- Status
As one of the core guys who created Threadripper, it's totally my favorite. 
And thanks for a great design, I now have 4 TR rigs, and now a 2700X rig. Will be one of the first to get TR2
1) Are there any news in the open source Firmware front?
I'm not of the "open source everything" crowd as I am aware that there are things that involves the Intellectual Property of several companies, so any such effort would fall into a lawyer's maze. However, AMD could still provide the require binary blobs and documentation to initialize the Hardware. If my memory serves me correctly, last time I checked, Coreboot seems to not support Zen because AMD didn't provided the bare minimum to do so, and reverse engineering would take ages.
2) Will there be SoC-style Zen based products for entry level consumer?
As much as I love Zen itself, a thing that annoys me to no end is that there are no Motherboards that I know of that uses Zen as a pure SoC. AMD announced the A300 and X300 Chipsets back when Ryzen was launched, intended for mITX Motherboards, but I think that no one did products with them.
The thing is, Ryzens using Socket AM4 have 24 PCIe Lanes available (4 are used by the Chipset), while the Zen die itself has 32 and I think that the recently announced EPYC Embedded products can actually use all of them. Zen also has embedded USB3, SATA Controller and even a hidden 10 GBit NIC. A mATX sized embedded Ryzen or Raven Ridge would need no Chipset at all, it should be rather cheap and great for some minimalist builds (Even including Internet Cafes), and still pack a punch.
A full Chipset adds Motherboard complexity, power consumption, latency and cost, and all of them are stupid considering that in entry level builds Zen itself as a SoC could happily carry the entire build. It really gets me on my nerves that Zen is such a great piece of silicon and most Motherboards manufacturers doesn't realize its potential.
And that leads me to...
3) Will AMD try at some point vertical integration of the product stack, like designing reference Motherboards but for production instead of internal testing?
The best way to solve points 1 and 2, would be that AMD tried designing a few reference Motherboards and make them available for the typical Motherboards partners to manufacture and support, as to not compete with them. This would be similar to what happens with Video Cards, so I don't see that it is an impossible idea.
If AMD designed a Motherboard from the ground up, it could pick the other components of the platform based on the availability of end user documentation. This way, instead of providing just some code to initialize the Hardware then leaving the Coreboot community the task of having to reverse engineer one by one a few Motherboards to implement whatever support chips they use (And most never get to the fully feature complete stage), you could have a Motherboard that from the ground up is intended to cater to that audience, so everything about it is functional and documented.
I think that an embedded Raven Ridge in mATX Form Factor with open source Firmware would be a one-of-a-kind platform. Simple, cheap, auditable, somewhat expandable...
I'm not of the "open source everything" crowd as I am aware that there are things that involves the Intellectual Property of several companies, so any such effort would fall into a lawyer's maze. However, AMD could still provide the require binary blobs and documentation to initialize the Hardware. If my memory serves me correctly, last time I checked, Coreboot seems to not support Zen because AMD didn't provided the bare minimum to do so, and reverse engineering would take ages.
2) Will there be SoC-style Zen based products for entry level consumer?
As much as I love Zen itself, a thing that annoys me to no end is that there are no Motherboards that I know of that uses Zen as a pure SoC. AMD announced the A300 and X300 Chipsets back when Ryzen was launched, intended for mITX Motherboards, but I think that no one did products with them.
The thing is, Ryzens using Socket AM4 have 24 PCIe Lanes available (4 are used by the Chipset), while the Zen die itself has 32 and I think that the recently announced EPYC Embedded products can actually use all of them. Zen also has embedded USB3, SATA Controller and even a hidden 10 GBit NIC. A mATX sized embedded Ryzen or Raven Ridge would need no Chipset at all, it should be rather cheap and great for some minimalist builds (Even including Internet Cafes), and still pack a punch.
A full Chipset adds Motherboard complexity, power consumption, latency and cost, and all of them are stupid considering that in entry level builds Zen itself as a SoC could happily carry the entire build. It really gets me on my nerves that Zen is such a great piece of silicon and most Motherboards manufacturers doesn't realize its potential.
And that leads me to...
3) Will AMD try at some point vertical integration of the product stack, like designing reference Motherboards but for production instead of internal testing?
The best way to solve points 1 and 2, would be that AMD tried designing a few reference Motherboards and make them available for the typical Motherboards partners to manufacture and support, as to not compete with them. This would be similar to what happens with Video Cards, so I don't see that it is an impossible idea.
