• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

64-Core Threadripper and ultra-HEDT X599 Coming End of 2019

mikeo said:
Yeah, that's what I do now, but in the past with nvenc and intel quick sync the quality was trash.
Click to expand...

I agree, and I do hope that they continue to improve, in terms of quality as well as flexibility. I'd love to see the full range of codec quality settings be available and be able to choose based on output need and resource usage, and I'd love the option in the hardware and drivers to expose a 'hybrid' option to software such that any and all fix-function, CPU, and GPU resources can be used in parallel as desired.
 
  • Like
Reactions: mikeo
like this
This is classic industrial misdirection. By announcing this, AMD is forcing Intel to go with a craaaaaazy 128 core/256 thread design. They will expend their hordes treasure on it, roll it out, exhausted, and AMD will keep flooding the market with Zen 2. Practicing The Art of War, as preached by Zen Two.

;)

/s. I am impressed and would love to have one.
 
c3k said:
This is classic industrial misdirection. By announcing this, AMD is forcing Intel to go with a craaaaaazy 128 core/256 thread design. They will expend their hordes treasure on it, roll it out, exhausted, and AMD will keep flooding the market with Zen 2. Practicing The Art of War, as preached by Zen Two.

;)

/s. I am impressed and would love to have one.
Click to expand...

Patience. You're time to upgrade will come, young padawan.
 
What this news means to me: I get to watch LTT, Paul, Kyle and Jay all run the same rendering benchmarks that will now display onscreen as quickly as their respective jpeg porn collections do when them cameras are off......meanwhile, GTA V might pick up 4-12 FPS if you're running a GPU with enough overhead......
 
HeadRusch said:
What this news means to me: I get to watch LTT, Paul, Kyle and Jay all run the same rendering benchmarks that will now display onscreen as quickly as their respective jpeg porn collections do when them cameras are off......meanwhile, GTA V might pick up 4-12 FPS if you're running a GPU with enough overhead......
Click to expand...

Hahahahahaha, so true.
 
HeadRusch said:
What this news means to me: I get to watch LTT, Paul, Kyle and Jay all run the same rendering benchmarks that will now display onscreen as quickly as their respective jpeg porn collections do when them cameras are off......meanwhile, GTA V might pick up 4-12 FPS if you're running a GPU with enough overhead......
Click to expand...

Speaking of LTT, Linus had an interesting take on this on Friday's WAN show. There's not a great reason to put this out there in almost any market (mainstream, workstation, or server, because... EPYC); there's no real costumers for this because it's so wildly out of line with those groups needs/constraints, whether it's cost or heat or whatever other compromises, it's just "extra" as the kids say these days. But here's the thing: if they release an updated 32 core TR, it will probably best Intel's 28 cores, although not by much. But... if the ryelease a 64core TR, and just smash all the multithreaded records in benchmarks, it gives them an undisputed halo product that Intel can't match, and even if no one buys it except Tony Stark... er... I mean Elon Musk, they will see a big bump in sales.
 
tempertantrum said:
There's not a great reason to put this out there in almost any market (mainstream, workstation, or server, because... EPYC)
Click to expand...

This depends entirely on how well it scales with current applications on current operating systems. TR 1 and TR 2 were a bit of a fail at 24c and 32c on Windows due to scheduling challenges; they regularly ran slower.

Given that both a) AMD has completed significant work to get the Zen / Zen+ architectural challenges out of the way and b) software developers have been working to bridge the gap, including Microsoft, I expect a 64c TR to be of more utility to consumers upon availability than the TR 2990WX was at 32c.
 
I just had a frightening thought. If the prices we've seen on X570 motherboards turn out to be true, X599 boards could easily crack $1,000. :eek:
 
Dan_D said:
I just had a frightening thought. If the prices we've seen on X570 motherboards turn out to be true, X599 boards could easily crack $1,000. :eek:
Click to expand...

Jeez.

AMD's lower CPU pricing may - at some price points - be nullified by motherboard prices.

When I bought my $600 i7-3930k and $400 Asus P9x79 WS in 2011 I thought the pricing was nuts. In the end it wasn't too bad as I've kept them in my main rig longer than any previous CPU and motherboard I've owned, but wasn't AMD becoming competitive again supposed to drive pricing DOWN? :p

Also, I wonder if there is any truth to the rumor of a higher end x590 chipset becoming available for AM4 boards. That would also make for more expensive boards, I'd imagine.
 
