Intel & AMD Processor Hierarchy

[guru7of9]

Yes, there are going to be some poor Passmark submissions for every processor, including every Alder Lake CPU. As I have mentioned before, when I build systems I try to choose the best performing components to try to create a highly performant balanced system. The results of that effort are usually a system that not only performs measurably better in the applications I run but also scores much higher than the average score in Passmark for a given CPU. So I believe there can be a good performance correlation between applications and some synthetic benchmarks.

I believe one reason the 5800X3D has not scored as well is that it cannot be easily overclocked. Many of the online articles on 5800X3D performance were done with stock CPU settings, yet I believe many Passmark submissions were done with overclocked, optimized systems, thus better representing what you can expect with optimal system components. And not all online sites evaluating the 5300X3D show a 7% performance difference between 5800X and 5800X3D in single-thread performance. For example, if you scroll down to the Cinebench R23 single-core test at this link, there is a 12% difference when running stock. The gap would be wider if the 5800X was overclocked. And if you do the math, it shows the 12900ks is 53% faster (2141 vs 1398), which is close to what the Passmark single-thread reports.

AMD Ryzen 7 5800X3D Review: Gaming-First CPU

Making CPU cores faster rather than adding more cores is the best way to boost PC gaming performance. That's why AMD has supercharged their 8-core, 16-thread CPU...

www.techspot.com

You are saying that the lack of overclocking is the reason why single scores in passmark are low.
Not quite sure why you are so fixated about lack of overclocking with the 5800X3D and also obsession with passmark? Benchmarks are way to measure system performance but at the end of the day, some have more credibility than others!
Overclock setups and 5800X3D cpus will vary in performance !
The standard 5800X doesnt overclock that well (its already cranked pretty hard from factory) and for gaming a lot of the time it hurts performance. Can boost apps ok at times but the improvement is still not massive.
Most reviewers would agree that from their testing they have found that the Ryzen 5800X3D is marginally slower in single thread and multithreaded apps and faster in quite a few games . Some quite a bit and some negligible!

 

[DataMeister]

In what world is a i7 2600k lower tier than a Ryzen 3 1200?

In this world only.

Passmark reflects the same conclusion.
i7-2600k = (MT: 5497 / ST: 1743)
Ryzen 3 1200 = (MT: 6304 / ST: 1927)

 

[M42]

Not quite sure why you are so fixated about lack of overclocking with the 5800X3D and also obsession with passmark? Benchmarks are way to measure system performance but at the end of the day, some have more credibility than others!
Most reviewers would agree that from their testing they have found that the Ryzen 5800X3D is marginally slower in single thread and multithreaded apps and faster in quite a few games . Some quite a bit and some negligible!

I believe it is likely that AMD sent cherry-picked 5800X3D's to reviewers, so the performance in the reviews may not be representative of what a consumer sees. Thus, I like Passmark because it provides results across many samples of each CPU instead of just the single CPU used in most online reviewers' benchmarks. Thus, a comparison of the 5800X's Passmark score to the 5800X3D's score better represents the average performance ratio between them. The same for the 5800X3D vs 12900ks. As of today, the 5800X3D, 5800x, and 12900ks score 2959, 3483, and 4314, respectively in the Single Thread performance benchmark. This indicates the 5800x is 17.7% faster, and the 12900ks is a whopping 45.8% faster. Thus, I suspect the average 5800X3D's gaming performance is also poorer than the online reviewers were able to obtain.

 

[cryoburner]

I believe it is likely that AMD sent cherry-picked 5800X3D's to reviewers, so the performance in the reviews may not be representative of what a consumer sees. Thus, I like Passmark because it provides results across many samples of each CPU instead of just the single CPU used in most online reviewers' benchmarks. Thus, a comparison of the 5800X's Passmark score to the 5800X3D's score better represents the average performance ratio between them. The same for the 5800X3D vs 12900ks. As of today, the 5800X3D, 5800x, and 12900ks score 2959, 3483, and 4314, respectively in the Single Thread performance benchmark. This indicates the 5800x is 17.7% faster, and the 12900ks is a whopping 45.8% faster. Thus, I suspect the average 5800X3D's gaming performance is also poorer than the online reviewers were able to obtain.

