This Is BAD, AMD Basically Lies About CPU Performance

Z

ZeroBarrier

[H]ard|Gawd
Joined
Mar 19, 2011
Messages
1,032
I didn't see this posted or even mentioned here; forgive me if it has been.

It seems AMD is back to misrepresenting their first party benchmarks that so many users here swear can be trusted. As always, do not believe a faceless corporation and wait for trusted third parties to reviews, aggregate and make as an informed decision as possible.


View: https://www.youtube.com/watch?v=BoCYyg6ecLM
 
Maybe I'm weird. But has anything really changed?

Window sucks. Oh, really? Never knew that.
 
Bigbacon said:
Don't they all misrepresent benchmarks? Isn't that the point of marketing.
Click to expand...
I mean... yes and no. Usually what you see is picking and choosing the benchmarks that look best to you. So like if an Intel CPU were faster at gaming and SPECPerf but slower at Cinebench and x264, then they'd choose to show the games and SpecPerf but leave off Cinebench and Handbrake. That's not completely truthful, but it is a normal level of cherry picking. This is a different level of dishonesty: This is deliberately setting things up to manipulate the results, not just putting your best foot forward. They are using another component to hamstring the performance to bring down the competition to their level. It's real shady.

Red Falcon said:
I would accept bunk benchmarks from any large/megacorp over hiding severe hardware flaws like Meltdown, Spectre, Foreshadow, etc.
Click to expand...
Intel didn't hide Meltdown, etc, they were new flaws discovered by Google Project Zero team, which Intel then patched. If you think you can make something with no security bugs, ever, well then, be my guest, but I'm going to bet you fail. Also of note, Meltdown isn't just an Intel issue, it affects ARM Cortex A75 cores, and IBM Power. Likewise Spectre affects ARM and AMD in addition to Intel. Also if you want to pick at AMD directly, they have one of their own called Zenbleed which is peculiar to their CPUs only, not Intel as well as plenty of others that never got a fancy name, just CVEs.
 
Last edited:
Lakados said:
Misrepresenting benchmarks is one thing, but changing flags to hinder non AMD CPU’s to make your own ones look better is a straight up Intel play.

We shit on Intel for doing it and you better bet we’ll shit on AMD for doing it too.
Click to expand...

I'm not taking a side, I'm just saying this is normal and to be expected, this is mind blowing.
 
Yeah, I suppose this is kinda news. Like the weather.
Whether or not this was malicious, we'll never know. As stated by others, these are insanely complicated things.

Regardless, who would trust any Corps own benchmarks? It's like trusting the used car salesman. You just don't do it.
 
Sycraft said:
Intel didn't hide Meltdown, etc, they were new flaws discovered by Google Project Zero team, which Intel then patched. If you think you can make something with no security bugs, ever, well then, be my guest, but I'm going to bet you fail. Also of note, Meltdown isn't just an Intel issue, it affects ARM Cortex A75 cores, and IBM Power. Likewise Spectre affects ARM and AMD in addition to Intel. Also if you want to pick at AMD directly, they have one of their own called Zenbleed which is peculiar to their CPUs only, not Intel as well as plenty of others that never got a fancy name, just CVEs.
Click to expand...
Intel was well aware they weren't doing the necessary memory checks within their instructions and rings, unlike AMD had been doing the whole time.
This is why Intel held the performance crown until 2018, and why these hardware-level exploits go back as far as the Pentium Pro from 1995.

Even if Intel "didn't know", they certainly knew it was worth going with the risks for the performance gains, and were complicit with doing so until they were caught, which they eventually were.
 
Red Falcon said:
Intel was well aware they weren't doing the necessary memory checks within their instructions and rings, unlike AMD had been doing the whole time.
Click to expand...
...so why then did Google discover the same issues in ARM and Power?
Red Falcon said:
Even if Intel "didn't know", they certainly knew it was worth going with the risks for the performance gains, and were complicit with doing so until they were caught, which they eventually were.
Click to expand...
Again, are you going to give this same level of hate to AMD? They have plenty of AMD-only exploits (and some cross architecture ones) they just don't tend to make the news and get fancy names. This year they had a couple fairly serious ones, one being an arbitrary code execution via SMM (CVE-2023-20577). They have a page listing all the ones they are notified about, what it affects, fixes, etc, as do all good vendors.

