Question How in the world has AMD got the Ryzen 7600X and 7700X priced same when they are inferior even in P cores only compared to 13600K and 13700K

[Wolverine2349]

I mean the Ryzen 7700X is an 8 core CPU and Ryzen 7600X is a 6 core CPU. And the 7700X is $399 and 7600X is $299.

Intel has the Core i7 13700K priced at $399 and Core i5 13600K priced at $399. And those CPUs have better P cores being 8 and 6 core counterparts with slightly better IPC than Zen 4 and can clock as high or higher with similar power usage. And for those who do not like e-cores (I am one of them, but I love Intel P cores) can disable them and you get better 6 and 8 core CPUs form Intel Raptor Lake than AMD Ryzen. And for those who want e-cores you get then as well for the same price and better P cores of equal core counts.

SO what is AMD thinking and they still have not budged on the prices of the 7600X and 7700X. They are pricing the like their 6 and 8 Zen 4 cores are better than Intel's Raptor Cove cores of equal count even though they are not any better and in fact not as good?? Or is that debatable??

The Ryzen 7900X and 7950X prices make more sense as then you get more than 8 strong cores and AMD has those by the balls who want more than 8 cores and do nit want to go hybrid route. SO yeah 7900X and 7950X prices make sense.

But 7600X and 7700X are almost a ripoff unless you just have not have AMD as they do nothing better than 13600K and 13700K for exact same price and have slightly weaker P cores and no additional e-cores for those that like the e-core options (And for those that do not it is easy peasy to disable and you get the better 6 and 8 core chips for the same price)

Its puzzling to me AMD is behaving as if they are still superior in all ways like they were with Ryzen 5000 from November 2020 to November 2021 when Intel was of no competition on core count nor per core IPC performance which was only for 1 year. I mean AMD is still much smaller and was underdog for years and hard to believe they think they can act they are premium brand in the 6 and 8 core CPU segment when the 7600X and 7700X are worse than Intel counterparts even with the e-cores off.

Your thoughts

 

[coercitiv]

As additional engineers are added to a project I think it's safe to assume the prospect of significant design flaws is reduced. So I think the point is that is is very unlikely that a feature as important as the one being discussed ended up being a "screw up."

The number of engineers has little to do with the quality of the final product except for situations when the team is severely understaffed. The foundations of the project, the timing, the quality of the engineering and management corps are far more important when aiming for excellency. A competent collective knows to set the scope of the project anyway, thus aims for adequate scale relative to no. of engineers. Intel is a company in crisis for many years now, with MANY engineering failures. Lakefield and Sapphire Rapids are notable and relatively recent examples, but they're not the only ones.

 

[scannall]

I have to say, these kinds of threads get tedious. CPU's from either company are just fine. If an Intel product is best for your situation, then buy it, it will be just fine. AMD fits my needs better currently, so that's what I bought.

 

[Hulk]

One example in the CPU space that directly contradicts this reasoning is the Bulldozer design with CMT. AMD was a very successful company at the time with top class design teams, yet it happened. The wrong path taken with disastrous consequences. There are more.

Yes and Intel with Netburst and Itanium were also wrong turns. But these are anecdotal. There are literally thousands of correct decisions if not more for every wrong turn. Major architectural mistakes are rare, especially when you consider the millions/billions of transistors in modern CPU designs.

Back to the topic. Carfax83's argument point that it would be very unlikely for them to add said feature without solid reasoning is a worthy of being mentioned and your retort that it was "one of the dumbest things written and using it in an argument to support your view is transparently dishonest" is wildly hyperbolic. You might not agree with is argument of "authority" for the Intel engineers but it is surely worthy of mention in a debate.

 

[DAPUNISHER]

Back to the topic. Carfax83's argument point that it would be very unlikely for them to add said feature without solid reasoning is a worthy of being mentioned and your retort that it was "one of the dumbest things written and using it in an argument to support your view is transparently dishonest" is wildly hyperbolic. You might not agree with is argument of "authority" for the Intel engineers but it is surely worthy of mention in a debate.

You mean back to the off topic. The topic was about the audacity of AMD to price their 6 and 8 core CPUs the way they did. Hilariously, right at the very moment there was a price cut. Since that fake news was debunked immediately, everything else has been the usual versus nonsense you Ford v. Chevy folks love so much. Enjoy.

 

[maddie]

Yes and Intel with Netburst and Itanium were also wrong turns. But these are anecdotal. There are literally thousands of correct decisions if not more for every wrong turn. Major architectural mistakes are rare, especially when you consider the millions/billions of transistors in modern CPU designs.

Back to the topic. Carfax83's argument point that it would be very unlikely for them to add said feature without solid reasoning is a worthy of being mentioned and your retort that it was "one of the dumbest things written and using it in an argument to support your view is transparently dishonest" is wildly hyperbolic. You might not agree with is argument of "authority" for the Intel engineers but it is surely worthy of mention in a debate.