As I pointed out on the last page, another "more popular" but openly anti-AMD benchmarking site showed the single-threaded non-gaming performance difference between the 5800X and the 5800X3D as being right in-line with their difference in clock rates, as expected, based on hundreds of user-submitted results. So that's more evidence that the Passmark results should not be trusted to provide an accurate representation of the relative performance of these processors. Especially since after all these weeks, there are still only 32 samples for the 5800X3D on Passmark, and it's hard to say whether or not they come from separate people, or from properly-configured systems, or even from individuals trying to purposely skew the results, something that would be easy to do with such a low sample size. You seem to be putting way too much trust in that benchmark's results, despite all other sources appearing to disagree with it, at least as far as this processor's performance is concerned.

 

[M42]

As I pointed out on the last page, another "more popular" but openly anti-AMD benchmarking site showed the single-threaded non-gaming performance difference between the 5800X and the 5800X3D as being right in-line with their difference in clock rates, as expected, based on hundreds of user-submitted results. So that's more evidence that the Passmark results should not be trusted to provide an accurate representation of the relative performance of these processors. Especially since after all these weeks, there are still only 32 samples for the 5800X3D on Passmark...

You may be proving my point... as far as we know that's just another reviewer with an AMD-supplied cherry-picked 5800X3D sample. And, how does that dispell the average of 30+ submitted benchmarks from people with no outside sponsors and nothing to gain? And just to be clear, I'm not saying every 5800X3D system is slow, The 5800X3D's performance is tied to the performance of its cache, but not every CPU's cache performs equally well.

 

[cryoburner]

You may be proving my point... as far as we know that's just another reviewer with an AMD-supplied cherry-picked 5800X3D sample. And, how does that dispell the average of 30+ submitted benchmarks from people with no outside sponsors and nothing to gain? And just to be clear, I'm not saying every 5800X3D system is slow, The 5800X3D's performance is tied to the performance of its cache, but not every CPU's cache performs equally well.

It's not a reviewer, but another user-based synthetic benchmarking site with close to 20 times the number of submissions as Passmark for the 5800X3D. I can't say I trust them to provide meaningful numbers comparing different architectures, even less so than Passmark, but at the very least their results comparing the 5800X with the 5800X3D more or less line up with the reviews, and are based on a much larger sample size of hundreds of users.

You also assume the people submitting those 32 benchmarks have "no outside sponsors and nothing to gain", when you have no way of knowing that. There could very easily be a number of submissions among them from companies using a relatively well-known benchmark to convince people like yourself to buy and promote their product over the competition, just like how there are companies that do the same to make a brand's products appear better or worse in user reviews on sites like Amazon. Or maybe multiple results were from someone who tested the processor with different combinations of memory and other hardware to test how much on an affect it had on performance. Or someone messing with the numbers just to troll others. Anyone who has browsed through the comments on certain tech news sites probably wouldn't find that possibility surprising.

Cherry-picked processors for reviews are technically a possibility as well, but that's more likely to influence overclocking results, and not so much a processor running at stock clocks. If there were widespread cases of the processors not hitting similar performance levels as what was shown in reviews when properly configured, you would likely hear about that.

 

[M42]

It's not a reviewer, but another user-based synthetic benchmarking site with close to 20 times the number of submissions as Passmark for the 5800X3D. I can't say I trust them to provide meaningful numbers comparing different architectures, even less so than Passmark, but at the very least their results comparing the 5800X with the 5800X3D more or less line up with the reviews, and are based on a much larger sample size of hundreds of users.

You also assume the people submitting those 32 benchmarks have "no outside sponsors and nothing to gain", when you have no way of knowing that. There could very easily be a number of submissions among them from companies using a relatively well-known benchmark to convince people like yourself to buy and promote their product over the competition, just like how there are companies that do the same to make a brand's products appear better or worse in user reviews on sites like Amazon. Or maybe multiple results were from someone who tested the processor with different combinations of memory and other hardware to test how much on an affect it had on performance. Or someone messing with the numbers just to troll others. Anyone who has browsed through the comments on certain tech news sites probably wouldn't find that possibility surprising.

