Q: Availability and devices
A: Super happy about OEM engagement. 150+ platform designs across consumer and commercial. 50+ for consumer, starting october. 20 on Evo.
A: Super happy about OEM engagement. 150+ platform designs across consumer and commercial. 50+ for consumer, starting october. 20 on Evo.
Q: Prev gens had power consumption reduction. Explain efficiency?
A: Fundamental reduction in TGL. With that power efficiency, we decided to use it with headroom - added IP and capability in the same TDP range.
A: Fundamental reduction in TGL. With that power efficiency, we decided to use it with headroom - added IP and capability in the same TDP range.
Q: How does Intel deliver speed up in these AI situations
A: Comes through capabilities and datatypes, then software on top. New hardware in CPU and GPU with DL Boost is at the heart.
A: Comes through capabilities and datatypes, then software on top. New hardware in CPU and GPU with DL Boost is at the heart.
Q: Is the UHD non-iris graphics same as Gen 9?
A: No it's the Xe across the entire product line. At the higher end, to get the iris, you have to get the dual channel (memory?)
A: No it's the Xe across the entire product line. At the higher end, to get the iris, you have to get the dual channel (memory?)
Q: Why dual ring-bus? Was done previously with HCC CPUs.
A: Opportunity based on transistors and metal stack. Used the same circuit design techniques to double ring bandwidth - helps BW for GPU and CPU we wanted to make sure it was never a bottleneck. Had little impact on power
A: Opportunity based on transistors and metal stack. Used the same circuit design techniques to double ring bandwidth - helps BW for GPU and CPU we wanted to make sure it was never a bottleneck. Had little impact on power
A: Ring can scale in voltage/bw based on workload requirement
Q: You said driving ring frequencies ? From what to what?
A: Ice Lake at 3.8 GHz, ring freq was statically bound to cores or graphics. What we've done in TGL can sense what freq is needed and adjust as required. Ring can get close to 4.8 GHz
A: Ice Lake at 3.8 GHz, ring freq was statically bound to cores or graphics. What we've done in TGL can sense what freq is needed and adjust as required. Ring can get close to 4.8 GHz
Q: What is the benefit of OpenVINO?
A: Can train on OpenVINO or supply pre-train, and OpenVINO can optimize for Intel inference on each platform inc quantization or model size etc Supports AVX-512 and GNA
A: Can train on OpenVINO or supply pre-train, and OpenVINO can optimize for Intel inference on each platform inc quantization or model size etc Supports AVX-512 and GNA
Q: How soon will 11th Gen Desktops catch up?
A: Stay tuned for those announcements. We'll be scaling out 11th Gen for mobile into Celeron/Pentium/vPro. Market is demanding for mobile right now, so that's our focus.
A: Stay tuned for those announcements. We'll be scaling out 11th Gen for mobile into Celeron/Pentium/vPro. Market is demanding for mobile right now, so that's our focus.
Q: Differences between Intel Iris Xe and Iris Plus?
A: Doubled performance, and that's the 2nd time in a row. A lot about using Superfin. Arch work and hardcore design work has pushed the core of what's new. Also more in media, AV1 etc
A: Doubled performance, and that's the 2nd time in a row. A lot about using Superfin. Arch work and hardcore design work has pushed the core of what's new. Also more in media, AV1 etc
Q: If customers want Gen4 dGPU and Gen4 SSD, what lanes?
A: TGL today has 4 PCIe 4.0 lanes, so direct SSD you would use those. It's either or, or put the SSD on the PCH. Though on these platforms we don't think you need a dGPU :)
A: TGL today has 4 PCIe 4.0 lanes, so direct SSD you would use those. It's either or, or put the SSD on the PCH. Though on these platforms we don't think you need a dGPU :)
Q: Is UP3/UP4 related to Y and U?
A: (Intel said something different about package choices and platform level, but UP3 is U, UP4 is Y)
A: (Intel said something different about package choices and platform level, but UP3 is U, UP4 is Y)
Q: How can the consumer know when a laptop CPU will operate at a given power when OEMs don't state?
A: TGL is all about the dynamic range. We don't force branding about thermal indication so it will come down to the performance reviews.
A: TGL is all about the dynamic range. We don't force branding about thermal indication so it will come down to the performance reviews.
Q: What is the die size and transistor count?
A: We're keeping it to ourselves, it's different. We'll leave it to others to find out.
A: We're keeping it to ourselves, it's different. We'll leave it to others to find out.
Q: PCH in 14nm or 10nm?
A: 14nm
A: 14nm
Q: Will all mobile processors in future be 10nm?
A: For mobile with 11th Gen we've transitioned to Superfin and going forward we'll use Superfin and beyond for mobile.
A: For mobile with 11th Gen we've transitioned to Superfin and going forward we'll use Superfin and beyond for mobile.
Q: Will Intel enable its RUGs for press?
A: Yes. We believe strongly in this. We want the press to validate our testing.
A: Yes. We believe strongly in this. We want the press to validate our testing.
Q: When can we expect to see vPro?
A: Early next year, 11th Gen Core with vPro and 11th Gen Evo vPro
A: Early next year, 11th Gen Core with vPro and 11th Gen Evo vPro
Q: Is the best gaming performance in the Evo spec systems?
A: What we saw today, the Evo validation has to use Xe Iris and our top performance SKUs.
A: What we saw today, the Evo validation has to use Xe Iris and our top performance SKUs.