I'm not trying to shit on AMD, but you are on an un-informed Intel hate train as though they are the only company that has CPU vulnerabilities. Hardware attacks happen and are an increasing area of security research. All companies have issues with them. AMD is no different.
 
N
Bigbacon said:
I'm not taking a side, I'm just saying this is normal and to be expected, this is mind blowing.
Click to expand...
Not Really though, adding a discrete GPU then failing to mention it being included and making no mention of it, then presenting those scores as your APU aren’t exactly the same as cherry-picking and tuning favourable settings.

This is some peak level Intel Shenanigans, that I had really hoped AMD would not sink too.
It wreaks of either desperation or incompetence and I don’t like thinking AMD capable of either.
 
Sycraft said:
...so why then did Google discover the same issues in ARM and Power?
Click to expand...
Other corporations started looking at their designs once the exploits started being demoed and tested on other ISAs and platforms.
It was less of a shock about them since those were mostly newer designs made within the decade, you know, not spanning back to 1995 like Intel's design choices.

Sycraft said:
Again, are you going to give this same level of hate to AMD?
Click to expand...
If hardware-level exploits, that when patched lost 40-60% of performance depending on the task, yes.
If that were to happen to AMD and ARM, it would make me consider erek's hard-on for RISC-V.

Sycraft said:
I'm not trying to shit on AMD, but you are on an un-informed Intel hate train as though they are the only company that has CPU vulnerabilities.
Click to expand...
I was there in Q1 2018 when this shit went down and had to patch more systems than I care to remember and lost planned time off for it.
That's not even counting the personal devices that were negatively effected and had to be replaced out of pocket, which Intel certainly didn't pay for.

Sycraft said:
All companies have issues with them.
Click to expand...
You mean like Intel does in this thread right next to this one that we are having this very discussion in, of which has flawed microcode which may potentially have damaged CPUs?

Uninformed my ass.
Get over yourself.
 
Last edited:
Just because it is "business as usual" for the likes of Intel, AMD, and practically everyone else that makes cpus/gpus doesn't mean we should fail to point it out, or just accept the status quo.
 
Have they not had a not so bad track record for a while ?

RX 6600 CPU benchmark and that level of manipulation is not what AMD have been doing , maybe because they are old not for enthusiast CPU they did not thought people would check-care.

We would expect picking application-game that make you look good, not make much effort when choosing competitor ram and so no, etc... but there some boundaries
 
Yap, no worse that what Intel does too.

Same marketing bullshit they all pull.
 
ZeroBarrier said:
It seems AMD is back to misrepresenting their first party benchmarks that so many users here swear can be trusted.
Click to expand...
I must be blocking some people that you don't; I haven't seen a single case of a user here who doesn't have the attitude that one should always wait for independent benchmarks.
 
Red Falcon said:
Get over yourself.
Click to expand...
Get over yourself. This is such a fanboy rant. Intel had exploits, a new class of exploit, that affected a number of architectures, that the researchers themselves said would only be the beginning of a class of exploits (they were right) and yet Intel is the only bad one here in your book. Acting like you had to replace persona devices because of this is ridiculous. If you chose to... ok, but to me all that says is that you got spooked by the news stories and didn't look into the reality of the vulnerabilities or the mitigations and that you continue to not look into the others that happen with CPUs but don't get big stories, just CVEs rated high (Meltdown was a medium, BTW, as are most of the Spectre ones).

