Discussion Intel current and future Lakes & Rapids thread

[Hitman928]

AMD have a vPro alternative as a part of Ryzen Pro. Did Intel even mention it or was it totally glossed over?

Both AMD and Intel have built in malware protection in their pro line of processors. What Intel was demonstrating was a new feature only in Tigerlake pro processors they call CET which prevents control flow hijacking. It still needs to be supported at least at the OS level though and Microsoft has support for it in the build pipeline, not sure when it will hit the mainline release though. The funny thing is that because 10 nm is such a mess, this appears to be a mobile only feature for now as the Xeon line won't have this feature in the upcoming Icelake release and will probably not be seen in the Xeon line until Sapphire Rapids.

Here's a paper covering what it does and does not protect against: https://i.blackhat.com/asia-19/Thu-...sisted-Control-Flow-Integrity-Enforcement.pdf

I'm 100% sure AMD will have an equivalent soon, very likely in Zen4 and it will be ubiquitous throughout their pro/enterprise tier processors.

Intel also glossed over the fact that their AMT is absolutely riddled with security holes and bugs.

This is probably a much bigger deal than CET.

 

[Dayman1225]

AMD have a vPro alternative as a part of Ryzen Pro. Did Intel even mention it or was it totally glossed over?

I’m pretty sure the laptop they compared to was using a 4750U PRO.

 

[Hitman928]

Also, these guys did another test on RKL-S now that they have PCIe 4.0 working:

(88) 11代酷睿11900K游戏性能提前测试 网游被5600X吊打 单机不如十代i9 只能价格战 - YouTube

Looks like even with pcie4 working their Rocketlake sample is performing less than their 10900K. Looks like their sample is running lower frequencies at stock than the 10900K as well. Disappointing results but it's still an early sample so I'll wait to see how the final product with the release BIOS ends up.

 

[Abwx]

Here's a paper covering what it does and does not protect against: https://i.blackhat.com/asia-19/Thu-...sisted-Control-Flow-Integrity-Enforcement.pdf



This is probably a much bigger deal than CET.

In other words than Intel s Mc Afee s paper :

Intel has announced that its 10nm Tiger Lake CPUs will be boast a new hardware-based security feature to protect against Spectre-like malware attacks.

The laptop processors will be the first to come with would be the new Intel Control-Flow Enforcement Technology (CET), which Intel claims offers protection against the misuse of legitimate code through control-flow hijacking attacks, a type of malware that has been notoriously difficult to mitigate through software.
Jointly developed by Intel and Microsoft, the technology provides two new key capabilities to help guard against control-flow hijacking malware: Shadow Stack (SS) and Indirect Branch Tracking (IBT).

Intel Tiger Lake processors will thwart future Spectre and Meltdown attacks

Intel Control-Flow Enforcement Technology (CET) will protect against control-flow hijacking attacks

www.techradar.com

Viral marketing, litteraly, i told you...

 

[coercitiv]

What Intel was demonstrating was a new feature only in Tigerlake pro processors they call CET which prevents control flow hijacking.

Are you sure they made an actual demonstration though? Did they take over that system remotely based solely on a browser connection?

As of right now, I am not aware of a hardware based vulnerability on current AMD and/or Intel processors that enables taking over the system. I know of browser vulnerabilities which would allow that, and they get patched every month.

 

[mikk]

Press Center - TSMC to Kick off Mass Production of Intel CPUs in 2H21 as Intel Shifts its CPU Manufacturing Strategies, Says TrendForce | TrendForce - Market research, price trend of DRAM, NAND Flash, LEDs, TFT-LCD and green energy, PV

Intel has outsourced the production of about 15-20% of its non-CPU chips, with most of the wafer starts for these products assigned to TSMC and UMC, according to TrendForce’s latest investigations. While the company is planning to kick off mass production of Core i3 CPUs at TSMC’s 5nm node in...

www.trendforce.com

This is huge if true. I'm happy Intel is doing it.

Does it mean we will see CPUs from TSMC before Intel 7nm?

This news looks dubious. Realistically and logically, what Core i3 could be TSMC 5nm based and coming this year in mass production? As we know Alder Lake will be based on Intels own enhanced SuperFin and there is no sign that Intel is switching a mainstream Core i3 to TSMC this year. Furthermore the news claims Intel is switching to TSMC 3nm in 2021 for the mid-range and high-end which I don't see either, the timeframe is completely unrealistic. I can see TSMC for GPUs (DG2+) and partially server, however no chance for TSMC 3nm in 2022. I think Intel is going with enhanced SuperFin 10nm on consumer CPUs for a while. If 7nm really fails they may have to consider TSMC for consumer CPUs but I doubt it will happen before 2023.