Q: Gen 4 PCIe SSDs launched until are now bulky. What's the point of Gen 4 on TGL
A: They're not all bulky! There are thinner form factors that give the performance you desire.
A: They're not all bulky! There are thinner form factors that give the performance you desire.
Q: Integrated the WiFi, what about cellular? Why is Cellular not part of Athena?
A: We provide 4G and partner with 5G - when we look at the attach rate, having a discrete solution enables flexibility for partners and improve time to market.
A: We provide 4G and partner with 5G - when we look at the attach rate, having a discrete solution enables flexibility for partners and improve time to market.
Q: Do you expect to see TGL support 28W? Very few ICL.
A: As you see the larger chassis, you'll see higher operating ranges. It tracks with form factor.
A: As you see the larger chassis, you'll see higher operating ranges. It tracks with form factor.
Q: Will the transistor changes and stack changes spread over Intel, or just mobile?
A: They'll be available for all on Superfin and above. We've seen a large benefit of mobile. Scalability helps enable this.
A: They'll be available for all on Superfin and above. We've seen a large benefit of mobile. Scalability helps enable this.
Q: Is there a gearing ratio between CPU/GPU/ring to sync or async?
A: Separate voltage domains and clocks. There are clock crossings. They can operate autonomously. Used to be a ratio to agent, but that's gone now. Completely independent
A: Separate voltage domains and clocks. There are clock crossings. They can operate autonomously. Used to be a ratio to agent, but that's gone now. Completely independent
Q: PCIe 4 on UP3 and UP4?
A: Yes!
A: Yes!
Q: Evo sub-brand enables reliability and performance removing bottlenecks?
A: A big part of that promise is responsiveness and responsiveness on all power. We have 25 KEIs to pass to be EVO branded in multiple scenarios.
A: A big part of that promise is responsiveness and responsiveness on all power. We have 25 KEIs to pass to be EVO branded in multiple scenarios.
Q: Developer interest on AI use cases on client?
A: Lot of 3rd party research; All developers are integrating some form of AI in knowledge or development. AI is not standalone, it has unique integration. Seeing a lot of breakthroughs now to help users
A: Lot of 3rd party research; All developers are integrating some form of AI in knowledge or development. AI is not standalone, it has unique integration. Seeing a lot of breakthroughs now to help users
A: We're at the beginning of this era. Lots of new use cases coming through that leverage AI. Infusing in everything across the board to make things better and easier. Not just in apps, but also at the OS level and SoC management
Q: Architecture can support LPD5 - when available?
A: Stay Tuned on future products.
A: Stay Tuned on future products.
Q: PCH is PCIe 4 or PCIe 3?
A: PCIe 3
A: PCIe 3
Q: dGPU on TGL?
A: You will see this, but as people move to thin and light with battery life and efficiency, we expect to see a lot more Xe only
A: You will see this, but as people move to thin and light with battery life and efficiency, we expect to see a lot more Xe only
20 mins of Q&A left
So ask questions here
Q: Benefit of AV1?
A: All about power efficiency and perf. In software, it pegs the CPU and very inefficient. In hardware, can get smooth AV1 playback.
A: All about power efficiency and perf. In software, it pegs the CPU and very inefficient. In hardware, can get smooth AV1 playback.
Q: When are the ML-Perf results going to be officially submitted?
A: In a short time frame, during September
A: In a short time frame, during September
Q: How should users think about 11th Gen and Evo vs laptops today?
A: Evo approaches a lot of discrete solutions in market today. Will always be a market for the super high end, but Xe enables AAA titles playable in thin and light designs.
A: Evo approaches a lot of discrete solutions in market today. Will always be a market for the super high end, but Xe enables AAA titles playable in thin and light designs.
A: THis isn't just on a few select titles - it's across a wide range of games at 1920x1080
Q: Can you share more about additional security mitigations in 11th gen
A: We Enabled CET along with TME.
A: We Enabled CET along with TME.
Q: MTK branded 5G on TGL?
A: Making progress to deliver a 5G solution. Sampling those modules to partners very soon. Early next year we expect products with that solution
A: Making progress to deliver a 5G solution. Sampling those modules to partners very soon. Early next year we expect products with that solution
Q: Now L2 and L3 are larger, latency?
A: Yes, but our caches are much faster, so latency is actually less. Most people associate larger cache with larger latency.
A: Yes, but our caches are much faster, so latency is actually less. Most people associate larger cache with larger latency.
Q: What AMD systems in the comparisons?
A: We went above and beyond to get the best AMD systems for our comparison. We believe the comparison is the best possible given time constraints.
A: We went above and beyond to get the best AMD systems for our comparison. We believe the comparison is the best possible given time constraints.
Q: 1.25MB L2 does what?
A: Enables larger workloads and sets straight out of the core.
A: Enables larger workloads and sets straight out of the core.
Q: Which AI quantization levels does Intel feel be relevant to this form factor?
A: Depends on workload. Some high precision defaults to FP32, e.g. medical imaging and visual effects, where speed isn't important. But INT8 will be a sweet spot.
A: Depends on workload. Some high precision defaults to FP32, e.g. medical imaging and visual effects, where speed isn't important. But INT8 will be a sweet spot.
Q: What should influencers look for with TGL?
A: (Chris Walker) This is our answer to what influences, our customers, and what our users have asked for. It is the best across everything you do, in fantastic form factors.
A: (Chris Walker) This is our answer to what influences, our customers, and what our users have asked for. It is the best across everything you do, in fantastic form factors.
That's a wrap on the TGL Q&A.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
