The Ryzen "ThreadRipper"... 16 cores of awesome

[Edrick]

I am not much of an AMD fan, and generally prefer Intel due to platform maturity, but AMD is kicking the snot out of Intel with Threadripper.

$999 for a 16 core CPU with a 4Ghz boost OC is insane.

Say goodbye to the stagnation CPU world of the past 6 years.

 

[estarkey7]

Because you can buy a better CPU for that MT workload with lower price (EPYC, Xeon) than their HEDT counterparts.
And I consider EPYC and Xeon as workstation/ server CPU, not server CPU.
So if that workload is THE one, HEDT 16C makes no sense to use price/performance and pure performance wise.

That is they way I am looking at it. No offence and hard feelings OFC.

That CPU might be lower price on the 24/48 Epyc , but that motherboard ain't! Remember, Epyc and Therefore have the same pin count, but Epyc has octal channel ram. That means those pins are either floating or (more likely) grounded on Threadripper. Unless AMD has a way in the BIOS to swap between quad channel memory and octal channel, an Epyc CPU in a Threadripper motherboard may burn up!

Sent from my rk3188 using Tapatalk

 

[maddie]

This. I still cant comprehend why would they downgrade from solder to TIM the exact moment they finally have some competition. It goes against all logic.

Hubris.

A lot of how Intel have approached markets with their pricing and segmentation strategies smacks of it. We can do what we want as no one can touch us.

Remember Andrew Grove "only the paranoid survive". He must be turning in his grave.

Granted, paranoia is a terrible state of existence.

 

[maddie]

It makes sense if you consider their enterprise customers won't care and it probably saves them a good amount of money. They think they can "weather the PR storm" and come out on the other side with bags of extra cash, no worse for wear.

I wouldn't be surprised if they are right.

Then again arent enterprise customers buying Xeons instead of HEDT CPUs? But i agree, they may be right. I am still considering it myself. But i would be past that and already bought, if not for these thermal issues - they saved few bucks on solder, but lost (at least for now) entire grand on the missed sale cause of it. Past few years, i would have no other choice, and if i wanted/needed it, my money would end in their pocket anyway. This year they may not.

Timmah! is right and it might be possible to ballpark estimate this.

If we assume that it costs $2 to apply solder instead of Silicone TIM and we take an average $800 for a HEDT.

Then for every AMD Threadripper sale instead of an Intel Skylake-X due to the TIM fiasco, it will take Intel to make 400 sales to recoup that lost tiny extra profit.

I think my general logic is sound even if the assumed numbers are incorrect, which leads us to believe that Intel loses overall.

 

[krumme]

Timmah! is right and it might be possible to ballpark estimate this.

If we assume that it costs $2 to apply solder instead of Silicone TIM and we take an average $800 for a HEDT.

Then for every AMD Threadripper sale instead of an Intel Skylake-X due to the TIM fiasco, it will take Intel to make 400 sales to recoup that lost tiny extra profit.

I think my general logic is sound even if the assumed numbers are incorrect, which leads us to believe that Intel loses overall.

Agree its a huge loss given the market size. But its always easy to make the right decision in hindsight.
When TIM was decided and set in stone there was no competition in sight especially for those huge cores; it was the right decision. Where it goes haywire is the decision to overclock those processors out the gate when it was known threadripper was comming. Was vrm overload, bad efficiency, trotling and high temp worth the perf benefit? Imo no. Its here it went wrong. Better to take the hit on shortterm margins than burn your brand imo.

The decisions and eg those childish recent ppt make the impression its made in haste. Stupid and ugly. Far from the Intel with confidense and control.
You also sell what you are. And that - Be it factory oc cpu or childish ppt is certainly not optimal b2b branding.
Add its not the picture you want to communicate internally in an organization like Intel that is loaded with highly competent senior engineers and technicians. With all due respect but its not a burger chain and the imployees demand more.

 

[lobz]

and this is why I said I am missing the high clock 12C TR. 10C at 4,6 GHz is most of the time more than 16C at 3.5GHz

Don't forget that the 7900X consumes a RIDICULOUSLY high amount of power at 4.6 GHz, not to mention you have to purchase a very beefy cooling, so you don't burn down the house trying to achieve that frequency.

 

[lobz]

It makes sense if you consider their enterprise customers won't care and it probably saves them a good amount of money. They think they can "weather the PR storm" and come out on the other side with bags of extra cash, no worse for wear.