If your response to The Spectre/Meltdown thing was to blame Intel and get mad at them for having to patch systems, and replace all your personal devices, ya I'm going to call that uninformed. My reaction was to read the CVE, read the analysis by the researchers, conclude that it was an issue, but not a bad one, mostly of concern to our VM servers. Then write up a summary for leadership, be annoyed with the media since they were blowing up something way beyond its actual impact, when they've ignored far more serious issues, and roll out patches in order of system risk level. You know, my job. As for my personal devices, I didn't worry as the impact was minimal. Meltdown is of little concern to a single-user environment, particularly one where you are running with a privileged account (as it would be much easier to simply use that privilege to hook debug another process, rather than do a side channel attack) and Spectre was mostly a risk on a single user system for websites, which the browsers updated against quickly.

[Spectre] said:
What did you say about me?
Click to expand...
Come on, we know you pick fights with all the processors. We've seen you drunk, outside the bar, yelling at ARM to come fight you :D
 
Remember! AMD would gladly slit your throat if it garenteed a 2% gain in their share price (and if it was legal)... In fact they would be feduciarily obligated to do so!

Just like Intel.


They don't care about you. You are not their customer. You are the one of the many products they sell to their shareholders. Their shareholders are the real customers, and if they don't care about cheating in benchmarks, than it's not an issue.

It's only an issue when shareholders care.

You don't matter to them as long as you still throw them money.
 
ZeroBarrier said:
I didn't see this posted or even mentioned here; forgive me if it has been.

It seems AMD is back to misrepresenting their first party benchmarks that so many users here swear can be trusted. As always, do not believe a faceless corporation and wait for trusted third parties to reviews, aggregate and make as an informed decision as possible.


View: https://www.youtube.com/watch?v=BoCYyg6ecLM
Click to expand...

The specifics pointed out in this video, are interestingly bad optics for AMD. Apparently they have claimed in their computex slides, that the Zen 3 refresh parts 5800XT, 5900XT---are somehow as good or slightly faster in gaming, than Intel 13th Gen. (5800XT Vs. 13600k and 5900XT Vs. 13700k). CPUs which are 2 generations newer. There should be virtually no situations where either Zen 3 CPU would be faster than the compared 13th gen CPUs.

The supposed data was taken from a test setup where they used DDR4 3200 for those 13th gen CPUs. And both CPUs were gaming with an RX 6600. So they basically found some sort of low performance anomoly which may have been slightly repeatable, and then published that as a slide to show off at Computex.

Hardware Unboxed says they should have simply announced the CPUs as a goodwill legacy support/value option for AM4. And not published any performance data. Especially not something so specifically manufactured.
 
Sycraft said:
I mean... yes and no. Usually what you see is picking and choosing the benchmarks that look best to you. So like if an Intel CPU were faster at gaming and SPECPerf but slower at Cinebench and x264, then they'd choose to show the games and SpecPerf but leave off Cinebench and Handbrake. That's not completely truthful, but it is a normal level of cherry picking. This is a different level of dishonesty: This is deliberately setting things up to manipulate the results, not just putting your best foot forward. They are using another component to hamstring the performance to bring down the competition to their level. It's real shady.
Click to expand...
Without really disagreeing, I have to ask who's going to pair those CPUs with something like a 4080. Yes, it's shady, but IF it's representative of the kind of GPU people would pair with that CPU, it could be useful knowledge.
 
1_rick said:
Without really disagreeing, I have to ask who's going to pair those CPUs with something like a 4080. Yes, it's shady, but IF it's representative of the kind of GPU people would pair with that CPU, it could be useful knowledge.
Click to expand...
I mean, you have to if you want to rate CPU performance in games. The truth of the matter is most of the time the CPU matters only a small amount, if at all, in games because you are GPU (or scanout) limited. That's why review sites will tell you to not get a big chonk of a CPU if all you are doing is gaming. You don't need a 7950X3D or 14900K to play games. To see the difference, what sites have to do is not only pair with a chonk of a GPU, but also turn down rez. So like Hardware Unboxed does 1920x1080 on a 4090 when they want to test CPUs. That way you see the CPU scaling the best. Doesn't mean you should pair the low end CPUs with a chonk GPU, just lets you see how they compare in CPU limited applications. They chat some about it here and why they do it they way they do, and what useful information you get.

