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

[John Carmack]

Could be. But that's the way I received it. I'm proceeding under the assumption that TR isn't even in good enough shape to share clock speeds, that "summer" means "October", and that this is all hat no cattle. You're right, I also don't see any reason you couldn't slap two 1700s together, but they didn't even show a number after Blender. Down right weird, man. One would figure, with Vega too late to compete with Pascal, that you'd want to highlight the stuff that is shipping, but even the server specs were really thin.

That said, as I said, I don't know if this is AMD's MO. Maybe this kind of dysfunction is normal?

There were no cache counts, core speeds, or Blender scores shown for the other brand either. Must be non competitive.

 

[mattiasnyc]

I would think that hype and demos and teasers would work well with the gaming crowd, but demoing 4k gaming on an hedt? Why? You want workstation software companies up on stage with you telling your potential customers (their existing customer base) why you're a force to be reckoned with: CAD, 3D modeling/animation, media editing, data visualization (bonus points for running queries against your servers), maybe machine learning? If you want the professional market, you'll have to start acting that way.

It might be an issue with Computex itself though. Do the types of people who would invest in an AMD HEDT really attend Computex and get swayed by a presentation on stage?

I would think that for many bigger customers, i.e. businesses, they'll either wait for people to sort out the issues or buy an assembled system from a vendor. So the question is if vendors would have been impressed by a presentation here, or if they'd be more interested in getting review samples and a great line of communication with AMD's development / support teams. I think it's the latter, so I'm not entirely surprised they're playing this angle.

I do agree though that there are people like me who have been holding out on an upgrade and who now want to see how things perform, people who are professionals using their workstations to earn money, even if it isn't full time. On the other hand, many of us are pretty pragmatic about it and are "ok" with waiting anyway.

So perhaps AMD has simply found that this is a good thing to do for marketing, and they're also pursuing other avenues at the same time.

But yes, absolutely AMD shook Intel loose. There would never have been more than 12 cores, and there might not have even been a 12 core from Intel given their previous path. It's what I've said to anyone who asks me why I bought a Ryzen system -- if my purchase helps bring better hardware sooner, I'd do it again in a heartbeat.

Commendable attitude. You have my vote.

 

[piesquared]

There were no cache counts, core speeds, or Blender scores shown for the other brand either. Must be non competitive.

I don't recall seeing any release time line for intel's line up either. For sure not the 12+ core chips that they added last minute (TBA?). I'm guessing sometime towards Christmas?

 

[dark zero]

Threadripper's success will come down to price, of course. I think that 16 core Ryzen at $1,000 would put Intel in a tight spot. If clocks are only in the 2ghz range or something like that, then threadripper will anemic. Vega is coming way too late though. 1080ti already cleaned house and Volta is right around the corner. I think VEGA is basically DOA.

That is unless Vega is 25% more than 1080ti...

 

[dnavas]

There were no cache counts, core speeds, or Blender scores shown for the other brand either. Must be non competitive.

No, there were no core speeds shown for 12-core and up. Would you like me to take Intel to task as well? Ok, fine, I am assuming, as with AMD that these skus are not ready, and that they are likely 1) a response to AMD's TR and vaporware OR 2) not yet optimized enough for prime time. Specifically, I assume the 12-core not being ready is why Intel would have waited until later in the year to release Skylake-X, and the 14-core and up are responses to AMD and may well be some jury-rigged piece of junk that I wouldn't touch with a ten foot pole. TBD.

What we did get from Intel, however, was a statement of what they intend to bring core-count-wise to market. With pricing. We didn't get that from AMD. Intel had something to demonstrate regarding their pricing. They did that. They don't need to demonstrate performance, because they already have the per-core, per-clock performance advantage as well as high-clocking and OC headroom. To demonstrate competitiveness, they needed to lower pricing.

What we got from AMD was a promise not to release a brain-damaged product like that 6-core piece of sadness that Intel wants to sell. That's actually really good. But they had a more complex task wrt competitiveness, and they didn't even ante-up.

It might be an issue with Computex itself though. Do the types of people who would invest in an AMD HEDT really attend Computex and get swayed by a presentation on stage?