If AMD designed a Motherboard from the ground up, it could pick the other components of the platform based on the availability of end user documentation. This way, instead of providing just some code to initialize the Hardware then leaving the Coreboot community the task of having to reverse engineer one by one a few Motherboards to implement whatever support chips they use (And most never get to the fully feature complete stage), you could have a Motherboard that from the ground up is intended to cater to that audience, so everything about it is functional and documented.
I think that an embedded Raven Ridge in mATX Form Factor with open source Firmware would be a one-of-a-kind platform. Simple, cheap, auditable, somewhat expandable...
Raven Ridge has been a popular offering for both desktop and—to a lesser extent—for laptops markets in its first generation iteration as a budget gaming offering. I have some questions regarding its future.
1. Will Raven Ridge’s driver releases (speaking of Windows drivers) ever be integrated or consolidated into the same releases as the main discrete graphics driver branch?
2. Some creative users have proposed using ThreadRipper’s multi-die, Infinity Fabric-based package design for a super Raven Ridge of sorts. They propose placing a GPU die and CPU die—and perhaps dedicated memory as well—in a single ThreadRipper-like package. Has AMD considered this for future Raven Ridge releases—or if you are not at liberty to speak freely, minimally, has this idea at least been acknowledged by AMD?
3. In what ways has Raven Ridge been a success for AMD and what does it mean for this product line?
4. In what ways do you see Raven Ridge pushing the boundaries even further in the future?
5. Do you see a Surface or a Surface-like 2-in-1 device using Raven Ridge in the near future?
1. Will Raven Ridge’s driver releases (speaking of Windows drivers) ever be integrated or consolidated into the same releases as the main discrete graphics driver branch?
2. Some creative users have proposed using ThreadRipper’s multi-die, Infinity Fabric-based package design for a super Raven Ridge of sorts. They propose placing a GPU die and CPU die—and perhaps dedicated memory as well—in a single ThreadRipper-like package. Has AMD considered this for future Raven Ridge releases—or if you are not at liberty to speak freely, minimally, has this idea at least been acknowledged by AMD?
3. In what ways has Raven Ridge been a success for AMD and what does it mean for this product line?
4. In what ways do you see Raven Ridge pushing the boundaries even further in the future?
5. Do you see a Surface or a Surface-like 2-in-1 device using Raven Ridge in the near future?
Last edited:
Are there any plans to focus more on single threaded performance vs overall multithreaded capacity?
It looks like AMD chose to make a multithreading beast where as Intel made a very fast single threaded CPU then made it multithreaded with unused execution pipelines. AMD's integer pipeline seems to be far more symmetrical where Intel's is not.
It looks like AMD chose to make a multithreading beast where as Intel made a very fast single threaded CPU then made it multithreaded with unused execution pipelines. AMD's integer pipeline seems to be far more symmetrical where Intel's is not.
1 - With the 2nd gen line up taking advantage of 12nm, Precision Boost 2, and Zen+ architecture, we were able to make a single product that outperforms both 1700X and 1800X. We also had market feedback that three versions of the eight core, were a little confusing. So we opted to simplify the stack a little bit, and have two models; same core and thread and cache amounts, but different TDP and different stock performance. At the same time we went to a product line up that had all processors with included cooler, which meant it was simpler again to decide which model you want - now it's about looking for a specific TDP, or most performance out of a core count.
1 - Yes, and you should see this happen this month.
2 - It's an interesting idea, for sure.
3 - "Raven Ridge" has been a success in that it has helped AMD get new product lines with all the major global OEMs, plus be able to more successfully compete in the commercial and consumer desktop markets. The product line is still continuing to ramp and has lots of room to grow past its initial success. This will help to fuel further development of similar style products for the future.
4 - The biggest opportunities are the obvious ones, getting more of a presence in retail and mainstream buying marketplaces with good notebooks and desktops that familys and businesses want to use. It's a given, but that's the boundary that needs to be moved. Then we think about space probes or deep sea drones or lava proof computers (I don't know, I'm just making this part up).
2 - It's an interesting idea, for sure.
3 - "Raven Ridge" has been a success in that it has helped AMD get new product lines with all the major global OEMs, plus be able to more successfully compete in the commercial and consumer desktop markets. The product line is still continuing to ramp and has lots of room to grow past its initial success. This will help to fuel further development of similar style products for the future.
4 - The biggest opportunities are the obvious ones, getting more of a presence in retail and mainstream buying marketplaces with good notebooks and desktops that familys and businesses want to use. It's a given, but that's the boundary that needs to be moved. Then we think about space probes or deep sea drones or lava proof computers (I don't know, I'm just making this part up).
1 - No new news, sorry.
2 - Can't speculate on future products, sorry.