 

[dmens]

This news looks dubious. Realistically and logically, what Core i3 could be TSMC 5nm based and coming this year in mass production? As we know Alder Lake will be based on Intels own enhanced SuperFin and there is no sign that Intel is switching a mainstream Core i3 to TSMC this year. Furthermore the news claims Intel is switching to TSMC 3nm in 2021 for the mid-range and high-end which I don't see either, the timeframe is completely unrealistic. I can see TSMC for GPUs (DG2+) and partially server, however no chance for TSMC 3nm in 2022. I think Intel is going with enhanced SuperFin 10nm on consumer CPUs for a while. If 7nm really fails they may have to consider TSMC for consumer CPUs but I doubt it will happen before 2023.

What if Intel has been doing double duty and designing everything to both processes in anticipation of their own process failing miserably?

 

[shady28]

This news looks dubious. Realistically and logically, what Core i3 could be TSMC 5nm based and coming this year in mass production? As we know Alder Lake will be based on Intels own enhanced SuperFin and there is no sign that Intel is switching a mainstream Core i3 to TSMC this year. Furthermore the news claims Intel is switching to TSMC 3nm in 2021 for the mid-range and high-end which I don't see either, the timeframe is completely unrealistic. I can see TSMC for GPUs (DG2+) and partially server, however no chance for TSMC 3nm in 2022. I think Intel is going with enhanced SuperFin 10nm on consumer CPUs for a while. If 7nm really fails they may have to consider TSMC for consumer CPUs but I doubt it will happen before 2023.

Looks like clickbait to me. Intel outsourcing non-CPU, frankly this would not be the first time. They have a long history of using foundries at different points in time to augment their own capacity, but not for CPUs.

In the current envionment with jacked up demand, I would look at that as normal.

The CPU part would be news, but I don't believe that. I think a lot of folks who don't know Intel's history of using foundries here and there during transition or spike demand times took some talk about Intel getting ready to use a foundry and extrapolated into something that wasn't there.

It would be perfectly normal for Intel to, for example, use a foundry to produce its chipsets while it is upgrading a few of its own fabs. Since they are planning to upgrade their one remaining 22nm node (used for low end desktop chipsets) to 14nm, it may simply be to keep supply up while they do that.

 

[Hulk]

I get the impression Intel did 11900K vs 5900X comparison because that is the price point they want for 11900K, $550. I would be more interested in the 11700K vs 5800X @$450.

This is an interesting observation. I think Intel is fooling itself if it believe that people are going to be cross shopping the 8 core 11900k against the 12 core 5900x. It's wishful thinking and an attempt to protect their pricing structure for RL. Problem is the real competitor for 11900k is the 5800X. They are going to have to realize that the top price of the 11900k is going to have to be $450 and the rest of the stack priced down from there.

And they better hope Zen 3 isn't widely available and AMD decides to really hurt them by pricing the 5800X at $350. Not likely I know but they could get mean if they wanted to because they have high end parts above and Intel does not.

 

[shady28]

This is an interesting observation. I think Intel is fooling itself if it believe that people are going to be cross shopping the 8 core 11900k against the 12 core 5900x. It's wishful thinking and an attempt to protect their pricing structure for RL. Problem is the real competitor for 11900k is the 5800X. They are going to have to realize that the top price of the 11900k is going to have to be $450 and the rest of the stack priced down from there.

And they better hope Zen 3 isn't widely available and AMD decides to really hurt them by pricing the 5800X at $350. Not likely I know but they could get mean if they wanted to because they have high end parts above and Intel does not.

Agree for the most part. 11900K might be ok at a price point between 5800X and 5900X, but not at the same price as the 5900X. Even if it eeks out small gains on single thread. it'll get smashed in multi-thread vs a 5900X. I'd rather have 8 faster cores than 12 slower ones, and I think that is far more useful for the vast majority of users. However, if it's the same price and performance difference is 3%? That is not enough to matter.

The 2nd point I disagree with. AMD is not in a position to do anything to affect pricing vs Intel until it can get more allocation from TSMC, or somehow extract themselves from the PS5/Xbox vortex that is sucking down their allocations. That's unlikely to happen in the near future. The only thing they could do is make the price go up, by charging more.

But in 3 months, if RKL is in (relatively) plentiful supply as I suspect it will be, how many will keep waiting on AMD?

 

[Hulk]

I'd rather have 8 faster cores than 12 slower ones,

Do you think the per core speed difference between RL and Zen 3 will be significant, more than 4 or 5%? I was thinking overall the two cores were going to more or less comparable.

