Intel Benchmark Shows PCIe 4.0 Storage Performance of Core i9-11900K Exceeding AMD Ryzen 9 5950X by 11 Percent

Tsing

The FPS Review
Staff member
Joined
May 6, 2019
Messages
12,222
Points
113
intel-11-th-gen-desktop-rocket-lake-s-1-1024x576.jpg
Image: Intel



The PCIe Gen 4 storage performance of Intel’s 11th Gen Core i9-11900K processor is substantially better than that of AMD’s Ryzen 9 5950X. This is according to a new benchmark shared by Intel chief strategist Ryan Shrout and storage technical analyst Allyn Malventano, which show the Rocket Lake-S processor with a 11 percent lead over red team’s Zen 3-based CPU in PCMark 10’s Quick System Drive Benchmark. A Samsung SSD 980 PRO (1 TB) was used in the test.



“At #CES21 we looked at Rocket Lake-S gaming,” Shrout tweeted. “Here’s a sneak peek of Core i9-11900K PCIe...

Continue reading...
 
How many side-channel attacks did Intel leave open to get the performance boost?
Edit Re: using the x16 slot - I expected better of Shrout and Malventano.
 
Last edited:
All running the drive on the x16 slot does is bypass the PCH and go directly to the CPU's PCIe controller. I agree it isn't real world, but AMD CPU's have been doing this for awhile as they have dedicated PCIe lanes for storage as it is. Interestingly enough, using an SSD in Gen 3.0 mode via the dedicated CPU based PCIe lanes hasn't ever yielded any real advantages over the PCH based slots on the Intel boards. I think its an interesting test, but it would have made more sense to repeat the same test with the SSD installed in a conventional PCH based M.2 slot and show the disparity between the two if there is one.
 
How many side-channel attacks did Intel leave open to get the performance boost?
I'll say that I reacted poorly upon reading this... and that I'm working on getting over that. It ain't you, it's me.

I will also say that most of these are related to poor engineering decisions that Intel never expected to have to face the music for. They really did not plan to use Skylake as long as they did, and they really did not expect Skylake to have the install base that it still does, which is what made its flaws so lucrative to target.

What I draw from that is that, with all the fervor and all the pivoting Intel has had to do to try to keep up with demand while their manufacturing advancements have spun in a circle, Intel isn't likely to repeat these mistakes.

They have paid a heavy price for their recent stumbles and it's reasonable to me to think that they've already worked to avoid repeats.
 
I'll say that I reacted poorly upon reading this... and that I'm working on getting over that. It ain't you, it's me.

I will also say that most of these are related to poor engineering decisions that Intel never expected to have to face the music for. They really did not plan to use Skylake as long as they did, and they really did not expect Skylake to have the install base that it still does, which is what made its flaws so lucrative to target.

What I draw from that is that, with all the fervor and all the pivoting Intel has had to do to try to keep up with demand while their manufacturing advancements have spun in a circle, Intel isn't likely to repeat these mistakes.

They have paid a heavy price for their recent stumbles and it's reasonable to me to think that they've already worked to avoid repeats.
Fair point about Skylake. The original Specter and Meltdown attack vulnerabilities go all the way back to Nehalem, though. There was a pattern through a few CPU redesigns. I would hope they have kept that in mind for their more recent designs.
 
The original Specter and Meltdown attack vulnerabilities go all the way back to Nehalem, though.
The root of these attacks, or the technique behind them, is based on how SMT works.

It's interesting that the redesigns needed to mitigate the attack vectors likely don't have any performance impact, they just require a bit more engineering to implement, and companies (not just Intel!) didn't think at the time of design that the feature would be targeted and thus didn't feel that the investment to engineer protections was needed.

They were wrong this time, but it cannot be understated that all modern technology is riddled with vulnerabilities that stem from risk management decisions throughout their design and life cycles.

And it should be understood that these are unavoidable!

They are artifacts of 'human in the loop' design. Even AI that designs AI that designs AI would still have human design at its root too.

So while Intel very much should take flak for these choices, we should also understand that these choices aren't an 'Intel' thing, they're a 'human' thing :)
 
So while Intel very much should take flak for these choices, we should also understand that these choices aren't an 'Intel' thing, they're a 'human' thing :)

While I do agree a mistake was made... I have issue with a few things.

1. Intel continued to release CPU's that were vulnerable and needed performance impacting microcode to alleviate.
2. their new CPU's were sold as better performers but only without the microcode update.
3. They released NEW CPU's that had this problem AFTER it was discovered.