If you hamstring the CPU with a low-end GPU, particularly at higher resolutions, it can lead to the misleading result that they are the same when they aren't. Same with any component. Like CPU speed actually matters to network transfer speed... but only for high-speed interconnects. If I published a chart of network transfer speed vs CPU using a gig NIC, it would show there had been basically no performance improvement in well over a decade. However, if I stuck a 400gig NIC in there, suddenly you'd see big differences as now it was enough to overtax CPUs.

Axman said:
Intel knew.
Click to expand...
Why did they know, but nobody else did? Meltdown was Intel, one ARM core, Power 8 and 9, as well as zArchitecture. Spectre was even wider and worked on basically any processor with speculative execution, including Intel, AMD, ARM, Power, z, even VIA (yes they still make CPUs, surprisingly). Spectre, or rather speculative execution side channel attacks, are also an ongoing class of attack, new ones crop up now and again, they just rarely make news past something like a SANS bulletin.

So again, I ask: Intel knew, but everyone else didn't?
 
Red Falcon said:
Because Intel and Apple totally never ever do this. :whistle:
I would accept bunk benchmarks from any large/megacorp over hiding severe hardware flaws like Meltdown, Spectre, Foreshadow, etc.
Click to expand...
Apple is the worst with this. It's absolutely infuriating when they compare their current benchmarks to years old hardware nobody cares about.
 
KazeoHin said:
Remember! AMD would gladly slit your throat if it garenteed a 2% gain in their share price (and if it was legal)... In fact they would be feduciarily obligated to do so!
Click to expand...
This is a misunderstanding that tend to be something repeated enough that I am not sure if people are half-serious, just joking, etc...

KazeoHin said:
You don't matter to them as long as you still throw them money.
Click to expand...
That a bit of a distinction without a difference, wanting people to still throw them money is exactly what mattering to them mean.
 
LukeTbk said:
This is a misunderstanding that tend to be something repeated enough that I am not sure if people are half-serious, just joking, etc...


That a bit of a distinction without a difference, wanting people to still throw them money is exactly what mattering to them mean.
Click to expand...
I'm to understand that feduciary duty means that if a entity that has feduciary duty to a party is made aware of a legal opportunity to benefit the party financially but chooses not to (even in the absence of any direct financial harm) , the party has the legal right to press charges and demand compensation.

So in other words, if it were legal to slit your throat, and doing so was highly likely to financially benefit AMD or Intel (especially at a level of 2% increase in their per-share price), they would have a LEGAL OBLIGATION to slit your throat, lest shareholders could demand compensation from them for not taking the opportunity presented.
 
Shoganai said:
Apple is the worst with this. It's absolutely infuriating when they compare their current benchmarks to years old hardware nobody cares about.
Click to expand...
But half the time that years old hardware was what’s powering their previous generation. Apple for their Intel and AMD stuff was 3 generations old when they finally got around to replacing it. Apple was terrible for releasing new products with 2 year old parts.
 
KazeoHin said:
I'm to understand that feduciary duty means that if a entity that has feduciary duty to a party is made aware of a legal opportunity to benefit the party financially but chooses not to (even in the absence of any direct financial harm) , the party has the legal right to press charges and demand compensation.
Click to expand...
It is really large on how long of a time horizon people in charge can defend a decision.

Try to find a big company that do not give money to charity for example, they could obviously make more money next quarter (and would be perfectly legal) by stopping all those charities, but an argument can be make that those good social standing help recruitment, protection from government and social pressure, as a publicity and when talking the next 50 years are a good for the company bottom line. Not manipulating benchmark result by a lot would be very easy to defend by the executives (and would be hard to prove that it would have been a benefit to do so to start with, the blowback being always a risk).