I wouldn't be surprised if they are right.

using TIM instead of solder saves them a good amount of money? you're just kidding, right?

 

[dnavas]

Adobe Premiere Pro to video rendering! After doing color grading, Neat Video noise reduction and a few special effects, I once did a render out of a 4K H.264 project down scaled to 1080p for an hour and a half video output that took >47 hours!

Exactly this -- video editing. 4k60p footage or 4:2:2 1080p footage are both a pain just to decode (the latter because the bozos spec'ing video card support for accelerated decode haven't woken up to the fact that 4:2:2 is a thing). My last springboard diving video was done in 4k with a slo-mo inset, and I could easily use ten *times* the number of cores that my 1800X has during an encode, if my video package could use it (which it can't, afaik). My last concert was a 45 minute video, and after you deal with grading and re-centering, and all the rest of it, it did take something like 10 hours to render (during which my 1800X black-screened twice :yay: so really more like two days). And that was with only one cam. Throw a few more cams in the mix, and it just balloons.

I'm convinced that if you took one of those two empty spots on TR and stuffed a dedicated encoder/decoder chip that could handle a wider range of formats than the standard video card enc/dec can and (say) 10 8k streams, at higher bandwidths, it would see a market. I don't expect it's a particularly *large* market, alas :]

In 4k60p video, there's half a billion pixels per second, and for non-GPU effects (which a lot of them are) that's a crap-ton of data to deal with if you want to keep your timeline real-time. So even with dedicated enc/dec hardware, there's still a real need for many cores and fast processors (sadly, not all of these effects scale with cores either). 8k is going to be four times that bandwidth. CPU makers can't possibly keep up at this rate.

 

[maddie]

Agree its a huge loss given the market size. But its always easy to make the right decision in hindsight.
When TIM was decided and set in stone there was no competition in sight especially for those huge cores; it was the right decision. Where it goes haywire is the decision to overclock those processors out the gate when it was known threadripper was comming. Was vrm overload, bad efficiency, trotling and high temp worth the perf benefit? Imo no. Its here it went wrong. Better to take the hit on shortterm margins than burn your brand imo.

The decisions and eg those childish recent ppt make the impression its made in haste. Stupid and ugly. Far from the Intel with confidense and control.
You also sell what you are. And that - Be it factory oc cpu or childish ppt is certainly not optimal b2b branding.
Add its not the picture you want to communicate internally in an organization like Intel that is loaded with highly competent senior engineers and technicians. With all due respect but its not a burger chain and the imployees demand more.

How long to correct this?

2-3 quarters for share price to be affected and a further 2-3 quarters for reorganization to begin. I agree that they have excellent engineers, but it seems they're not in charge.

AMD has some room to grow and consolidate their position. At least Lisa is not a Hector.

 

[IndyColtsFan]

This. I still cant comprehend why would they downgrade from solder to TIM the exact moment they finally have some competition. It goes against all logic.

Remember, x299 and SKL-X (but only 6, 8, and 10 core models) had been planned well before Ryzen was released and Intel didn't think Ryzen could touch them. Contrary to what some say, the 12-18 cores models were a direct reaction to Threadripper. I have no doubt that Intel had the 12-18 core models in mind as a contingency plan but I have little doubt that they thought they'd never have to use them. They're in serious jeopardy of losing MT performance to the 16 core TR IMO because I think the 16 core i9 won't have much, if any, clockspeed advantage. TR may end up with enough clock speed advantage to mostly overcome the IPC advantage and be within 5% of the 16 core i9 but for hundreds less.

I think we're are seeing 12-18 delayed precisely because Intel didn't think they would need to exercise that contingency and it wouldn't surprise me if we did see soldered i9s in the next wave. It might be too late for a soldered 12c as I assume it is already in production, but they may have no choice on the 14-18 models.

 

[IndyColtsFan]

I saw on Slickdeals that the 6850 is now $300 at Microcenter. Could this be the start of a price war between AMD and Intel?

 

[krumme]

How long to correct this?

2-3 quarters for share price to be affected and a further 2-3 quarters for reorganization to begin. I agree that they have excellent engineers, but it seems they're not in charge.

AMD has some room to grow and consolidate their position. At least Lisa is not a Hector.

Ppt and factory overclock is certainly not gona cut it when 7nm zen2 is here in 1.5 year. Therefore its shortsighted.
They need to run it like it was family owned and not give a shit about shareprice for the next 3 quarters.
Keep cool and focus.
Ajust margins expectations. Take the hit.
Cut cost.
Cut portfolio.
Cut capex.
Invest in core.
Invest in tech where they are unique. Eg integrate new mem tech into cpu tech.