Intel would have had MUCH more respect from me if they halted. Tool their chips that were problematic, admitted that they are and sold them to integrators super cheap in order to get them out and then not release a new CPU until they had a working unit that was not exposed to the existing vulnerabilities.

Businesses and individuals that refreshed their hardware AFTER the vulnerability was in the wild for a not short period of time, with CPU's that were released AFTER the vulnerability was out there, were buying security compromised hardware in order to do business. Many of them with a reasonable expectation that the new chips did not have the vulnerability.

From a business standpoint I see why they did it. People didn't know and those that did were locked into Intel for many reasons from Policy to lack of knowledge to.. just too lazy to change.

For retail consumers.... this was horrific. Great now you have to count on bob and tom patching out the many vulnerabilities around the hardware that were at the time of purchase. YEARS old in many cases. And doing research to be sure that the BIOS was updated was placed on the consumer many of which just didn't know to THINK about a year+ old vulnerability.

Intel took the money and abandoned consumers to suffer the ill effects counting on their enterprise partners to be more savvy and be ok.

That is why I don't like Intel right now. And that is why I would tell any consumer in the market for a PC or CPU today to go AMD or Apple now with their new M1 based systems.
 
were buying security compromised hardware in order to do business.
Here's the reason: Intel was selling these parts because if they didn't, their customers wouldn't be able to do business.

It's risk management, pure and simple, and the business risks of not being able to do business are pretty high!
 
Here's the reason: Intel was selling these parts because if they didn't, their customers wouldn't be able to do business.

It's risk management, pure and simple, and the business risks of not being able to do business are pretty high!
If only another company had been selling CPUs that had fewer vulnerabilities.
 
If only another company had been selling CPUs that had fewer vulnerabilities.
If only another company had been selling CPUs that could do everything else Intel's CPUs could!

Like get work done :)

I get that Ryzen is essentially there in most cases (more in some, less in others), but Skylake was up against the Dozers.
 
If only another company had been selling CPUs that could do everything else Intel's CPUs could!

Like get work done :)

I get that Ryzen is essentially there in most cases (more in some, less in others), but Skylake was up against the Dozers.

It's pretty telling that Intel only pulled their heads out and said. "Ohhh yea security" once AMD had them at an IPC disadvantage while being more secure.

So Intel is in this ****ter and in my book needs to prove itself to be high performance AND security designed before I personally buy another of their CPU's. And before I recommend another Intel CPU for the company I work for as an Engineer/Architect.
 
It's pretty telling that Intel only pulled their heads out and said. "Ohhh yea security" once AMD had them at an IPC disadvantage while being more secure.
Not at all. Skylake predates Zen by four or five years, and AMD had the benefit of the research that had been published in the interim. Intel had to backport all kinds of architectural improvements because they couldn't make 10nm or 7nm work. They never planned to do anything on 14nm after Skylake.

So Intel is in this ****ter and in my book needs to prove itself to be high performance AND security designed before I personally buy another of their CPU's. And before I recommend another Intel CPU for the company I work for as an Engineer/Architect.
I guess that depends.

On security alone, I cannot disagree, but as I noted above there's risk involved that must be balanced. It should also be said in this context that a perfectly secure system will never do any work. Business comes first, otherwise there is nothing left to 'secure'.

Where AMD can fall short at the enterprise level is... well, worthy of a PhD study. Everything from supply to compatibility to stability with legacy code is in play, and that's not just the CPUs, but also the platforms that they must come with, and the hardware that they all must also operate with.

Intel has the 'establishment' advantage here, not just in terms of mindshare, but also in literally everything else. And to put that somewhat into perspective, the hardware cost, while hitting five-figures easily when it comes to servers, is still just a small fraction of the value that is generated. Price versus performance can matter very, very little once a minimum performance threshold has been met, and other things then matter far more.

Intel is still deserving of criticism, especially given the follow on effects of their negligence in the enterprise, but they also still deliver value.


Another perspective: mobile.

I can't get a Latitude with a nice AMD APU. Can't get a Macbook Pro with one either, or an XPS. Or whatever HP pretends are equivalent in their lineup. I'm leaving Lenovo out on purpose, and have no idea if anyone else makes something comparable; for certain they don't have Dell's service if they do.

And I do blame AMD for that a bit. Platform stability, drivers, battery life... that all counts, on top of actually getting their products in vendors' hands, in volume.