Cherry-picked processors for reviews are technically a possibility as well, but that's more likely to influence overclocking results, and not so much a processor running at stock clocks. If there were widespread cases of the processors not hitting similar performance levels as what was shown in reviews when properly configured, you would likely hear about that.

Without benchmarking, how would anyone be able to compare objectively? The performance of the memory cache can be different between individual CPUs, even at stock clock rates, just like the choice in dynamic RAM can affect performance. The difference here is the 5800X3D's cache is almost everything that makes it faster, so if a particular test forces it to go outside of its 3D V-cache, it will be at a disadvantage. Or, if some CPUs have lower-performing cache, they will be at a disadvantage. AMD has developed a new process, so there are bound to be hiccups, which might be the reason this is the only CPU with 3D V-cache, and overclocking is not allowed.

Regarding "no outside sponsors and nothing to gain," I think it is evident that online reviewers fall outside that category. For example, if they got a CPU from AMD, it was likely cherry-picked because if AMD didn't test and validate review samples, AMD might be in for a publicity nightmare if one of those CPUs was not up to spec. AMD had everything to gain by ensuring good samples. I am sure Intel does the same with the review samples they hand out. In contrast, the people submitting to Passmark are usually just end-users using retail-purchased CPUs.

 

[guru7of9]

I believe it is likely that AMD sent cherry-picked 5800X3D's to reviewers, so the performance in the reviews may not be representative of what a consumer sees. Thus, I like Passmark because it provides results across many samples of each CPU instead of just the single CPU used in most online reviewers' benchmarks. Thus, a comparison of the 5800X's Passmark score to the 5800X3D's score better represents the average performance ratio between them. The same for the 5800X3D vs 12900ks. As of today, the 5800X3D, 5800x, and 12900ks score 2959, 3483, and 4314, respectively in the Single Thread performance benchmark. This indicates the 5800x is 17.7% faster, and the 12900ks is a whopping 45.8% faster. Thus, I suspect the average 5800X3D's gaming performance is also poorer than the online reviewers were able to obtain.

Your cherry picking argument can be used but, quite often the reviewers go out and buy the cpu themselves through normal retail channels to stop exactly that , which would negate this completely !
I don't think cherry picked cpus would stack up for all the reviews ! AMD would get found out pretty quickly and then look like fools !

 

[M42]

Your cherry picking argument can be used but, quite often the reviewers go out and buy the cpu themselves through normal retail channels to stop exactly that , which would negate this completely !
I don't think cherry picked cpus would stack up for all the reviews ! AMD would get found out pretty quickly and then look like fools !

Here's my reasoning for AMD's cherry-picking: most reviewers claim the single-threaded performance of 5800X3D is about 7% less than 5800X. However, Passmark submissions from end-users give a different number:

5800X3D - average score 2957
5800X - average score 3482 (17.7% faster)

I believe part of the 17.7% results from users being able to overclock the 5800X, which is something that cannot be easily done with the 5800X3D. However, it wouldn't account for 10.7% of the difference from 7% to 17.7%.

 

[Vox]

Why are all listed DDR5 cpus slower than DDR4 ones?

Ex.

Intel Core i9-12900K DDR4 / DDR5

100% / 93.51%

100% / 95.86%

Intel Core i5-12600K DDR4 / DDR5

90.89% / 84.32%

96.94% / 92.33%

 

[King_V]

Why are all listed DDR5 cpus slower than DDR4 ones?

Ex.

Intel Core i9-12900K DDR4 / DDR5

100% / 93.51%

100% / 95.86%

Intel Core i5-12600K DDR4 / DDR5

90.89% / 84.32%

96.94% / 92.33%

Because that's what testing revealed.

 

[M42]

Because that's what testing revealed.

But, using what speed DDR5 RAM? Other sites using high-performance DDR5 show gains over the best DDR4.

 

[shady28]

You guys need to re-do your DDR5 tests. A lot of sites got away with using crippled DDR5-4800 at first, because DDR5 was expensive and hard to obtain.

The landscape has changed, DDR5-6000 is now about the same price as some good DDR4-3600 C14 (~$270) and generally can tie with the best of DDR4. And of course, for another $125, it's easy to find even faster DDR5-6200 which is where the DDR5 AL rigs beat all the DDR4 rigs.

