Drop in will be up to the OEMs. But yes this platform should be good for 3-5 years before any major updates.So Rome is Zen2 Epyc and should plug-in to current Epyc servers if I'm reading correctly. As well as currently being worked on and being debugged.
Seeing that GloFlo is leading with a high power 7nm and TSMS with a low power one, what do see as the time difference for TSMC to have the HPC version?Here we have multiple sources confirming that GF 7LP is launching with a process for high performance .
http://www.anandtech.com/show/11558...nm-plans-three-generations-700-mm-hvm-in-2018
The 7 nm platform of GlobalFoundries is called 7LP for a reason — the company is targeting primarily high-performance applications, not just SoCs for smartphones, which contrasts to TSMC’s approach to 7 nm. GlobalFoundries intends to produce a variety of chips using the tech, including CPUs for high-performance computing, GPUs, mobile SoCs, chips for aerospace and defense, as well as automotive applications. That said, in addition to improved transistor density (up to 17 million gates per mm2 for mainstream designs) and frequency potential, GlobalFoundries also expects to increase the maximum die size of 7LP chips to approximately 700 mm², up from the roughly 650 mm² limit for ICs the company is producing today. In fact, when it comes to the maximum die sizes of chips, there are certain tools-related limitations.
For their newest node, the company is focusing on two ways to reduce power consumption of the chips: implementing superior gate control, and reducing voltages. To that end, chips made using GlobalFoundries' 7LP technology will support 0.65 – 1 V, which is lower than ICs produced using the company’s 14LPP fabrication process today. In addition, 7LP semiconductors will feature numerous work-functions for gate control.
https://www.semiwiki.com/forum/content/6837-globalfoundries-7nm-euv-update.html
GF however is leading with a high performance (LP equals Lead Performance in IBM speak) version of 7nm for AMD while TSMC is first with a low power version of 7nm for Apple, Qualcomm, MediaTek, and the other SoC vendors.
https://www.globalfoundries.com/sites/default/files/product-briefs/7lp-product-brief.pdf
http://semimd.com/chipworks/2017/01/18/iedm-2016-setting-the-stage-for-75-nm/
Four Vt options are available in the TSMC 7-nm technology [1] . There are four device Vt options with a range of ~200 mV.
GF 7LP provides choice of 5 core device Vt compared to TSMC which provides 4. Here again the requirement to support very high performance 5 Ghz and low power mobile is driving the need to provide a broader range of Vt and higher Vt.
TSMC is launching first with a low power mobile version of N7 for its primary customer Apple and GF is launching with a high performance version of 7LP for AMD. Its going to be interesting to see the comparison of products manufactured at TSMC N7 HPC for Nvidia vs GF 7LP for AMD in 2019.
I'm not sure they would really. If your fabs are full and making money, then why would you change horses so to speak? Comes down to money at the end of the day. If your fabs are filled, why spend money trying to fill your fabs more...Seeing that GloFlo is leading with a high power 7nm and TSMS with a low power one, what do see as the time difference for TSMC to have the HPC version?
Will be interesting how GloFo's and TSMC's different approaches pan out.TSMC is launching first with a low power mobile version of N7 for its primary customer Apple and GF is launching with a high performance version of 7LP for AMD. Its going to be interesting to see the comparison of products manufactured at TSMC N7 HPC for Nvidia vs GF 7LP for AMD in 2019.
Or AMD could be using TSMC for future console SoCs or Navi.
Later edit: I don't think AMD has the resources to build Zen2 on 2 silicon architectures. I also don't think GF has the capacity to sustain IBM and AMD's Ryzen and Radeon ramp up on 7nm.
As we already have mid-gen refreshes for both consoles and Sony has stated they'll still do fully new gens a future console SoC may well be Zen based. And AMD did well in the past having development in the semi custom business feed back into their main business.Or AMD could be using TSMC for future console SoCs or Navi.