Feducirie obligation between management and stockholder does include financial return but it is not the only variable and if a company publicly have some social responsibility goal (like limiting environment damage) known by the investor that agreed to it by still investing, obviously now it opens the door a lot for a long list of action that do not optimize profit.

KazeoHin said:
So in other words, if it were legal to slit your throat, and doing so was highly likely to financially benefit AMD or Intel (especially at a level of 2% increase in their per-share price), they would have a LEGAL OBLIGATION to slit your throat, lest shareholders could demand compensation from them for not taking the opportunity presented.
Click to expand...
And like I said, this is a common misconception meme on the Internet.

Courts typically defer to the business judgment of directors, meaning as long as their decisions are made in good faith, with reasonable care, and in what they believe to be in the best interest of the company, those decisions are protected. This can include decisions that prioritize long-term value over short-term profits, investments in employee welfare, corporate social responsibility initiatives, or other stakeholder considerations. ... Directors and officers are expected to act in a manner that they reasonably believe to be in the best interest of the corporation, which can include a variety of strategic and ethical considerations.

A ceo would easily win the battle that even if next quarter profit would have been 2% higher, killing customer over the next 50 years would have been a bad goal, the company reputation (to get loan at good interest rate, good price from business partner-provider, habitability to recruit the best talents at a reasonable price, cities and states collaboration or at least not active fighting their next project, not having one day politicians trying to score political point by starting antitrust lawsuit or other against them, are not worth that short term money), anything that hurt the reputation of the company to customer would be an easy decision to defend. They can obviously vote to replace them if they disagree, but sue them and have them pay fine for those type of action, no, obviously not.
 
We mustn't forget, AMD are the PEOPLES processor company. They are supposed to be above these types of shady shenanigans. For many it's painful to see them lower themselves(myself included).
 
  • Like
Reactions: Niner
like this
Brent over at The FPS Review is covering it now.

I found it on the youtube channel, haven't seen it on the actual website yet.


View: https://www.youtube.com/watch?v=M1p8WFhiwHc

Edit, just finished watching his video and I have to disagree with Brent. He says they aren't lying because they are not fabricating the numbers, but in my opinion, they are lying by portraying both CPUs are equal when they know full well they are are not equal.
 
Can’t believe no one has posted the “anyway” meme yet LOL
I’ve never put much stock in any first party benchmarks, they all want to show their product in the best possible light even if it’s completely unrealistic. Just ignore them and move on with your life and 3rd party benches.
 
gtrguy said:
Can’t believe no one has posted the “anyway” meme yet LOL
Click to expand...
Specially for old product we can almost get perfect estimate of their actual performance just looking at their specs. (and maybe why they did not think people would care and even catch it)
 
ZeroBarrier said:
Brent over at The FPS Review is covering it now.

I found it on the youtube channel, haven't seen it on the actual website yet.


View: https://www.youtube.com/watch?v=M1p8WFhiwHc

Edit, just finished watching his video and I have to disagree with Brent. He says they aren't lying because they are not fabricating the numbers, but in my opinion, they are lying by portraying both CPUs are equal when they know full well they are are not equal.
Click to expand...

What AMD shows, is the difference between ReBAR (which is universal) and Smart Access Memory (which is proprietary).
SAM is better for GPUs, hands down, AMD tailor-fits the solution to the problem, and it's good which does give an advantage to keeping to an all-AMD system in the mid to low class.
But this is a highly misleading result at best because they are not showing CPU results, they are showing platform results. the CPU is not the limiting factor in those tests and AMD knows it!

They might not be lying, but they are pulling their truth out of a technicality, and it's a "feels bad" all around.
Like that guy who insists it's not cheating because "Rules as Written", it's not cheating via semantics, but you certainly don't feel good being on the receiving end of it, that's some D&D villain fine print level shenanigans, and it not only deserves to be called out but should be called out wherever possible, regardless of who is doing it.

Side note:
Great Video, Liked and Subscribed for more.
 
You must log in or register to reply here.
Back
Top