EPYC almost runs on X399

[Skanky Sylveon]

23 minutes ago, ARikozuM said:

A car with a starter that rotates itself, but does not start the car's ignition, is not almost starting. It's quite literally not starting. 

That is brilliant.

 

Edit: Although saying that the car is almost starting, is better then saying it's starting.  Since it's not.

[Jito463]

1 hour ago, Cinnabar Sonar said:

der8auer stated in his video that EPYC's pinout is identical to Threadripper's

In the video, what he says is that "because it boots partially it has to be identical".  If that was all he was basing it on, that would have been an error to make such an absolute statement.  It wasn't until I called him on that in the YT comments, that he mentioned checking the actual pin layout from AMD.

[Skanky Sylveon]

1 minute ago, Jito463 said:

In the video, what he says is that "because it boots partially it has to be identical".  If that was all he was basing it on, that would have been an error to make such an absolute statement.  It wasn't until I called him on that in the YT comments, that he mentioned checking the actual pin layout from AMD.

Either way, the pinout is still identical, correct?

[Jito463]

Just now, Cinnabar Sonar said:

Either way, the pinout is still identical, correct?

According to Der8auer.  I wonder why he didn't mention checking the pinout from AMD in the video?

[Skanky Sylveon]

3 minutes ago, Jito463 said:

According to Der8auer.  I wonder why he didn't mention checking the pinout from AMD in the video?

I agree.  Perhaps he didn't want to generate anymore ire from AMD?

AMD already doesn't seem pleased from his findings.

[mr moose]

3 hours ago, Cinnabar Sonar said:

@Drak3 @mr moose

Can we at least agree that this thread's title is better then the article's?

 

 

I can, I'm pretty sure it's just a title and no one is going to lose their job or be litigated for it.

 

3 hours ago, ARikozuM said:

A car with a starter that rotates itself, but does not start the car's ignition, is not almost starting. It's quite literally not starting. 

In this instance it turns the motor but nothing happens.

 

 

As far as I am concerned (coming from an EE education) If you feed a bunch of silicon 90 pixie watts of magic juice and the thing doesn't go Chernobyl on you then you are half way there.  And that''s pretty good for something complex like a CPU.  For all we know the only thing required to go from here to fully working could be a bios hack.  

[Skanky Sylveon]

5 minutes ago, mr moose said:

or be litigated for it.

DON'T give anyone ideas.  

[BluJay614]

1 hour ago, VegetableStu said:

@BluJay614 update from der8auer: nope. ECC didn't work either. experiment will nevertheless carry on

 

 

I know. I saw the new video where he showed the dies that where active, and pointed out that he couldn't get the dies hot enough on Epyc to show the active dies. I haven't updated as I've actually hoped the thread would be shut down(thank the people who seem intent on bashing everything)

[Drak3]

On 2/2/2018 at 7:47 PM, Cinnabar Sonar said:

 

Edit: Although saying that the car is almost starting, is better then saying it's starting.  Since it's not.

No, it's not. It is still a complete inaccuracy. It did not start.

 

EPYC does not work on X399.

 

Until we see BIOS mods that get EPYC wprking, assuming that it does share a pinout with Threadripper, that's the end of the story. No 'almost.'

[handymanshandle]

On 1/29/2018 at 2:00 PM, BluJay614 said:

Agreed..... Though I guess you could say shits and giggles, or home server with something like the ASRock X399 Professional Gaming

What kind of home server needs more than 16c/32t? 

[BluJay614]

2 minutes ago, wcreek said:

What kind of home server needs more than 16c/32t? 

The kind for bragging rights along with shits and giggles. It's a "because I can" land type thing, if they got it fully functional.

[TopHatProductions115]

If it actually did happen (EPYC somehow working on X399), imagine how bad it would be. AMD would probably start modifying their CPU sockets again...  

[Drak3]

1 hour ago, wcreek said:

What kind of home server needs more than 16c/32t? 

Multiple modded Minecraft intance servers.

[TopHatProductions115]

12 minutes ago, Drak3 said:

Multiple modded Minecraft intance servers.

With Skyrim running in the background, @1440p

Edited February 6, 2018 by TopHatProductions115

[Jito463]

1 hour ago, TopHatProductions115 said:

If it actually did happen (EPYC somehow working on X399), imagine how bad it would be. AMD would probably start modifying their CPU sockets again...  

Perhaps, but I doubt we'd see too many people who would do it.  I'm sure there would be some, but the extra cost of Epyc chips vs Threadripper chips would deter most people.  That would probably change once Epyc has been out a few years and one could pick up the older chips second-hand, but by that point AMD will have newer stuff out.

 

Having said that, I am a bit surprised AMD didn't add an extra notch (or remove one) to prevent installation either way.  It wouldn't take a complete modification of the socket, just a slight alteration to physically impair installation.

[TenThousand]

Why would you even want to? TR4 boards aren't wired for 128 pcie lanes. All you'll have is a gimped processor that can't really be overclocked.

[Jito463]

1 hour ago, TenThousand said:

Why would you even want to? TR4 boards aren't wired for 128 pcie lanes. All you'll have is a gimped processor that can't really be overclocked.

The only advantage would be the extra core count, and if you really need the core count that badly, it's not much more (comparatively speaking) to get a server board over a TR board.

[mr moose]

11 hours ago, Drak3 said:

No, it's not. It is still a complete inaccuracy. It did not start.

 

EPYC does not work on X399.