 

[Exist50]

I'd rather have 8 faster cores than 12 slower ones

Because of power limits, Ryzen will be significantly faster in anything using 8 cores.

 

[Exist50]

This news looks dubious. Realistically and logically, what Core i3 could be TSMC 5nm based and coming this year in mass production? As we know Alder Lake will be based on Intels own enhanced SuperFin and there is no sign that Intel is switching a mainstream Core i3 to TSMC this year. Furthermore the news claims Intel is switching to TSMC 3nm in 2021 for the mid-range and high-end which I don't see either, the timeframe is completely unrealistic. I can see TSMC for GPUs (DG2+) and partially server, however no chance for TSMC 3nm in 2022. I think Intel is going with enhanced SuperFin 10nm on consumer CPUs for a while. If 7nm really fails they may have to consider TSMC for consumer CPUs but I doubt it will happen before 2023.

Dubious is putting it generously. The article is wrong, plain and simple.

 

[shady28]

Do you think the per core speed difference between RL and Zen 3 will be significant, more than 4 or 5%? I was thinking overall the two cores were going to more or less comparable.

No, I think it will be single digit % faster in single core, but in that range it doesn't matter much. Hence the post about other components, like choice of motherboard, can make 5% difference on the same CPU.

I do think it will have a lot more potential for an enthusiast, through power unlocking and overclocking or all-core max turbo type of tweaks. Same as with all Intel K chips.

So it comes down to price and availability. Hypothetically, if I had to choose between a 5900X and a 11900K at the same price point, I'd get a 5900X. The problem is, finding the 5900X, and bigger problem - finding it at MSRP.

I still think Intel will wind up pulverizing AMD simply by having a product that they can deliver in volume.

 

[stebler]

This website lists the RKL and CML-R CPUs, including cache config and base clocks.

Luxembourg

 

[coercitiv]

Are you sure they made an actual demonstration though? Did they take over that system remotely based solely on a browser connection?

I did a little more digging on this subject, and findings are a bit dubious: as required to disclose by Intel, their claims are based on a study commissioned by Intel from IOActive. The study itself does not describe in any kind of detail the steps they took to prove the AMD system is unprotected from malware, but it does describe the technologies they believe give Intel the competitive advantage in security.

The main feature is considered to be Threat Detection Technology (TDT), which based on another article from ArsTechnica enables antivirus software to scan memory contents using GPU acceleration and also helps heuristic algorithms detect suspicious behavior based on system telemetry data. Both features are nice and all, but they are 100% bound to the performance of the active antivirus software.

Based on the data I have so far, I'm tempted to think that what Intel demoed in their CES presentation was the effectiveness of Windows Defender running on two different systems, one of which had a different build of Defender.

 

[Hitman928]

I did a little more digging on this subject, and findings are a bit dubious: as required to disclose by Intel, their claims are based on a study commissioned by Intel from IOActive. The study itself does not describe in any kind of detail the steps they took to prove the AMD system is unprotected from malware, but it does describe the technologies they believe give Intel the competitive advantage in security.

The main feature is considered to be Threat Detection Technology (TDT), which based on another article from ArsTechnica enables antivirus software to scan memory contents using GPU acceleration and also helps heuristic algorithms detect suspicious behavior based on system telemetry data. Both features are nice and all, but they are 100% bound to the performance of the active antivirus software.

Based on the data I have so far, I'm tempted to think that what Intel demoed in their CES presentation was the effectiveness of Windows Defender running on two different systems, one of which had a different build of Defender.

The claim of having the best hardware based security comes from that laughable 'study' from IOActive. The actual demo was about CET, unless they are flat out lying in their presentation. There are software mitigations for things that CET is supposed to protect against but (in theory at least) there are ways around the software mitigations. It's possible that Intel used builds without the software mitigations in place on the AMD system or that having complete control of the system they setup a very unrealistic attack vector on the AMD system. As far as I can tell, Intel gave no configuration details of either system other than Intel CET being enabled on the Intel laptop. Hardware based control flow protection is a good thing, but (IMO) AMD not having it (for now) is not that big a deal as actually getting hacked this way with a properly configured system is extremely unlikely.

Think about it this way, if this were such a big threat and so easily used as Intel is trying to make it seem, they are then telling you that you should not buy any Intel based desktop, workstation, or server, as no Intel solution in these areas provide this protection. You should also avoid any pre-Tigerlake laptop. Only Tigerlake based laptops will provide it whenever Windows puts it in their mainline build. I think Intel has bigger security problems to worry about than AMD at this point in time and while CET helps Intel in this regard, it doesn't flip the script here.