 

[King_V]

You guys need to re-do your DDR5 tests. A lot of sites got away with using crippled DDR5-4800 at first, because DDR5 was expensive and hard to obtain.

No they do not.

The landscape has changed, DDR5-6000 is now about the same price as some good DDR4-3600 C14 (~$270) and generally can tie with the best of DDR4. And of course, for another $125, it's easy to find even faster DDR5-6200 which is where the DDR5 AL rigs beat all the DDR4 rigs.

PCPartPicker today, for 2x16GB RAM:

• Cheapest DDR4-3600: $99.99
• Cheapest DDR5-4800 : $189.52
• Cheapest DDR5-5200 : $199.99
• Cheapest DDR5-5600 : $219.44
• Cheapest DDR5-6000: $279.99

Not all of the older systems even support DDR4-3600, but, if you're going to demand that they test the older systems at DDR4-3600 CAS 14, the LOWEST CAS available for 3600, then you'd probably better hold DDR5-6000 to that same standard, so, CAS30

• Cheapest DDR4-3600 CAS 14: $227.82
• Cheapest DDR5-6000 CAS 30: $359.99

Then, in the 12900k and 12600k review, there's also this note:

Alder Lake chips support both DDR4 and DDR5 memory, but there are several caveats tied to DDR5. As a default, DDR5 runs in Gear 2 mode, resulting in higher latency. Additionally, standard motherboards only support DDR5-4800 if the motherboard has only two physical slots. Therefore, at stock settings, Alder Lake only supports DDR5-4400 on any motherboard with four slots — even if only two slots are populated. Support drops as low as DDR5-3600 if four slots are filled with dual-rank memory DIMMs
...
In contrast, Alder Lake supports DDR4-3200 in Gear 1 mode for all processors. That can yield latency and performance advantages for the tried and true memory.

So . . exactly what are you demanding of them for testing now?

 

[shady28]

No they do not.



PCPartPicker today, for 2x16GB RAM:

• Cheapest DDR4-3600: $99.99
• Cheapest DDR5-4800 : $189.52
• Cheapest DDR5-5200 : $199.99
• Cheapest DDR5-5600 : $219.44
• Cheapest DDR5-6000: $279.99

Not all of the older systems even support DDR4-3600, but, if you're going to demand that they test the older systems at DDR4-3600 CAS 14, the LOWEST CAS available for 3600, then you'd probably better hold DDR5-6000 to that same standard, so, CAS30

• Cheapest DDR4-3600 CAS 14: $227.82
• Cheapest DDR5-6000 CAS 30: $359.99

Then, in the 12900k and 12600k review, there's also this note:


So . . exactly what are you demanding of them for testing now?

DDR5-6000 C30 would destroy DDR4-3600 C14. DDR5-6000 C36 is much cheaper and in games, about ties the higher end DDR4-3600 C14. In most everything else (productivity), it destroys DDR4-3600 C14.

Also if you take one more step on that DDR4-3600 C14, you'll find it ships from overseas in '4 to 6 weeks'. The lowest price appears to be $239 from PCPartPicker, coming from Newegg.

Regardless of the cost, my point remains, their chart states that DDR5 is 8% slower than DDR4 setups. That is purely because they used garbage DDR5-4800.

People erroneously get the impression that "DDR4 is better for games" - a true statement when these first came out and one could only get 4800 or if lucky 5200, at a massive (like $800) price premium.

That simply is not the case now. Reasonably priced DDR5 is now superior to DDR4 for almost all uses.

People like you are perpetuating this obsolete thought process.

 

[shady28]

To illustrate, this is what happens when you use DDR5-6000 C36 Trident Z5, $299.99 RAM, on a 12900KS vs a 3800X3D.

The 3800X3D is not the fastest gaming CPU, it never has been.

 

[King_V]

Your "proof" shows, what, exactly? What speed of DDR4 was that 5800X3D using, exactly? And why focus on getting that hard-to-find CL14 3600 DDR4 when you could just get CL18 4000MHz for much cheaper? First search turned up a GSkill 2x16 kit for $129.99.

Most Vermeer chips can hit 4000MHz RAM support in 1:1:1 mode.