And I say that while typing on an XPS 15 that I had initially searched for an AMD solution instead of the i7 I wound up getting. Not only are the AMD parts unavailable in nicer laptop chassis, but I also ran across multiple reputable reviews that listed stability issues with content creation workflows.

Yeah, I'll deal with the i7. CPU performance actually wasn't a primary concern; stability and ergonomics were.
 
I can't get a Latitude with a nice AMD APU. Can't get a Macbook Pro with one either, or an XPS. Or whatever HP pretends are equivalent in their lineup. I'm leaving Lenovo out on purpose, and have no idea if anyone else makes something comparable; for certain they don't have Dell's service if they do.

And I do blame AMD for that a bit. Platform stability, drivers, battery life... that all counts, on top of actually getting their products in vendors' hands, in volume.
Wasn't there a law suit against Intel for strong arming OEMs into not using AMD CPUs in their products like 15-20 years ago? The first paragraph I quoted there makes me ... suspect.

Sure, part of it could be all the things you talk about in the second paragraph, but... the lack of reasonable AMD products I think has more to it than just that.
 
Yea for business purposes AMD hasn't cracked that market yet with their CPU's in the mobile space.

For density and scientific/business servers AMD is right there unless customers are running some truly legacy code. (Admittedly MANY are.)

Businesses need to go to their vendors and insist on it. But if you're putting a T1 card in a computer today... you'll want Intel, and you'll want older CPU's with supporting controllers. These things are the most sensitive pieces of crap I've had to deal with in a VERY long time.

So yes for MOST enterprise compute... Intel is still the way the cookie crumbles. But as tech continues to move forward and Intel continues to slip soon enough those contracts are not going to seem as palatable to the vendors like Dell.

Look at the Dell rack servers... you can build out an AMD unit pretty easily now. Year and a half ago you needed a dousing rod of HTML to find the page that had AMD cpu's and sku's on it for servers.

AMD is gaining mindshare and as they increase that they get market. As that grows the software vendors start writing code to take advantage of the small but growing install base. It feeds into itself until the point that we have stability parity between Intel and AMD.

Once that moment comes Intel will know they are in it for real and probably start spending some real money to get a tech advantage again.

Oh... and hopefully be security minded.
 
Wasn't there a law suit against Intel for strong arming OEMs into not using AMD CPUs in their products like 15-20 years ago? The first paragraph I quoted there makes me ... suspect.
Yup.

I don't want to excuse Intel's ethical blunders, but as a business major and a veteran... if you ain't cheatin', you ain't tryin'. Quoted from a JAG officer.

In business, if you have to run to the courts for something that isn't straight up corporate espionage, i.e. your competitor broke into your facility and stole your technology, it means that they outmaneuvered you.

Sure, part of it could be all the things you talk about in the second paragraph, but... the lack of reasonable AMD products I think has more to it than just that.
Think about how many more CPUs Intel produces. Intel is also limited by their own fabs, while AMD is limited by the whims of TSMC, for whose capacity they have to compete with other industry behemoths like Apple and Nvidia, among many others, including rumored Intel business.

If AMD wants to get their products in Dell's premiere product lines, they're going to have to compete with Intel across the board. Not just at the enthusiast level, where we're willing to work around certain shortcomings, but also things like mobile where the whole platform matters so much more.
 
Once that moment comes Intel will know they are in it for real and probably start spending some real money to get a tech advantage again.

Oh... and hopefully be security minded.
They never stopped; they just totally boondoggled their fab advancements. Doesn't matter what they design if they can't fabricate physical parts, right?

Which is why we're getting heavy backports in the form of Rocket Lake. I'd recommend paying attention to how RKL does security wise, I know I will be.

AMD is gaining mindshare and as they increase that they get market. As that grows the software vendors start writing code to take advantage of the small but growing install base.
AMD needs more production. They have stupidly attractive products, but one cannot depend on being able to actually buy them!

And that's true for regular consumers. How do you convince an enterprise that needs a stable SKU for a deliverable to take a chance on an AMD spec? How can Dell or HP even assure them that there will be more of that SKU available when needed?
 
And that's true for regular consumers. How do you convince an enterprise that needs a stable SKU for a deliverable to take a chance on an AMD spec? How can Dell or HP even assure them that there will be more of that SKU available when needed?
Dell must have some confidence because amd cpus are featured in parity with Intel cpus now.
 
Become a Patron!
Back
Top