You know, you don't have to buy the $700 X570 mobos just because they exist.
The cheapest X570 mobo is going to sell for ~$160 according to reports. The prices on the really high end mobos are of course exorbitant, but before X570 there were no really high end AM4 mobos.

Also if you don't like paying for X570/X599, you can still run it in existing B450/X470/X399/etc. mobos (for X399 not officially confirmed but likely).
 
ChronoDetector said:
I wonder if this new chipset will support more than quad channel memory.
Click to expand...

I doubt it, but I am sure we will see some big clock increases on the memory akin to the rest of Zen 3. Quad channel at 4400+ mhz with an efficient memory controller should do the trick.
 
Back in the old days when you bought a new game and you didn't have a 3D accelerator there was an option to play the game in software render mode.

How long before there is enough extra parallel CPU power to render via the extra cores and not need 3d acceleration?
 
odditory said:
"Now it's just getting stupid. And really, what's the point? They've gone crazy with cores (I call her Crazy Lisa) and remember that Cyberpunk will only use four to eight".
--Intel VP
Click to expand...

You are not the targeted audience, these are specifically designed for professional users like me which is a godsend honestly having access to that many cores.
 
IdiotInCharge said:
This would be like trying to run ray tracing on a 1050Ti, but even slower.
Click to expand...
Yeah, just try running any modern game on mesa's software fallback on linux for an idea. Will have to set a few env variables to bypass opengl version checks, some features wont work, and it'll still be slow. Some games are almost playable, though.
 
It's hard to guess where this increase in core counts is going to end up. With clock speed decreasing with die shrinks and IPC gains seemingly being less every generation this could get fun to watch. My compute needs have never been so far below the best of the best. I'll be debating an upgrade from a quad core to truly ridiculous core counts for gaming and IT work. I got a stutter in an open world environment in Metro Exodus earlier this week. That was the first stutter I've noticed since purchasing a 1080 ti on launch day along with a 7700k.
 
ochadd said:
It's hard to guess where this increase in core counts is going to end up. With clock speed decreasing with die shrinks and IPC gains seemingly being less every generation this could get fun to watch. My compute needs have never been so far below the best of the best. I'll be debating an upgrade from a quad core to truly ridiculous core counts for gaming and IT work. I got a stutter in an open world environment in Metro Exodus earlier this week. That was the first stutter I've noticed since purchasing a 1080 ti on launch day along with a 7700k.
Click to expand...

Clock speeds are not decreasing in the consumer space. They are going up. Actually, they are going up everywhere on all but the highest end core count CPU's like Intel's Platinum 9282 etc. Those have really low clocks, but they have an insane number of cores.
 
Dan_D said:
Clock speeds are not decreasing in the consumer space. They are going up. Actually, they are going up everywhere on all but the highest end core count CPU's like Intel's Platinum 9282 etc. Those have really low clocks, but they have an insane number of cores.
Click to expand...

My understanding as a layman is that the smaller the electrical path the less voltage and ghz a processor can have. There are optimizations but the fundamental restraint remains.
 
ochadd said:
My understanding as a layman is that the smaller the electrical path the less voltage and ghz a processor can have. There are optimizations but the fundamental restraint remains.
Click to expand...
It may be your understanding, but aren’t basically all Zen 2 parts clocked same or higher than their comparative Zen+ counterparts? And isn’t part of this bump attributable to the die shrink to 7nm?
 
cdabc123 said:
I guess I just dont see any point in a many core chip like when it very much acomplishes the exact same thing as many core server chips do.



tdp is normally less restricted on servers. I have server sockets that can push 250w+ so they can fit more on the die. cooling is also less limited on servers because you can just strap a jet engine to it and call it a day. There should be no reason the epyc chip shouldnt be able to turbo just as high on a finit amount of cores as this chip. Honestly the only way this would make since would be if they were trying to charge substantially more for the ability to run multiple chips but even then windows begins to hit limits at 256 threads and on massively parallel tasks they can normally be distributed over multiple servers
Click to expand...