3 - The interest levels on AMD reference platforms being available in the marketplace has been very low, plus our motherboard partners are not sure they want us to compete - and we currently have not only very good relationships with them, but also a pretty good set of products in the market today. There's been a few references to few SFF boards being available, but that will change as Ryzen becomes more popular and the market drives the need for the less popular (but still very necessary) form factors. There's also the resouce angle to consider - do you want AMD engineers building products that are already in market, or do you want them focused on the next generation of products? Or, if you imagined hiring new staff, do you want those new staff members reinventing products already on sale, instead of being people that deliver the new IP the drives the computing market forward? I think most people pick innovation and new products. But there's certainly still requests AMD can make to the existing partners to see if there's something else they can do.
2 - Can't speculate on future products, sorry.
3 - The interest levels on AMD reference platforms being available in the marketplace has been very low, plus our motherboard partners are not sure they want us to compete - and we currently have not only very good relationships with them, but also a pretty good set of products in the market today. There's been a few references to few SFF boards being available, but that will change as Ryzen becomes more popular and the market drives the need for the less popular (but still very necessary) form factors. There's also the resouce angle to consider - do you want AMD engineers building products that are already in market, or do you want them focused on the next generation of products? Or, if you imagined hiring new staff, do you want those new staff members reinventing products already on sale, instead of being people that deliver the new IP the drives the computing market forward? I think most people pick innovation and new products. But there's certainly still requests AMD can make to the existing partners to see if there's something else they can do.
When the GPU driver thread is on a separate CCX from the thread it's communicating with, Ryzen suffers a significant drop in draw call performance. Even when paired with blazing fast RAM, it only manages to be as fast as Core 2 at processing draw calls.
We did a benchmark to see the draw call performance of specific CPU architectures, which gave us the above findings.
And even when the driver thread is communicating to another thread on the same CCX, with blazing fast RAM used, draw call performance is around Ivybridge's. Considering that Ryzen has single thread performance on par with Broadwell, why is it that Ryzen is not as good at processing draw calls?
xLake Intel CPUs are very good at draw calls, with them being 23% faster than Ryzen at processing draw calls despite only being ~5% faster in single thread performance. Will draw call performance be targeted for significant improvement in future architectures?
We did a benchmark to see the draw call performance of specific CPU architectures, which gave us the above findings.
And even when the driver thread is communicating to another thread on the same CCX, with blazing fast RAM used, draw call performance is around Ivybridge's. Considering that Ryzen has single thread performance on par with Broadwell, why is it that Ryzen is not as good at processing draw calls?
xLake Intel CPUs are very good at draw calls, with them being 23% faster than Ryzen at processing draw calls despite only being ~5% faster in single thread performance. Will draw call performance be targeted for significant improvement in future architectures?
If you look at the extreme overclocker results, you can submissions on HWBOT of all cores/threads running at 5.5GHZ+. You just have to be sub-zero, get the heat out of the chip. So the clock frequencies are really limited by thermal dissipation, and the voltage required to drive that clock.
Architecture does make a difference to clocks, you can see that on the HWBOT fastest processors - 18 of the top 20 are "Bulldozer" family architecture chips. So likely there are some things we can look into for future cpu core architectures alongside working with our foundry manufacturing partners.
Architecture does make a difference to clocks, you can see that on the HWBOT fastest processors - 18 of the top 20 are "Bulldozer" family architecture chips. So likely there are some things we can look into for future cpu core architectures alongside working with our foundry manufacturing partners.
Hi James, i noticed mentioning about the boot kit.
It is inevitable that some stores all over the world because of stocked motherboards, may still sell boards with an older bios that does not support ryzen 2 series. And might not update the boards.
Is AMD going to extend the boot kit solution to ryzen 2 cpu as well ?
It is inevitable that some stores all over the world because of stocked motherboards, may still sell boards with an older bios that does not support ryzen 2 series. And might not update the boards.
Is AMD going to extend the boot kit solution to ryzen 2 cpu as well ?
- Status
Similar threads
-
Question FX9590 intermittent no post -- but runs fine once started. Ruled everything but CPU. ?????
- Started by DONINAUSTIN
- Replies: 35
-
[EETimes] TSMC starting 5 nm EUV Risk Production in April?
- Started by jpiniero
- Replies: 81
-
AMD 2990WX 32 core 64 thread Builders thread Pics starting build !- Started by Markfw
- Replies: 141
TRENDING THREADS
-
[Rumor, Tweaktown] AMD to launch next-gen Navi graphics cards at E3
- Started by Hitman928
- Replies: 502
-
-
-
-