 

[scannall]

Remember, x299 and SKL-X (but only 6, 8, and 10 core models) had been planned well before Ryzen was released and Intel didn't think Ryzen could touch them. Contrary to what some say, the 12-18 cores models were a direct reaction to Threadripper. I have no doubt that Intel had the 12-18 core models in mind as a contingency plan but I have little doubt that they thought they'd never have to use them. They're in serious jeopardy of losing MT performance to the 16 core TR IMO because I think the 16 core i9 won't have much, if any, clockspeed advantage. TR may end up with enough clock speed advantage to mostly overcome the IPC advantage and be within 5% of the 16 core i9 but for hundreds less.

I think we're are seeing 12-18 delayed precisely because Intel didn't think they would need to exercise that contingency and it wouldn't surprise me if we did see soldered i9s in the next wave. It might be too late for a soldered 12c as I assume it is already in production, but they may have no choice on the 14-18 models.

I think it's unlikely that they will solder anything. They will just move the base clock and boost clock at stock down a bit, and then give the disclaimer that "Overclocking isn't advised and will void your warranty."

 

[Timmah!]

Remember, x299 and SKL-X (but only 6, 8, and 10 core models) had been planned well before Ryzen was released and Intel didn't think Ryzen could touch them. Contrary to what some say, the 12-18 cores models were a direct reaction to Threadripper. I have no doubt that Intel had the 12-18 core models in mind as a contingency plan but I have little doubt that they thought they'd never have to use them. They're in serious jeopardy of losing MT performance to the 16 core TR IMO because I think the 16 core i9 won't have much, if any, clockspeed advantage. TR may end up with enough clock speed advantage to mostly overcome the IPC advantage and be within 5% of the 16 core i9 but for hundreds less.

I think we're are seeing 12-18 delayed precisely because Intel didn't think they would need to exercise that contingency and it wouldn't surprise me if we did see soldered i9s in the next wave. It might be too late for a soldered 12c as I assume it is already in production, but they may have no choice on the 14-18 models.

I wonder, if we could see soldered 7900x in the next wave too be it simply introduced as a new stepping or renamed to say 7910x. Thats another thing that keeps me from buying 7900x.

EDIT: Why would be too late for soldered 12core? I know its competely unlikely and unrealistic, but technically not impossible - they could withdraw all the 7900Xs from the market, delid them and have soldered, right? And 7920x is not even out, so they could push it to later date, if they required more time to solder it...

 

[moinmoin]

I saw on Slickdeals that the 6850 is now $300 at Microcenter. Could this be the start of a price war between AMD and Intel?

Microcenter is not Intel (nor AMD), their deals are one of a kind.

 

[IndyColtsFan]

Microcenter is not Intel (nor AMD), their deals are one of a kind.

Yeah, I know, but that seems really low. I am also wrong - it was the 6850 for $300, not the 6800.

 

[IndyColtsFan]

I think it's unlikely that they will solder anything. They will just move the base clock and boost clock at stock down a bit, and then give the disclaimer that "Overclocking isn't advised and will void your warranty."

I think you're right and that they'll rely on the strength of marketing to weather the Threadripper storm.

 

[krumme]

I wonder, if we could see soldered 7900x in the next wave too be it simply introduced as a new stepping or renamed to say 7910x. Thats another thing that keeps me from buying 7900x.

Dont know. But zen was needed to be stopped now on the low freq 14llp proces. Not a highly optimized 7nm high freq ibm process. Its hopeless in 2019. tim what not. They will be run over on the core.
How ironically it might seen its actually the new cache and mesh that is securing them scalability that is their strongest feature to defend servermarket.
Now they are forced to really innovate. TR and epyc simply exposed them.

 

[maddie]

Ppt and factory overclock is certainly not gona cut it when 7nm zen2 is here in 1.5 year. Therefore its shortsighted.
They need to run it like it was family owned and not give a shit about shareprice for the next 3 quarters.
Keep cool and focus.
Ajust margins expectations. Take the hit.
Cut cost.
Cut portfolio.
Cut capex.
Invest in core.
Invest in tech where they are unique. Eg integrate new mem tech into cpu tech.

Agree with everything you posted but realistically, not going to happen.