In regards to the single CPU with lots of threads vs multi CPU with combined equal threads, licensing for a single socket is cheaper. Licensing 64 cores in one CPU is about 25% cheaper than licensing 64 cores spread over 2.

And in regards to the TDP on servers yes jet engines.... but on a server it’s not uncommon to see 16+ phases per socket that isn’t something you are going to see in many desktop class boards and most of the server form factors are pre designed for multi concurrent PSU’s for consistent power delivery from a regulated source. You just have conditions in a server room that are not going to be delivered to 95% of the users. I am very tempted to liquid cool my next server rack. A couple of 1U 9x9 radiators should do nicely.
 
Nolan7689 said:
It may be your understanding, but aren’t basically all Zen 2 parts clocked same or higher than their comparative Zen+ counterparts? And isn’t part of this bump attributable to the die shrink to 7nm?
Click to expand...

I think it makes sense that we gain some back with optimizations but we were at near 5 ghz on water back in 2011. 4.5 ghz with the Pentium EE before that. All the engineering from the planet's semiconductor manufacturers and we're still in the same ballpark on clock speed.
 
Lakados said:
In regards to the single CPU with lots of threads vs multi CPU with combined equal threads, licensing for a single socket is cheaper. Licensing 64 cores in one CPU is about 25% cheaper than licensing 64 cores spread over 2.

And in regards to the TDP on servers yes jet engines.... but on a server it’s not uncommon to see 16+ phases per socket that isn’t something you are going to see in many desktop class boards and most of the server form factors are pre designed for multi concurrent PSU’s for consistent power delivery from a regulated source. You just have conditions in a server room that are not going to be delivered to 95% of the users. I am very tempted to liquid cool my next server rack. A couple of 1U 9x9 radiators should do nicely.
Click to expand...

In regards to Licensing it will depend on how they setup the chip and present it to the os. The Intel phi chips I have actually present each chip as 5 separate CPUs in windows and seeing how amds chips are pretty much just multiple well connected CPUs it may be presented similarly
 
cdabc123 said:
In regards to Licensing it will depend on how they setup the chip and present it to the os. The Intel phi chips I have actually present each chip as 5 separate CPUs in windows and seeing how amds chips are pretty much just multiple well connected CPUs it may be presented similarly
Click to expand...
Any and all licensing I have dealt with was simply Physical sockets and logical cores, I have never seen one that depended on how it was presented to the OS.
 
Lakados said:
Any and all licensing I have dealt with was simply Physical sockets and logical cores, I have never seen one that depended on how it was presented to the OS.
Click to expand...

it may have just een wierdness with the intel phi chips as windows was convinced i had 5 physical sockets
 
Is it just me or do the Hardware Tech News subforum make everything disorganized? Nobody puts CPU related stuff in the cpu sub thread and same goes for video cards...
 
Nightfire said:
Is it just me or do the Hardware Tech News subforum make everything disorganized? Nobody puts CPU related stuff in the cpu sub thread and same goes for video cards...
Click to expand...

I've seen stuff put into multiple forums- that would probably work best here, especially for news.

Granted the subforum doesn't really make any difference when you use 'New Posts'.
 
Nightfire said:
Is it just me or do the Hardware Tech News subforum make everything disorganized? Nobody puts CPU related stuff in the cpu sub thread and same goes for video cards...
Click to expand...
Nah, news should be in the news subforum. If there is something specific you want to discuss (about the architecture or a certain processor's features, for instance) that's a good candidate for the other subforum, as it may not relate directly to the news story.
 
So I have a question . Why isn't single core VPSs still not getting cheaper despite cheaper and larger core counts from AMD. Cheapest I found from a reputable company is 2.50$ but still tho it seems like it should be cheaper yes the prices are stuck for a while now
 
owned66 said:
So I have a question . Why isn't single core VPSs still not getting cheaper despite cheaper and larger core counts from AMD. Cheapest I found from a reputable company is 2.50$ but still tho it seems like it should be cheaper yes the prices are stuck for a while now
Click to expand...

Rent, salaries, licencing, electricity. Ultimately, hardware cost is a blip on the sheets compared to the ongoing costs.
 
PhaseNoise said:
640 really ought to be enough for anybody.
Click to expand...
You called? :cool:

m_img_56600.jpg
 
You must log in or register to reply here.
Back
Top