• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

*** AMD ThreadRipper ***

Wouldn't say sabotaged themselves - complacency and I think partly complacency that people would still buy Intel as a brand especially in established markets like servers, etc. even if AMD did come out with something competitive. AMD caught them on the wrong foot as well so they can't easily respond - despite some people having trouble understanding the concept - I suspect its more happy accident for AMD that they caught them at the point they did as you can't really strategise for something like that to that degree in the CPU market as there are limits how early you can bring a product to market and development time can be put back by things like having to respin, etc. which you can't really control and it makes no sense to delay a product in 99% of cases as it just gives the competition more time to develop their [next] product.
That's exactly what sabotaging themselves is...
 
That's exactly what sabotaging themselves is...

I was seeing sabotage as a more long term thing :s while I think Intel has only stumbled (partly doing it to themselves, partly AMD doing it to them) and I don't think will take as long to recover as people think or want to think.
 
Is Zen2 the 7nm one, which is different than Zen+ coming early next year yes?
Do we know, if DDR5 will be coming also with the Zen2? Because that would improve the performance even more.

I think DDR5 is set for 2020. Going by AMDs slide Zen 2 will arrive before DDR5. Yes Zen 2 is 7nm one. Not sure what zen+ is though.

Sonic. :p

As it is for Zen's Infinitiy fabric both are important, but the actual Speed, which affects the bandwidth is what's important.
Since the I.F runs at 50% RAM speed, which directly affects the bandwidth between the CCXs.

With high speed RAM Infinity Fabric runs from 22GB/s all the way to 50GB/s, so you can see why it's so important.
I.F can scale to 500GB/s according to AMD, which is why they're also going to use it for APUs, and Navi/GPUs.

Thanks to both of you for clearing that up.

Considering it was Kodak's employees who invented way modern digital camera sensors work it remind me more about this:

General Lefcourt: We took care of the Dilgar, we can take care of the Minbari.
Londo Mollari: Ah, arrogance and stupidity all in the same package. How efficient of you.

Where is that quote from?

Wouldn't say sabotaged themselves - complacency and I think partly complacency that people would still buy Intel as a brand especially in established markets like servers, etc. even if AMD did come out with something competitive. AMD caught them on the wrong foot as well so they can't easily respond - despite some people having trouble understanding the concept - I suspect its more happy accident for AMD that they caught them at the point they did as you can't really strategise for something like that to that degree in the CPU market as there are limits how early you can bring a product to market and development time can be put back by things like having to respin, etc. which you can't really control and it makes no sense to delay a product in 99% of cases as it just gives the competition more time to develop their [next] product.

You mentioned that previously and I was wondering how could they have caught Intel on the "right foot". Surely with out inside knowledge it would be impossible for Intel to have formulated a response in time?
 
You mentioned that previously and I was wondering how could they have caught Intel on the "right foot". Surely with out inside knowledge it would be impossible for Intel to have formulated a response in time?

Right now Intel are pretty much committed to going with what they've got (they are already in volume production, etc.) with the longest possible time to the next iteration or having to go to massive expense to bring something forward. A few months earlier or later would have given Intel more room for manoeuvring and/or competing with the next iteration. Trying to generalise a bit as there is more than one platform and product cycle involved on both sides.
 
Right now Intel are pretty much committed to going with what they've got (they are already in volume production, etc.) with the longest possible time to the next iteration or having to go to massive expense to bring something forward. A few months earlier or later would have given Intel more room for manoeuvring and/or competing with the next iteration. Trying to generalise a bit as there is more than one platform and product cycle involved on both sides.
I see the point your trying to make, but I don't think there is much Intel could have done with a few months.
For mainstream kabylake, the best they could have done is reduce the price. I don't think the 6 core from the HEDT side can be re-purposed for the mainstream boards.

For HEDT and servers its not just cores it other things like maximum RAM size and features like ram encryption (server only).
For HEDT apart from removing the artificial limitations its once again a price thing.
Server space it's the same thing, I highly doubt that even with more time Intel could have formulated a response to AMDs RAM encryption.

Considering these are minor iterations compared to the previous gen I'm guessing it takes about 2-3 years to design these CPUs. I think it would take a year at the least for Intel to properly react.
 
I see the point your trying to make, but I don't think there is much Intel could have done with a few months.
For mainstream kabylake, the best they could have done is reduce the price. I don't think the 6 core from the HEDT side can be re-purposed for the mainstream boards.