So, you're just trying to say that Intel is being treated unfairly, because people aren't willing to pay a lot more (33% to 140% more) for exactly HOW much performance gain? To gain a whopping TWO TENTHS of a percent over the 5800X3D, is that it?


You haven't illustrated anything, other than your grudge about the 5800X3D.

What you NEED to do is prove that there's a gain from running the SAME Intel CPU with DDR5 vs DDR4 that justifies the added cost of the DDR5.

Show that, with information that shows exactly what RAM specs were being used for both, then you might have a point. Show whether these speeds were achieved out of the box, or required tweaking/manual intervention. Otherwise, your gripe that "The 3800X3D is not the fastest gaming CPU, it never has been" (you mis-named the 5800X3D as the 3800X3D, twice) is showing that you believe people should be willing and eager to spend a lot more money to help "prove" that Intel is faster.

 

[shady28]

So, you're just trying to say that Intel is being treated unfairly, because people aren't willing to pay a lot more (33% to 140% more) for exactly HOW much performance gain? To gain a whopping TWO TENTHS of a percent over the 5800X3D, is that it?

No, I think what I said was very clear to most moderately intelligent and unbias folks.

The 5800X3D is not the "Fastest gaming CPU". It never was.

Perhaps what you mean to say, is that the 5800X3D is the fastest gaming CPU on the planet if your memory budget is $200???

You haven't illustrated anything, other than your grudge about the 5800X3D.

You appear to have only illustrated that you have to put artificial constraints around this fastest gaming cpu claim, ones not mentioned in most if not all of these articles, in order to give it false support.

As I demonstrated, it is not the fastest gaming CPU, it never was and it never will be. An Alder Lake rig with the right kit - and not even that much more expensive - is obviously faster.

What you NEED to do is prove that there's a gain from running the SAME Intel CPU with DDR5 vs DDR4 that justifies the added cost of the DDR5.

You mean like this? This type of information is all over the internet.

Show that, with information that shows exactly what RAM specs were being used for both, then you might have a point. Show whether these speeds were achieved out of the box, or required tweaking/manual intervention. Otherwise, your gripe that "The 3800X3D is not the fastest gaming CPU, it never has been" (you mis-named the 5800X3D as the 3800X3D, twice) is showing that you believe people should be willing and eager to spend a lot more money to help "prove" that Intel is faster.

I don't believe anything of the sort. What you do with your money is your business.

If you wish to remain willfully ignorant, have at it, it's the AMD way!

 

[King_V]

Yes, EXACTLY like that. Or, you could've posted a link to your reference. The whole point was the alleged performance benefit you were claiming of DDR5 for Intel. YOU are the one who brought up the 5800X3D.

I suppose those are the only games that show an advantage that you are trying to show, therefore you picked those.

So, you snag two charts for 2 games from Guru3d, which only used DDR4 3200, and even then, one of those two shows lowly 3200MHz, cheaper even than 3600Mhz, almost keeping up with DDR5. And you snagged two charts for two other games from a different site . . . specifically to "prove" your point? Is that the ENTIRE suite of games those sites tested, or did you cherry pick them to show the results you wanted?

You understand that Tom's uses more games than just two, right? Or would you like to tell them what to use for testing? Is that it? Would you like to personally be in charge of choosing what games they test on? How much to spend on RAM? How much manual tweaking/overclocking to do?

Please, tell us EXACTLY what Tom's is doing wrong, and how, EXACTLY you want it resolved!

 

[cryoburner]

Perhaps they spent big money on a 12900KS DDR5 system for gaming, and are trying to justify their purchase with charts showing margin-or-error performance differences.

Really, it's questionable whether any differences shown there would be worth paying for. In that Horizon chart, there's no "perceptible" performance difference between a $100 32GB kit of DDR4 and a $400 kit of DDR5-6200, so I'm not sure what point is being made there. And the Cyberpunk chart is based on running the game at 1080p medium settings, apparently on a 6900XT, something no one is likely to ever do, so it's more of a synthetic result than a real-world one. Sure, there may be some examples of games where DDR5 offers a performance edge, but there are also plenty of examples were DDR4 offers equal or better performance at a far lower price, due to its better latency.