Good question. Upon reviewing the entire presentation, it seems like the main thrust of the call was related to vendor support for Ryzen chips. You're right, they may be angling for the support of vendors and/or system builders. I didn't think much of that part of the presentation either, mind you, but the presentation is less useful than the post-meetups I guess?

 

[wildhorse2k]

So Computex has been a big dissapointment as AMD didn't release much information. I expected Threadripper frequencies and prices, looked forward to CPU battles and instead we get nothing.

I speculate that AMD was ready to do that, but when Intel announced its 14-18 core Skylake-X unexpectedly along with prices, AMD in the last minute decided to re-think it and delay. Perhaps there are still some problems that need solving. Intel managed it very well:

• when it was clear Threadripper was real, it rushed Skylake-X to be released sooner. Probably because lower core CPUs were basically ready, just not the 12 core and it was originally in no rush to release. It was nice to make it until Computex.
  
• customers will be able to buy up to 10 core Skylake-X in a month, while Threadripper will be released probably slightly later so AMD could miss some profit. Some customers may be willing to wait 1-2 months for reviews (just in case Intel screwed it up), but not willing to wait 3-4 months (to be sure Threadripper doesn't have Ryzen issues)
  
• Intel announced 14-18 core CPUs along with prices which are much lower than expected ($2000 for Intel 18 core is not a lot) to keep loyal customers
• Intel complicated AMDs position with historically low pricing (although still not enough for many here). AMD could have been hoping to sell Threadrippers for more to get good margins. Now they can't sell the 16C for more than $1200, probably less since Intels pricing above 10C is quite agressive (only $200 for each additional 2 cores).
  
• Intel will release 14-18 core CPUs slightly later than AMD will release Threadripper, giving it time to choose competitive frequencies
• Intel can pride themselves in having more cores

 

[IEC]

It's probably more likely AMD is just focusing on delivering with the server market first with Epyc, and the work on that will trickle down to the Threadripper platform. I don't think AMD is specifically responding to Intel's announcement.

 

[Glo.]

B&C have put a stick into ants nest with this tweet:

https://twitter.com/BitsAndChipsEng/status/870373386391891968

According to them entry level 16C/32T Threadripper CPU will cost 850$.

That would be actually mind blowing.

 

[IEC]

B&C have put a stick into ants nest with this tweet:

https://twitter.com/BitsAndChipsEng/status/870373386391891968

According to them entry level 16C/32T Threadripper CPU will cost 850$.

That would be actually mind blowing.

Considering two 1700s = $319 x 2 = $638, that's not much of a premium... depending on clocks. Obviously that discounts the increased cost of putting two dies together but the pins being on the motherboard rather than the processor should at least partially offset that.

 

[Puffnstuff]

I'm in for that price as it seriously undercuts Intel's leaked prices by a huge margin.

 

[Topweasel]

B&C have put a stick into ants nest with this tweet:

https://twitter.com/BitsAndChipsEng/status/870373386391891968

According to them entry level 16C/32T Threadripper CPU will cost 850$.

That would be actually mind blowing.

Mind blowing indeed. I didn't think it was possible for me to be disappointed in my Ryzen setup. And disappointed is probably the wrong word. But given the choice I would have held out for this if that price is even close to accurate. Luckily I have an ITX system I was thinking of doing a Ryzen build in early next year. Not having a use for the board would be the only major loss. Might wait till late next year for a TR 2.0.

Then again I went with an R7 because it was that sweet spot I was looking for in a 8c16t chip. I think I am just hooked on the idea of even more cost effective cores, to an almost junkie like level.

 

[Topweasel]

I'm in for that price as it seriously undercuts Intel's leaked prices by a huge margin.

Exactly let them have the extra 2c at 220% the cost.

 

[dnavas]

B&C have put a stick into ants nest with this tweet:

https://twitter.com/BitsAndChipsEng/status/870373386391891968

According to them entry level 16C/32T Threadripper CPU will cost 850$.

And .. just as I was figuring that one of the reasons they didn't talk price was because they wanted to sell R7 high until TR comes out, they go and drop R7 prices and then TR prices leak out. Motherboards are not going to be cheap, though. Probably half again as much for the CPU + motherboard combo?