Here we have multiple sources confirming that GF 7LP is launching with a process for high performance .
http://www.anandtech.com/show/11558...nm-plans-three-generations-700-mm-hvm-in-2018
The 7 nm platform of GlobalFoundries is called 7LP for a reason — the company is targeting primarily high-performance applications, not just SoCs for smartphones, which contrasts to TSMC’s approach to 7 nm. GlobalFoundries intends to produce a variety of chips using the tech, including CPUs for high-performance computing, GPUs, mobile SoCs, chips for aerospace and defense, as well as automotive applications. That said, in addition to improved transistor density (up to 17 million gates per mm2 for mainstream designs) and frequency potential, GlobalFoundries also expects to increase the maximum die size of 7LP chips to approximately 700 mm², up from the roughly 650 mm² limit for ICs the company is producing today. In fact, when it comes to the maximum die sizes of chips, there are certain tools-related limitations.
For their newest node, the company is focusing on two ways to reduce power consumption of the chips: implementing superior gate control, and reducing voltages. To that end, chips made using GlobalFoundries' 7LP technology will support 0.65 – 1 V, which is lower than ICs produced using the company’s 14LPP fabrication process today. In addition, 7LP semiconductors will feature numerous work-functions for gate control.
https://www.semiwiki.com/forum/content/6837-globalfoundries-7nm-euv-update.html
GF however is leading with a high performance (LP equals Lead Performance in IBM speak) version of 7nm for AMD while TSMC is first with a low power version of 7nm for Apple, Qualcomm, MediaTek, and the other SoC vendors.
https://www.globalfoundries.com/sites/default/files/product-briefs/7lp-product-brief.pdf
http://semimd.com/chipworks/2017/01/18/iedm-2016-setting-the-stage-for-75-nm/
Four Vt options are available in the TSMC 7-nm technology [1] . There are four device Vt options with a range of ~200 mV.
GF 7LP provides choice of 5 core device Vt compared to TSMC which provides 4. Here again the requirement to support very high performance 5 Ghz and low power mobile is driving the need to provide a broader range of Vt and higher Vt.
TSMC is launching first with a low power mobile version of N7 for its primary customer Apple and GF is launching with a high performance version of 7LP for AMD. Its going to be interesting to see the comparison of products manufactured at TSMC N7 HPC for Nvidia vs GF 7LP for AMD in 2019.
Yeah, if 7FF LP is GloFo only, as it appears, (no Samsung) then Zen2 won't be going anywhere - especially not to TMSC. On the plus side, Fab 28 is undergoing a 20% expansion. Even though some of us expect Zen2 to use 6 core CCXs, that would still mean ~33% more dice per wafer.
Not sure that a 6 core CCX makes that much sense- you're boosting the intra-CCX latency. My personal guess would be that they go to 3x4 per die, rather than 2x6.
But how do you make a die of 3 ccx? would have to be a very "long" die but nor sure that is a problem but maybe could be due to package size.
But how do you make a die of 3 ccx? would have to be a very "long" die but nor sure that is a problem but maybe could be due to package size.
I think the new AMD is going to make as few fundamental changes as possible for the needed performance. We are seeing their philosophy as getting the most out of as little as possible starting to payback in a big way. Don't expect that to change for the next few years at least.Hmm, good point. Maybe they could change the internal layout of a CCX, so that instead of a 2x2 tile it's a 1x4? *not a processor engineer*
Not sure that a 6 core CCX makes that much sense- you're boosting the intra-CCX latency. My personal guess would be that they go to 3x4 per die, rather than 2x6.
Not sure that a 6 core CCX makes that much sense- you're boosting the intra-CCX latency. My personal guess would be that they go to 3x4 per die, rather than 2x6.
Edit2: Dang slide is from a clickbait site
They have. Already announced and up on their webpage.If they build supermicro 1s boards I might consider buying EPYC. Seems Intel is leaving out 1s support with the Purley platform.... Not really happy about that...
You will find them here. Supermicro Epyc Servers and MotherboardsIf they build supermicro 1s boards I might consider buying EPYC. Seems Intel is leaving out 1s support with the Purley platform.... Not really happy about that...
Wow. I just figured this out.... The most I can find for DDR4 ECC registered is 64 gig modules. That motherboard I saw (H11DSi) has 16 slots and supports up to 2 TB memory 2666. So for 1 TB (64*16) DDR4 2133 (fastest I could find) and 2 32 core, 64 thread CPU's, and just guessing on $1000 motherboard with HSF's that 25k for 128 threads, and 1 TB memory, just WOW, not bad (for enterprise class hardware)
The memory was $1849 for 2 64 gig memory sticks = ~15k + mobo+ 8k for 2 processors.