And of course, the 12900KS is a $750 CPU, being compared against a $450 5800X3D, for what is essentially very similar performance in games, on average. And on that note, either one is arguably a waste, when other CPUs offer practically the same gaming performance for under $300. And that's at 1080p on an enthusiast-level graphics card. At higher resolutions, or with a lesser card, those small performance differences tend to evaporate.

 

[shady28]

Didn't think I needed to repost this but, with accusations of cherry picking from the AMD fan base I guess I do.

Also as I stated before, putting artificial restrictions on 'fastest gaming processor' is disingenuous. It may well be the 'fastest gaming processor for $450' or 'fastest gaming processor for DDR4 based systems' or something equally arbitrary.

However it is not the 'fastest gaming processor', period.

With DDR5-6200 C36 :

 

[King_V]

Didn't think I needed to repost this but, with accusations of cherry picking from the AMD fan base I guess I do.

Also as I stated before, putting artificial restrictions on 'fastest gaming processor' is disingenuous. It may well be the 'fastest gaming processor for $450' or 'fastest gaming processor for DDR4 based systems' or something equally arbitrary.

However it is not the 'fastest gaming processor', period.

With DDR5-6200 C36 :
(meaningless charts snipped)

Once again, not providing a link to the tests. Why is that?

Also, irrelevant. Nobody cares about your gripe with the declaration of the 5800X3D being the fastest gaming processor. You dragged that CPU into this, when the original complaint was that the Alder Lake chips should be re-tested now with the fastest possible DDR5 RAM available. These don't show DDR4 vs DDR5 for the same Intel CPU.

In case you need reminding, it was this post of yours, and then, this next post of yours, the latter of which included the ridiculous claim of, and I quote (emphasis changed by me):

That simply is not the case now. Reasonably priced DDR5 is now superior to DDR4 for almost all uses.

Yeah, reasonably priced? Not true. For almost all uses? Also not true.

Further, you wanted the tests redone with the fastest DDR5 available, compared to mainstream 3600MHz DDR4 RAM. Or slower, for the older CPUs. So, how about picking some kind of consistent standard? Say, mainstream/budget DDR4 vs mainstream/budget DDR5. Or, if you insist on the "OMG money is no object get the fastest DDR5" then compare it with the fastest, money-is-no-object level of DDR4, on the SAME Intel processor.

Holding both sides to the same standard, and by that, I mean, not AMD vs Intel because that was not the original complaint you made, but the same standard of DDR4 vs DDR5, provide solid data that compares their performance difference on an Alder Lake CPU and shows a meaningful gain. Gain that is worth the extra money.

 

[SidorovichJr]

allright, but can you do now the same CPU rank but in 4K? I play only in 4K and I want to see point proven that, once you go 4K in gaming you don't need to worry about cpu <Mod Edit> anymore cause they are more or less offering same performance ever since 9900k and up?

 

[M42]

Once again, not providing a link to the tests. Why is that?
--cut--
but the same standard of DDR4 vs DDR5, provide solid data that compares their performance difference on an Alder Lake CPU and shows a meaningful gain. Gain that is worth the extra money.

This site compares the performance of DDR5 vs DDR4 using a 12900K. It is clear that fast DDR5 RAM results in a significant performance advantage over DDR4 for Alder Lake.

DDR4 vs. DDR5 on Intel Core i9-12900K Alder Lake Review

The Intel Alder Lake platform has support for both DDR5 and DDR4 memory. We ran 38 application benchmarks and 10 games at multiple DDR4 configurations to learn what performance to expect when using DDR4 vs. DDR5 on 12th Gen, and whether there's a point at which DDR4 performance can beat the much...

www.techpowerup.com

 

[King_V]

Thanks for the link. Pretty amazing what happens with the whole story rather than cherry-picked test results.

Looks like the conclusion is absolutely NOT worth the price-premium of high end DDR5.

There is no doubt that choosing DDR5 gives you a faster overall experience than DDR4, but sticking to DDR4 isn't that big a performance loss, and more importantly, DDR4 does not significantly change the overall performance outlook of the Core i9-12900K against its competitors from the previous generation or the AMD camp. (emphasis mine)
...
Averaged across all CPU tests, the unlikely hero of this review has to be the humble DDR4-3200, which ended up just 3.7% slower than the DDR5-6000 reference.