I said nothing about Intel engineers specifically. In fact, my counter in another post was AMD with Bulldozer. My claim is that trusting the experts blindly is a dumb argument, especially for intelligent, critical thinkers. I can understand if an individual can't follow theory due to education or whatever, but experts are wrong all the time. Do you actually think otherwise? They are the ones that do the huge mistakes/disasters. Solid reasoning with the wrong theory/facts/assumptions, etc, still leads to failure.

I also don't understand your claiming Netburst and Itanium were anecdotal? Maybe it's a language issue.

 

[Hulk]

You mean back to the off topic. The topic was about the audacity of AMD to price their 6 and 8 core CPUs the way they did. Hilariously, right at the very moment there was a price cut. Since that fake news was debunked immediately, everything else has been the usual versus nonsense you Ford v. Chevy folks love so much. Enjoy.

Back to the topic to which Maddie responded!
Have a nice day!

 

[Schmide]

For FFT if the dataset is arranged in favorable way, you can saturate the AVX execution units to the fullest. You can achieve this by interleaving multiple streams of FFT into single processing section. I have done this in the past.

I agree. I too have packed multiple FFTs in such a way. For a 2d data set half the operations (columns) work out this way and you can transpose the data for the remaining rows. For small datasets like this it is the way to go. Though once the data expands out of the L1 cache, the sparse access can thrash the cache.

I made a mistake in my initial calculations. The memory footprint for a 32x32 2d FFT is 8k not 4k.

 

[DAPUNISHER]

Back to the topic to which Maddie responded!
Have a nice day!

You have a great Thanksgiving. And don't get angry, those exclamation marks are serious business.

 

[Hulk]

You have a great Thanksgiving. And don't get angry, those exclamation marks are serious business.

I'm not angry. You wouldn't like me when I'm angry.


(You stepped into that)

 

[Carfax83]

With this SVT-AV1 encoder 7950X is barely faster than a 5950X and both are well below a 12900K and you keep saying that it s due to CPU real perf...

I didn't run that test so I have no explanation for that, but in terms of base 256 bit loads and stores and execution units, Zen 4 and Zen 3 are exactly the same. Zen 4 will have faster execution due to higher clock speeds and more bandwidth however.

But the test itself was very short running only a few seconds so I wouldn't consider it very legitimate.

So you basically use an Intel designed encoder that show about no improvement from 5950X to 7950X to claim better perf for RL, have you any other such half legged tests..?.

You're cherry picking. That benchmark or whatever that TPU ran lasts only a few seconds, not long enough to come to any conclusions. The one that I showed originally from Computerbase.de was a much longer test and the 7950x was 22% faster than the 5950x.

Besides 7950X has 60% better SIMD throughput than the 12900K according to Sisoft s Sandra, so much for us not knowing about CPUs uarch.

SiSoft Sandra is a synthetic benchmark and not analogous to a real application. Zen 4 can be faster in SIMD workloads if the application has good core scaling as Zen 4 has more cores.

And we see that in the Phoronix benches, because the 7950x will often tie or slightly lead the 13900K in AV1 encoding despite having less raw SIMD capabilities.

 

[Carfax83]

So engineers can't screw up? Speaking as one, that is one of the dumbest things written and using it in an argument to support your view is transparently dishonest.

Look around you, any mistakes/screwups apparent?

Not saying engineers can't screw up obviously, but this is about displaying intention. Golden Cove was already a high bandwidth, high throughput architecture and Raptor Cove doubles down on it.

Raptor Cove nearly doubled L1 and L3 cache bandwidth as far as I can see, and having such enormous bandwidth is critical to feeding the execution units for these types of workloads as AVX2 is bandwidth reliant.

That's why RPL is notably faster than ADL in AV1 encoding despite having the same exact load/store capabilities and execution units.

 

[Carfax83]

I do program a lot in x64 Assembly as hobby including using AVX256. Load/store bandwidth hardly the limiting factor as Zen can do 2 loads and 1 store per cycle. It only matters for basic streaming type workload (like take 2 values, add together and save it). Zen4 also is 6 issue from uop-cache and it has a big uop-cache. Most of the critical loop will run from uop-cache. Therefore, having 6 wide decode is not a big advantage.

Well if you're a programmer then your understanding is well beyond mine for discussing this subject. I only came to my conclusion after researching why RPL was so fast in certain workloads like encoding, code compiling, compression etcetera.

And comparing Zen 4 to RPL, the latter has way more internal bandwidth, is wider, has a larger reorder buffer and can do 3x 256 bit loads and 2x 256 bit stores per cycle in tandem. Zen 4 can do 3x 256 bit loads OR 2x 256 bit stores per cycle OR 2x 256 bit loads and 1x 256 bit stores per cycle in tandem.

Golden Cove also has 2 cycle latency on FADDs compared to Zen 4 which has 3 cycles. I don't know if that makes a difference, but I am just comparing attributes.