 

[dacostafilipe]

I understand why they choose the 5900X, but because the 5950X is better binned and has clocks better, how do you guys think it will land in terms of gaming performance?

 

[Elfear]

I understand why they choose the 5900X, but because the 5950X is better binned and has clocks better, how do you guys think it will land in terms of gaming performance?

I'm no expert but I expect gaming results to be within a 5% spread amongst the various RL and Zen 3 chips. Not enough to really notice a difference unless the game scales very well with cores.

 

[jpiniero]

I understand why they choose the 5900X, but because the 5950X is better binned and has clocks better, how do you guys think it will land in terms of gaming performance?

5950X is in a much higher price tier than the 11900K. The 5900X is more comparable.

 

[Hulk]

Given what we know about Zen 3 and RL performance it seems that Zen 2 has a bit better IPC but lower clocks than RL, making them pretty much even.

I don't see how 11900k with 8 cores is going to be comparable to the 5900X with 12 cores. Sure, there will be some single threaded apps where they will be within spitting distance of one another but in any app that is multithreaded the 11900k will be crushed.

Regardless of what Intel wants us to believe the 5800X is the comparable part to the 11900k and should therefore be price at $450.

Intel just doesn't want to be boxed in by the tiny window squeezing all of the RL parts in below $450. But that's the reality, AMD has in fact squeezed them in.

You don't get to have your 8 core part compete with a 12 core part pricing-wise when per core performance is about the same because you're Intel and you say it's so. The old Intel mind trick.

 

[Zucker2k]

I think we are over-analyzing Intel's comparison of the 11900k to the 5900x here. AMD pushed the 5900x as their best gaming chip. Simple.

 

[blckgrffn]

I think we are over-analyzing Intel's comparison of the 11900k to the 5900x here. AMD pushed the 5900x as their best gaming chip. Simple.

It will be way more interesting to compare the lower tier SKUs anyway, whatever flagship-1 is from Intel (10850k/9700k SKU) is likely to be a much better value and used much more widely.

Given the 14nm process I expect that availability of this slightly lesser SKU to be excellent and for it to be abundant and present in tons of prebuilt systems.

I am also hoping to see a really competent sub $200 6C/12T part - ideally closer to $150. For quite a few people right now that's a great value spot, and if you can get a board and CPU for $300 - maybe even get a $25 aftermarket cooler too - if we could ever buy GPUs it would be solid starting point for the $800-$1k gaming rigs.

Right now I've helped build a couple 10400 based rigs because the 10th Gen six core K variant is just priced too high IMO. I'd build a 9700k/B365 for cheaper and count on the generous stock clocks to carry a ton of performance.

 

[Hulk]

I think we are over-analyzing Intel's comparison of the 11900k to the 5900x here. AMD pushed the 5900x as their best gaming chip. Simple.

Then why didn't Intel just go ahead and compare it to the 5950X? Or simply use an application that is not efficiently multithreaded and is suited to Sunny Cove architecture to show the 11900k in a better light than the 5950X. My point is that their comparison and the one I just imagined are not valid comparisons of overall performance among the parts. It's a mind trick they are trying to push and the more they say it the more real it becomes to people.

 

[dmens]

The claim of having the best hardware based security comes from that laughable 'study' from IOActive. The actual demo was about CET, unless they are flat out lying in their presentation. There are software mitigations for things that CET is supposed to protect against but (in theory at least) there are ways around the software mitigations. It's possible that Intel used builds without the software mitigations in place on the AMD system or that having complete control of the system they setup a very unrealistic attack vector on the AMD system. As far as I can tell, Intel gave no configuration details of either system other than Intel CET being enabled on the Intel laptop. Hardware based control flow protection is a good thing, but (IMO) AMD not having it (for now) is not that big a deal as actually getting hacked this way with a properly configured system is extremely unlikely.

Think about it this way, if this were such a big threat and so easily used as Intel is trying to make it seem, they are then telling you that you should not buy any Intel based desktop, workstation, or server, as no Intel solution in these areas provide this protection. You should also avoid any pre-Tigerlake laptop. Only Tigerlake based laptops will provide it whenever Windows puts it in their mainline build. I think Intel has bigger security problems to worry about than AMD at this point in time and while CET helps Intel in this regard, it doesn't flip the script here.

It is pure garbage. They claim to use "machine learning" to feed "heuristics models" to detect threats. How exactly? With what training? How do they distinguish between real threats and just regular legitimate software running at the hardware layer? The entire notion is laughable.