For HEDT and servers its not just cores it other things like maximum RAM size and features like ram encryption (server only).
For HEDT apart from removing the artificial limitations its once again a price thing.
Server space it's the same thing, I highly doubt that even with more time Intel could have formulated a response to AMDs RAM encryption.

Considering these are minor iterations compared to the previous gen I'm guessing it takes about 2-3 years to design these CPUs. I think it would take a year at the least for Intel to properly react.

And don't forget on the server market isn't encryption only.

Those 128pci lanes provide huge flexibility on everything. From storage to machine learning to expansion.
The best Intel can do on single socket server atm, is 40 lanes for more than double the price. And on dual socket, can only work 40+40 with limitations, for 3 times the price of the single socket AMD.
In addition all the legacy parts, USB3 etc, are inside the chip, and no dedicated chip on the board is required. Hence the boards can be more tightly packed for more stuff, and better performance of the overall chip. (there is a dedicated the server module on the EPYC CPUs.)
 
Last edited:
And don't forget on the server market isn't encryption only.

Those 128pci lanes provide huge flexibility on everything. From storage to machine learning to expansion.
The best Intel can do on single socket server atm, is 40 lanes for more than double the price. And on dual socket, can only work 40+40 with limitations, for 3 times the price of the single socket AMD.
In addition all the legacy parts, USB3 etc, are inside the chip, and no dedicated chip on the board is required. Hence the boards can be more tightly packed for more stuff, and better performance of the overall chip. (there is a dedicated the server module on the EPYC CPUs.)
I did forget about the PCIE lanes on the server even though I was watching the EPYC tech videos while writing it. Stuff like that requires a redesign of the CPUs I don't expect Intel having large PCIE capability till Jan 2019 in the earliest. Assuming they started redesigning them after the financial analyst day announcement.
 
Yes Zen 2 is 7nm one. Not sure what zen+ is though.

Where is that quote from?
Zen+ is both trying to get more clock speed out of current manufacturing process and no doubt tweaking those parts of architecture AMD didn't have time to do for Zen.
With all the downsizing and lack of high end products AMD hasn't had too much of resources or money available...
Without console APUs guaranteeing steady income doubt they could have kept developing Zen even this long.
So would definitely expect Zen+ being bigger improvement to Zen that Intel's annual update of model number.

Looks like you have hole in education...:p
http://tvtropes.org/pmwiki/pmwiki.php/Series/BabylonFive?from=Main.Babylon5
And approriate music for background while reading.
https://www.youtube.com/playlist?list=PLcVh-WWldt8n782rfHmR5VKyjdZom65yn




I think it would take a year at the least for Intel to properly react.
Even if architecture/design components are ready it might take closer to year to turn that into product in shop:
After finishing full logical level design there's no doubt lots of simulation of it for testing.
When being enough sure there likely aren't much of faults in it you have to start turning that logical design into physical chip design with transistors and other components, their positioning and all electrical wiring.
Then that tooling has to be made from so that you can start making engineering samples for further testing.
 
I see the point your trying to make, but I don't think there is much Intel could have done with a few months.
For mainstream kabylake, the best they could have done is reduce the price. I don't think the 6 core from the HEDT side can be re-purposed for the mainstream boards.

For HEDT and servers its not just cores it other things like maximum RAM size and features like ram encryption (server only).
For HEDT apart from removing the artificial limitations its once again a price thing.
Server space it's the same thing, I highly doubt that even with more time Intel could have formulated a response to AMDs RAM encryption.

Considering these are minor iterations compared to the previous gen I'm guessing it takes about 2-3 years to design these CPUs. I think it would take a year at the least for Intel to properly react.

That is the thing though - we all know Intel have been trickling out feature and performance updates so they can charge the most for them artifically delaying or restricting features so they can charge silly money for certain CPU models, etc. don't mistake it for Intel lacking advances in these areas they undoubtedly have many more advances working in the lab, etc.
 
Zen+ is both trying to get more clock speed out of current manufacturing process and no doubt tweaking those parts of architecture AMD didn't have time to do for Zen.
With all the downsizing and lack of high end products AMD hasn't had too much of resources or money available...
Without console APUs guaranteeing steady income doubt they could have kept developing Zen even this long.
So would definitely expect Zen+ being bigger improvement to Zen that Intel's annual update of model number.

