory via the infinity fabric , which normal threadripper does when running in single channel mode.

True, I suppose that makes sense. But, you have to also consider what the BIOS is programmed to recognize, which adds a layer of challenges.

thats probably a dead end atm , the me firmware is the limiting factor, unless someone manages to patch it ( basically impossible without intel's keys) or trick newer z370 me firmware to run on z270 /z170, not gonna happen with the 6 core chips anytime soon. the 4 core coffee lake chips do boot on z170 with microcode , since it appears they are just rebranded kabylake chips, they will even boot in ddr3 mobos.

Well, much like I just said, getting Epyc to work on a Threadripper motherboard is also likely limited by firmware. So when you consider that's the only limitation, you can think of the situation like this: A. TR and Epyc are the same generation, they share the same physical socket, and are largely capable of the same things. However, TR physically has less components to it (affecting it's functionality) and is designed for enthusiasts, while Epyc is meant for servers. B. When it comes to KL and CL, they may be different generations, but even companies like Asus admitted that there wasn't really any significant difference in how the 2 generations work. In other words, the broken compatibility is purely artificial, and the only thing preventing CL from working on Z170 is because that's what Intel wants. As you mentioned, the 4-core CL chips will boot on older boards (and presumably will boot on newer boards too) proving backward and forward compatibility is possible. Another thing to consider is who can be affected. Though it is VERY interesting to get an Epyc to run on a TR motherboard, the novelty pretty much ends right there. It is way too expensive for the average enthusiast to attempt such a thing. Meanwhile, there are likely thousands of people who already own a 100 or 200 series Intel motherboard, where a 6-core CL could be a substantial upgrade to many of them. If all it took was a firmware tweak, such a mod would be very beneficial to a large community. EDIT: It'd be interesting if someone got something like Coreboot to work on some of these motherboards (TR4 or 1151) to accept all of these CPUs. We likely won't be seeing that for 20 years though...

Even if its posible, need only to modificate bios microcode, memory updates, any other stuff don't need to be mod since the boards using different stuff, like bios information, audio, lan etc. you can't just get bios from another motherboard and flash into it and expect to work, worst case u will end up break the board.

p1stov:

you can't just get bios from another motherboard and flash into it and expect to work, worst case u will end up break the board.

Assuming they have multiple EEPROMs (which I'd be surprised if they didn't) the chances of you bricking the board are nearly impossible. But yeah, you probably wouldn't be able to take the BIOS from another board (even from the same manufacturer) and unless you're really good with a hex editor, I don't think there's much anyone could do to edit these. Maybe if this was back before the EFI days it could be done, but with today's firmware, forget about it.

schmidtbag:

True, I suppose that makes sense. But, you have to also consider what the BIOS is programmed to recognize, which adds a layer of challenges. Well, much like I just said, getting Epyc to work on a Threadripper motherboard is also likely limited by firmware. So when you consider that's the only limitation, you can think of the situation like this: A. TR and Epyc are the same generation, they share the same physical socket, and are largely capable of the same things. However, TR physically has less components to it (affecting it's functionality) and is designed for enthusiasts, while Epyc is meant for servers. B. When it comes to KL and CL, they may be different generations, but even companies like Asus admitted that there wasn't really any significant difference in how the 2 generations work. In other words, the broken compatibility is purely artificial, and the only thing preventing CL from working on Z170 is because that's what Intel wants. As you mentioned, the 4-core CL chips will boot on older boards (and presumably will boot on newer boards too) proving backward and forward compatibility is possible. Another thing to consider is who can be affected. Though it is VERY interesting to get an Epyc to run on a TR motherboard, the novelty pretty much ends right there. It is way too expensive for the average enthusiast to attempt such a thing. Meanwhile, there are likely thousands of people who already own a 100 or 200 series Intel motherboard, where a 6-core CL could be a substantial upgrade to many of them. If all it took was a firmware tweak, such a mod would be very beneficial to a large community. EDIT: It'd be interesting if someone got something like Coreboot to work on some of these motherboards (TR4 or 1151) to accept all of these CPUs. We likely won't be seeing that for 20 years though...

I see why you would want to focus on cf lake , but if intel blocks something on purpose via sofware, It becomes almost impossible, prime example is bclk oc on z170/z270/z370, bclk oc is perfectly possible on any chip , but intel intentionally restricts it to k skus after the skylake fiasco, no one has managed to break the locks and its been 2 years since the locks were first implemented, getting bclk oc is arguably a simpler task than getting a blocked proc to post on an usupported board that lacks proper firmware altogether. at least with epyc and x399 you can get a partial post, which is a whole lot more than you can get with 6core cf lake on older boards atm