AMD Epyc 7000 series revealed

[VerticalDiscussions]

Holy balls, AMD is moving at the speed of light when it comes to... Innovation!!

 

Intel = Ded meat if this keeps going on xD.

[ravenshrike]

Just now, KOMTechAndGaming said:

ive never had any problems with vmware or virtual box

Most of them involved graphics card pass through and splitting the output of a single graphics card among multiple VMs IIRC. Not sure as I didn't pay much attention to the issues.

[MorganO]

Just now, ravenshrike said:

Most of them involved graphics card pass through and splitting the output of a single graphics card among multiple VMs IIRC. Not sure as I didn't pay much attention to the issues.

ah ok

thanks

[Jito463]

3 minutes ago, KOMTechAndGaming said:

ive never had any problems with vmware or virtual box

 

Most end users wouldn't run into the issues, as it involved specialized scenarios that would typically only be encountered in an actual server environment (something Ryzen was never intended for anyway).  There was a reported issue with getting XP working in a VM, but there were workarounds for it, and it was - as I recall - unrelated to what @ravenshrike was referring to.

[MorganO]

Just now, Jito463 said:

Most end users wouldn't run into the issues, as it involved specialized scenarios that would typically only be encountered in an actual server environment (something Ryzen was never intended for anyway).  There was a reported issue with getting XP working in a VM, but there were workarounds for it, and it was - as I recall - unrelated to what @ravenshrike was referring to.

ah ok ,thanks for the information

[Humbug]

57 minutes ago, Taf the Ghost said:

It'll be a while before AMD makes too much progress in the Server market, but there'll be room.  That's a LOT of cores at a much lower, generally, price than Intel offers. The important part about Intel's server stack is that they have a huge list of server products available by list.  And all of them can be semi-custom bought as well. An absolutely massive "get" for AMD is if Amazon, Google or Facebook are interested in buying volume.

 

I also expect Linus to get a 2U server and OC it to 3.9 Ghz all-core. HAHA.

Even if AMD can get back 5% server market share (they will probably get a bit more) from Intel that is very big money for AMD.

[Sauron]

16 cores for less than 800$. 24 cores for around 2000. GG Intel.

[Taf the Ghost]

11 minutes ago, Humbug said:

Even if AMD can get back 5% server market share (they will probably get a bit more) from Intel that is very big money for AMD.

Yup, and the profit margin is also really high in that market. Those $2000 or higher CPUs probably aren't costing more than $150 to produce. They don't need high volume to make a lot of money on them. 

[PCGuy_5960]

2 hours ago, KOMTechAndGaming said:

in what way? 

Nvm, because apparently it's fixed now

[MorganO]

Just now, PCGuy_5960 said:

Nvm, because apparently it's fixed now

Ah ok, I was just curious as I use VMware/virtual box Albeit not regularly

[PCGuy_5960]

2 hours ago, NumLock21 said:

Skylake-E will not be cheap.

And will EPYC be cheap? Nope, both are expensive, because they are server grade parts. And those who buy server CPUs don't really care about price/performance

[PCGuy_5960]

Just now, KOMTechAndGaming said:

Ah ok, I was just curious as I use VMware/virtual box Albeit not regularly

I think they patched it with the latest microcode update, so if you update your BIOS you should be fine

[PlayStation 2]

Would make for part of a fucking amazing rendering rig.

[Master Disaster]

This is all great, really good stuff from AMD however I have reservations, not about AMDs hardware but about AMDs ability to penetrate the server market at all.

 

The last time AMD was relevant to server stuff was Opteron and even then they failed to put a dent into Intel's market dominance and now Intel have had 10+ years of complete and total dominance (like AMD haven't even had a competitor to Xeon) I struggle to imagine a scenario where many (if any) server farms are going to ditch their entire infrastructure and move over to a brand new and largely untested platform, especially in the one space where rock solid stability is a necessity.

 

I'm sorry but I really don't see these CPUs having much of an impact on Intel's sales at all.

 

I'd be very interested to know if @LinusTechwould consider swapping any of his servers over to AMD, given how much he stressed workstation stability in the recent editing machine update video I'm sure I could guess what his answer would be and he's only serving one room full of editors, imagine having to serve thousands of people.

[cj09beira]

4 hours ago, djdwosk97 said:

 

Why is virtualization bad? Are there pass through issues?

before agesa 1.0.0.6 you can't give vms a pci-e card 

[The Benjamins]

11 minutes ago, Master Disaster said:

 

I'd be very interested to know if @LinusTechwould consider swapping any of his servers over to AMD, given how much he stressed workstation stability in the recent editing machine update video I'm sure I could guess what his answer would be and he's only serving one room full of editors, imagine having to serve thousands of people.

He kind of already answer that, he sticks to being about 1 gen behind to maintain stability so all the kinks are ironed out before he uses it. (for his editors) he said he would use it on his (possible also luke's) machines at work.

 

I do think their is a market for AMD but it will take time to gain trust and to roll out very stable setups.

 

EDIT: oh not sure about servers. I feel he will do a project or 2 before thing of a server upgrade so he could try things out. he was talking about a possible virtualizing the work station on Epyc if he could get backing for that.

[DXMember]

the interesting thing is where they compare the lower tier 16 core EPYC to >$600 Xeon - would that be of any indication for what we can expect the Threadripper to be priced at?

[DXMember]

5 hours ago, The Benjamins said:

This basicly confirms they will always use a balanced setup so eypic has to be divisible by 8 and Thread ripper has to be divisible by 4.

there's a 10 core Threadripper with 3+2 + 3+2 CCX configuration

as well as 14 core Threadripper with 3+4 + 3+4

[The Benjamins]

Just now, DXMember said:

there's a 10 core Threadripper with 3+2 + 3+2 CCX configuration

no info officially.

officially alienware is only selling a 12 and 16 core version, why would they not sell 10 and 14 if it was a option.

I really really don't think we will see a 10 or 14 core.

[aakopa]

I would love to see an 8-core threadripper for like 500$, those pcie lanes would be really nice to have.

[DXMember]

5 hours ago, PCGuy_5960 said:

Those CPUs can't fit on a small board, they are too big for this

no worries, leave it to ASRock to come up with a mini ITX board that uses SODIMM

I think I saw one at Computex... or was it for Skylake-X, anyway, ASRock will find a way - I'm calling it

[THE_maverick]

CONFIRMED!!! #ThreadRipper 16/32 will be 750, as the pro counterpart is going to be around 850.

[cj09beira]

1 minute ago, aakopa said:

I would love to see an 8-core threadripper for like 500$, those pcie lanes would be really nice to have.

well if the 10/12 (lowest one) core is cheap, might as well get that

[DXMember]

5 hours ago, The Benjamins said:

ok thats why I said $1000, and it makes the rumors of the entry 16c at $850 make a lot of sense.

also isn't is a lot easier to get away with high premiums on the server platform then enthusiast.

What makes you think Threadripper will cost VASTLY more then Epyc?

yields

to make a 16 core EPYC you'd need four half dead dies, which are easy to get

to make a 16 core Threadripper you'd need two fully functional dies which are comparably harder to get

 

but then again... there's less silicon for Threadripper so going out of purely material cost it should be more than double the price in material to make EPYC and to trace the interposer

[PCGuy_5960]

1 minute ago, DXMember said:

no worries, leave it to ASRock to come up with a mini ITX board that uses SODIMM

I think I saw one at Computex... or was it for Skylake-X, anyway, ASRock will find a way - I'm calling it

It was X299. I don't think it's possible to make a mini ITX Threadripper board, I mean, just look at the size of that socket!