Looks like you have hole in education...:p
http://tvtropes.org/pmwiki/pmwiki.php/Series/BabylonFive?from=Main.Babylon5
And approriate music for background while reading.
https://www.youtube.com/playlist?list=PLcVh-WWldt8n782rfHmR5VKyjdZom65yn




Even if architecture/design components are ready it might take closer to year to turn that into product in shop:
After finishing full logical level design there's no doubt lots of simulation of it for testing.
When being enough sure there likely aren't much of faults in it you have to start turning that logical design into physical chip design with transistors and other components, their positioning and all electrical wiring.
Then that tooling has to be made from so that you can start making engineering samples for further testing.

The information i found point to a Zen refresh being zen 2, coming out in early 2018, I have seen some stuff about 14nm+ but i think that could be for the APUs. Also wikipedia says that zen 2 use to be called zen+.

Babylon 5, can't have been any good if i've never heard of it :p*
*Disclaimer: I have actually heard of babylon 5, i've just never watched it.



That is the thing though - we all know Intel have been trickling out feature and performance updates so they can charge the most for them artifically delaying or restricting features so they can charge silly money for certain CPU models, etc. don't mistake it for Intel lacking advances in these areas they undoubtedly have many more advances working in the lab, etc.
True but there is still the issue of turning that research into an actual mass producible product. Which is still takes a decent amount of time. All the research is good but you need to make it work together and IMO i think that's where the hard part is.
 
True but there is still the issue of turning that research into an actual mass producible product. Which is still takes a decent amount of time. All the research is good but you need to make it work together and IMO i think that's where the hard part is.
Lots of money and resources sure help but don't remove that huge amount of work and various steps needed.

Even actual manufacturing of completed silicon die takes time with lots of stages in it.
While hoping there's no flaws in it requiring physical changes/redesign to even test it.

Heck, tape out for production of first engineering samples of 7nm Zen 2 and Navi are in this year and maybe in couple months but we won't see products in shop until some time next year.



B5 is only two time Hugo Award winner, which is Oscar equivalent for scifi.
Along with handfull of other awards and whole pile of nominations.
And had more plot and thought than in dozens of standard TV-series, hence going badly for average consumer with worser than goldfish memory.
Including hints to things whole season ahead... or two. (which you only understand then)
 
B5 is only two time Hugo Award winner, which is Oscar equivalent for scifi.
Along with handfull of other awards and whole pile of nominations.
And had more plot and thought than in dozens of standard TV-series, hence going badly for average consumer with worser than goldfish memory.
Including hints to things whole season ahead... or two. (which you only understand then)

B5 was epic. And also might point out, that series like Caprica and Continuum (2012) got canned or chopped, because they were "too technical and difficult to understand".

----------------------------------

On topic, for those worried.
I had to dig through about waterblocks for X399, and found a two week old post by the EK rep at the OCNET forum, and I quote.....

We have the SP3(r2) socket water blocks ready for some time now (since end of December). They will be released on time
wink.gif
 
Was wondering how much OEMs, etc. would bother with aftermarket cooling for some of these bigger, less enthusiast, etc. segment targetted sockets.
 
I think DDR5 is set for 2020. Going by AMDs slide Zen 2 will arrive before DDR5. Yes Zen 2 is 7nm one. Not sure what zen+ is though.
Going by AMD's past practises, it's possible that whichever Zen comes out in 2020 (Zen 3 or 4) will have both a DDR4 and DDR5 memory controller and be compatible with both AM4 and AM5 sockets. Not sure if it'll be physically possible but I bet they'll try.
 
Apparently Japanese retailers are saying Threadripper will come stock with an AIO CLC; and will launch August 10th there.

http://www.gdm.or.jp/voices/2017/0716/214535

"7月14日(金)に発表されたAMDの新ハイエンドCPU「Ryzen Threadripper」シリーズ。国内では、7月27日(木)より予約受け付けがスタート。発売日は8月10日(木)に設定されている。なお、現時点で予価は未定。製品には標準で水冷クーラーが付属するという。"

Translation - AMD announced on July 14 (Fri) new high-end CPU series "Ryzen Threadripper". In Japan, the reservation is accepted starting July 27 (Thursday). Release Date is set to August 10 (Thursday). It should be noted that, at present tentative is undecided. The water-cooled cooler is supplied as standard with the product.
 
Back
Top Bottom