 

[Topweasel]

And .. just as I was figuring that one of the reasons they didn't talk price was because they wanted to sell R7 high until TR comes out, they go and drop R7 prices and then TR prices leak out. Motherboards are not going to be cheap, though. Probably half again as much for the CPU + motherboard combo?

But that is with the board makers to decide. The top end will hit $500+ but they can probably get away with $200-$300 offerings.

 

[ddogg]

Oh man, at that price I am definitely picking one up. I was expecting $1000+ but at $850 it's a sure buy for me.

 

[VirtualLarry]

B&C have put a stick into ants nest with this tweet:

https://twitter.com/BitsAndChipsEng/status/870373386391891968

According to them entry level 16C/32T Threadripper CPU will cost 850$.

That would be actually mind blowing.

I was darn close. I was hoping for $800, for the entry-level 16C/32T model. Still not too bad.

Edit: Still contemplating TR for the future. I do a lot of DC, or used to, so "MOAR CORES" is useful for what I do. DC will eat up as many cores as you can feed it, pretty-much.

Maybe after I save some money for my vehicle repairs.

(I just put together my Ryzen 5 1600 rig this morning, and it is SWEET! Don't have the blinged-out case with tempered glass, or 500GB 960 EVO PCI-E M.2 SSD yet though. More to come, I guess.)

 

[dnavas]

The top end will hit $500+ but they can probably get away with $200-$300 offerings.

Could be. I was figuring a board+CPU at $1200-$1500. It's the memory that will empty the wallet....

 

[T1beriu]

According to them entry level 16C/32T Threadripper CPU will cost 850$.

I just scrolled through B&C twitter feed. It's full of "hints" like this one that eventually proved to be complete BS. Have a look yourselves.

Again, as I said a few days ago, the guy behind the keyboard it's expressing his opinions as "leaks" or "facts", he's doesn't have an insider that feeds him leaks. You're reading too much into it. He uses the same tactics as w-junk-tech.

In a world where so called publications have forgot the basic rules of journalism are thriving by creating and spreading more and more fake news, we need to be more vigilant and critical of what we read and spread.

LE: I got to October 2016. Yeah, complete pile of BS leaks on that account. I can't stop laughing.

 

[krumme]

Can we have an entry level 32c epyc for 1799 usd then?
Lol

 

[IEC]

Can we have an entry level 32c epyc for 1799 usd then?
Lol

If anyone makes a Epyc-compatible E-ATX motherboard, I would totally buy one at that price. It's probably not that cheap though!

 

[Shamrock]

B&C have put a stick into ants nest with this tweet:

https://twitter.com/BitsAndChipsEng/status/870373386391891968

According to them entry level 16C/32T Threadripper CPU will cost 850$.

That would be actually mind blowing.

Has anyone read some of the comments on this link?

B&C says their source quoted 16c/32t cost AMD roughly $110-120 to make. So they are looking at $729-739 profit!

https://twitter.com/BitsAndChipsEng/status/870388109812355072

 

[wildhorse2k]

B&C says their source quoted 16c/32t cost AMD roughly $110-120 to make. So they are looking at $729-739 profit!

Do you think AMD employees work for free, no R&D cost, it doesn't have any debt and no bills to pay? Why is it that AMD never makes any profit per year?

 

[itsmydamnation]

Do you think AMD employees work for free, no R&D cost, it doesn't have any debt and no bills to pay? Why is it that AMD never makes any profit per year?

Because they had a crap CPU for the last 5 years.

But this is the whole point of zepplin, they made one SOC that is comparable in performance to intel which intel is currently/about to be using 4 SOC's for.

 

[Glo.]

Another portion of information from Fottemberg from BitsAndChips.
http://semiaccurate.com/forums/showpost.php?p=290705&postcount=594
If he is correct and the 12 core entry level Threadripper CPU will cost around 500(let it be even 599$) it is a bargain of the century.

P.S. Fott's information are sometimes hit and miss. But when he is correct, he is 100% correct .

 

[Glo.]

If anyone makes a Epyc-compatible E-ATX motherboard, I would totally buy one at that price. It's probably not that cheap though!

Epyc CPUs will fit in Threadripper socket. They have the same dimensions, and sizes, and are pin compatible.