Top executives have too much remuneration tied up in stock options. The board will have a lot of work to do. Hell, big shareholders might even substantially shake-up the board itself.

The recent amateurish Xeon marketing effort drives home the uphill task needed to fix this and this time, unlike the P4 days, there is no process advantage coming. I actually think that Intel might be in a more dangerous position now than then, especially when comparing technologies and top level management.

 

[krumme]

Agree with everything you posted but realistically, not going to happen.

Top executives have too much remuneration tied up in stock options. The board will have a lot of work to do. Hell, big shareholders might even substantially shake-up the board itself.

The recent amateurish Xeon marketing effort drives home the uphill task needed to fix this and this time, unlike the P4 days, there is no process advantage coming. I actually think that Intel might be in a more dangerous position now than then, especially when comparing technologies and top level management.

Agree its a far far more dangerous situation and due to amd unlimited production capability.
Zen core and IF just makes it possible to use this new production ecosystem.
But perhaps this bizarre reaction from Intel is just a result of them realizing that?
There will be turmoil. And lot of the historic obsession with margins will have to go. Certainly it looks like its a mess now and as the ceo is without control. This drifting sensation we all sense. Hopefully things will return to normal after the shakeup and a leaner, cooler and more creative Intel will emerge. More tech less segmenting. Frankly. Less greed. They will get bigger and live longer beeing less greedy.
But those top management stock options...how ironic. Shareholder selfinflicted pain.

 

[wildhorse2k]

There is also a major difference between situation now and many years back when AMD introduced Athlons and gave Intel bad beating. There is little difference between manufacturing process now and it is expected to even out in near future. Intel can no longer keep increasing frequency. So AMD can manufacture their cheap 8 cores and build bigger CPUs out of them. Due to doubling/quadrupling of surface area it doesn't get thermal issues either.

 

[IndyColtsFan]

Agree its a far far more dangerous situation and due to amd unlimited production capability.
Zen core and IF just makes it possible to use this new production ecosystem.
But perhaps this bizarre reaction from Intel is just a result of them realizing that?
There will be turmoil. And lot of the historic obsession with margins will have to go. Certainly it looks like its a mess now and as the ceo is without control. This drifting sensation we all sense. Hopefully things will return to normal after the shakeup and a leaner, cooler and more creative Intel will emerge. More tech less segmenting. Frankly. Less greed. They will get bigger and live longer beeing less greedy.
But those top management stock options...how ironic. Shareholder selfinflicted pain.

Yes, but I think everyone (me included) forgets one thing - how large is the market for HEDT anyway? Intel may be perfectly content with their status quo and let their marketing spin the situation for sales because it may not matter to them. IMO, HEDT is big for AMD - not because of volume but because it can help restore their reputation and consumer confidence.

 

[.vodka]

Yes, but I think everyone (me included) forgets one thing - how large is the market for HEDT anyway? Intel may be perfectly content with their status quo and let their marketing spin the situation for sales because it may not matter to them. IMO, HEDT is big for AMD - not because of volume because it can help restore their reputation and consumer confidence.

It's not about HEDT, it's about the server market. Epyc goes the same approach with four dies under that huge heatspreader. This is where Intel's panic comes from.

TR is just AMD applying the server solution (packaging and socket) to Intel's HEDT niche and being able to get a nice product out of the door with as little extra effort as possible.



AM4's credit is also there, Zeppelin as a building block provides a great mainstream platform with lots of power and connectivity for cheap.

TR is out of my price range, but I don't know if I can scratch the upgrade itch off much longer and my OC'd 2500k is too long in the tooth now, motherboards are now stable, AGESA 1.0.0.6a is out with even more fixes, a long upgrade path ahead in the future... What puts me off is the DDR4 prices, damn it.

 

[maddie]

At its most fundamental level, we have to realize that AMD is only fabbing one product.

We say Desktop. HEDT and Server, because we've been conditioned that way, but from AMD's point of view, they most probably see, 1 die product, 2 die product and 4 die product. The packaging differentiates them, not forgetting the appropriate glue of course.

No juggling of how many wafers to assign to a certain die, extremely simple inventory management. Do many realize how big an advantage this alone is?

 

[Tup3x]

IMO, Intel's segmenting is just awful overall. Very anti-consumer. Yet some are content with it apparently. Intel even segmented their AVX speed performance and PCI-Express lanes based on the CPU you buy even when its the same core overall.

It's starting to hurt them. Next year should be really interesting...