 

Until we see BIOS mods that get EPYC wprking, assuming that it does share a pinout with Threadripper, that's the end of the story. No 'almost.'

Must be pretty awful living in such a black and white world.

 

I am pretty sure you are confusing the word "almost" with the word "does", or even the concept of eventualities. Almost doesn't have any future eventualities tied to it's meaning, so even if he never gets it working, the word "almost" by definition is still accurate due to the nature of language and the context it is used in. 

 

 

[Drak3]

5 minutes ago, mr moose said:

Must be pretty awful living in such a black and white world.

 

I am pretty sure you are confusing the word "almost" with the word "does", or even the concept of eventualities. Almost doesn't have any future eventualities tied to it's meaning, so even if he never gets it working, the word "almost" by definition is still accurate due to the nature of language and the context it is used in. 

 

 

https://www.google.com/search?q=almost&oq=almost&aqs=chrome..69i57j69i60l3j0l2.1613j0j4&client=ms-android-att-us&sourceid=chrome-mobile&ie=UTF-8

 

This is not close to working. To say it almost worked is a lie.

[mr moose]

2 minutes ago, Drak3 said:

https://www.google.com/search?q=almost&oq=almost&aqs=chrome..69i57j69i60l3j0l2.1613j0j4&client=ms-android-att-us&sourceid=chrome-mobile&ie=UTF-8

 

This is not close to working. To say it almost worked is a lie.

 

You do realise that to get to memory initialization in the post the CPU has to load the bios, check itself and check the bios code and verify a few basic parameters on the motherboard?  Are you saying none of this happens, because to my way of thinking that if a CPU can self check, and load the bios without destroying itself then it is actually working to some degree.  The fact it makes it to memory is sufficient evidence that hardware issues are less likely to be the cause of it not making it through to loading the OS.

[Skanky Sylveon]

11 hours ago, BluJay614 said:

I haven't updated as I've actually hoped the thread would be shut down(thank the people who seem intent on bashing everything)

The mods aren't going to shut down this thread just because someone disagrees with you.

11 hours ago, wcreek said:

What kind of home server needs more than 16c/32t? 

A gaming server perhaps?

Although a gaming server would benefit from the extra PCIE lanes that EPYC provides, as well as the greater RAM potential.  So I would recommend using an EPYC CPU with an EPYC motherboard for that application still.

9 hours ago, Jito463 said:

Having said that, I am a bit surprised AMD didn't add an extra notch (or remove one) to prevent installation either way.  It wouldn't take a complete modification of the socket, just a slight alteration to physically impair installation.

I believe that I heard AMD stating they were saving on costs by using the same socket.  But I don't remember where I heard that, so take the information with a grain of salt.

3 hours ago, TenThousand said:

Why would you even want to? TR4 boards aren't wired for 128 pcie lanes. All you'll have is a gimped processor that can't really be overclocked.

The only reason that I can see is the cost savings.

I'm sure that EPYC motherboards cost more.

11 minutes ago, mr moose said:

You do realise that to get to memory initialization in the post the CPU has to load the bios, check itself and check the bios code and verify a few basic parameters on the motherboard?  Are you saying none of this happens, because to my way of thinking that if a CPU can self check, and load the bios without destroying itself then it is actually working to some degree.  The fact it makes it to memory is sufficient evidence that hardware issues are less likely to be the cause of it not making it through to loading the OS.

der8auer's newest video shows that EPYC is receiving power and "turning on" when in a TR4 motherboard.

While I wouldn't use the words "almost working" it's not the worst choice of words.

At least @BluJay614 didn't state that it was working, unlike the article that he posted.

[Drak3]

8 minutes ago, mr moose said:

 

You do realise that to get to memory initialization in the post the CPU has to load the bios, check itself and check the bios code and verify a few basic parameters on the motherboard?  Are you saying none of this happens, because to my way of thinking that if a CPU can self check, and load the bios without destroying itself then it is actually working to some degree.  The fact it makes it to memory is sufficient evidence that hardware issues are less likely to be the cause of it not making it through to loading the OS.

That's the most effort put into "it doesn't fucking work" I've ever seen.

 

Because despite the tests reporting an 'OK,' it still, and I know you'll love this little fact, doesn't fucking work.

[Jito463]

3 minutes ago, Cinnabar Sonar said:

I believe that I heard AMD saving on costs by using the same socket.  But I don't remember where I heard that, so take the information with a grain of salt.

I read the same thing, and it makes complete sense.  Still, it wouldn't take much for them to physically block installation either way.  Just as with the Intel sockets, there are several small notches on either side of the CPU.  Just moving/removing/adding one notch would make them physically incompatible without making any major alterations to the socket itself.

 

 

If you look on the two opposing sides just above the middle and towards the bottom, those notches there are what I'm referring to.  Just modifying one of those would be sufficient.

[Drak3]

1 minute ago, Jito463 said:

I read the same thing, and it makes complete sense.  Still, it wouldn't take much for them to physically block installation either way.  Just as with the Intel sockets, there are several small notches on either side of the CPU.  Just moving/removing/adding one notch would make them physically incompatible without making any major alterations to the socket itself.

 

 

If you look on the two opposing sides just above the middle and towards the bottom, those notches there are what I'm referring to.  Just modifying one of those would be sufficient.

Alternatively, they could have made the sled wider on EPYC.

[Skanky Sylveon]

@Jito463 It does make you wonder why AMD didn't do that in the first place.