This is taken from the Chips and Cheese Golden Cove deep dive article which helped me make sense of this and come to this conclusion:

Looking at a single thread bandwidth on Golden Cove, we see three 256-bit vector loads per cycle from L1D. That’s impressive, and explains how Intel can compete against AMD chips sporting more “P-Cores”. For comparison, AMD’s Zen 3 can do three loads per cycle, but only two of those can be vector loads. Golden Cove also has a wider 64 byte per cycle path to L2, while AMD uses a narrower 32 byte per cycle path. If you have a vectorized program that fits in core-private caches, Golden Cove gives you massive bandwidth.

Popping the Hood on Golden Cove – Chips and Cheese

 

[Carfax83]

I would say the extra e-cores do a lot to help intel's performance, much more than any cache/memory bandwidth.

At the core of any modern compression algorithm is a FFT. ( AV1 is a 32x32 2d FFT)

The memory used in this operation is 8k which easily fits in the L1 cache. (probably double as in place FFT implementations do not lend well to vectorized FFTs) In terms of unoptimized operations, a FFT of 32 has O(N log2 N) = 32 * 5 = 160 complexity. Expanding out to the 2d operation where each row and column is operated on, giving 160 * 64 = 10240 = ~10k complexity. Depending on the compression level this can be reduced by a least a quarter. (no need to finalize the edge bins). SIMD operations further reduce the number of operations but require the data be transposed adding back in some complexity. AVX and SSE help a fair amount, but probably only 2-3x performance gains over non-vectorized implementations.

This is a very computational bottleneck where load stores represent a tiny fraction of the total operations. Best case #loads (64-96) / #operations (10k) = 0.0064 - 0.0096.

If the E cores help out a lot, then shouldn't the 13900K be well ahead of the 13700K, and how is the 13700K which is technically a 16 core CPU but with 8 efficiency cores so close to the 7950x which has 16 big cores in the Computerbase.de benchmark?

The E cores are nowhere near as powerful as a Zen 4 core and by all appearances, don't really help out that much.

 

[Schmide]

If the E cores help out a lot, then shouldn't the 13900K be well ahead of the 13700K, and how is the 13700K which is technically a 16 core CPU but with 8 efficiency cores so close to the 7950x which has 16 big cores in the Computerbase.de benchmark?

The E cores are nowhere near as powerful as a Zen 4 core and by all appearances, don't really help out that much.

Ehh. If you look at the graphs everything scales kind of linearly. Especially in the 4k graph. You get 2 and change for 4 AMD cores or 8 intel e cores where L3 cache is similar. A little less than 4 for 4 AMD cores + the extra L3 cache. (7700x-7900x) Intel 13600k-13700k about same little less than 4 with relative cache and core difference.

The point is they do carry weight. Maybe P cores carry more but little differences like load store ratio are probably less than any of the other factors.

The more I look at the data the more the relative segmentation by both AMD and intel becomes apparent.

 

[DAPUNISHER]

TLDW - Price drop made Linus change his pick from Raptor to Ryzen.

GN continues to prove that they are out of touch with the budget market. I will mention potential bias, but I personally think being out of touch is the reality. This is the holiday shopping CPU buyer guide I will dissect. They must use PCPARTSPICKER or something for pricing. Because some of the recommendations and conclusions are poor.

Let's start with the fact you can still get the 3D for $329.99. Making it the choice he wanted it to be, at the price he wanted. Therefore should have been the de facto recommendation. Also, no inclusion of the actual prices after promo from Newegg at the time of his intro and outro even though they are right there on screen is sus. You can't wish away the real price for Holiday shoppers, it fosters accusations of bias.

He takes shots at AM4 being abandoned. Yet somehow does not do the same for Alder or Raptor. More fuel for the bias bandwagon.

Now let's really get that bias bandwagon moving. Choosing the 12100f and 10100 series for the best budget CPUs is what happens when you are either biased, or out of touch with the current market. The Ryzen 5500 is clearly the best sub $100 choice given it's extra threads, overclocking capability, and superior upgrade path. It absolutely bodies the 10100 series. The Ryzen 5600 is clearly the best budget CPU and most agree with me, as the sales stats reflect. It is currently #1 on Newegg, #2 on Amazon. At this tier, Ryzen again has more threads and overclocking in its favor.

BTW, today is the first time since launch day, that I have seen Zen 4 in the top 10 on either Amazon or Newegg. The value of the 7950X at current pricing is winning over holiday builders.

One more comment on the GN vid. He closes with Intel putting pressure on AMD. I agree, if we look at newest gen v. newest gen. However, Amazon is the biggest U.S. vendor, and AMD occupies all 10 top sellers spots right now. Influencers are trying to control the narrative, but it only carries so much weight. DIY builders on a tight budget know AM4 is where it's at.

 

[Carfax83]

Ehh. If you look at the graphs everything scales kind of linearly. Especially in the 4k graph. You get 2 and change for 4 AMD cores or 8 intel e cores where L3 cache is similar. A little less than 4 for 4 AMD cores + the extra L3 cache. (7700x-7900x) Intel 13600k-13700k about same little less than 4 with relative cache and core difference.

The point is they do carry weight. Maybe P cores carry more but little differences like load store ratio are probably less than any of the other factors.

The more I look at the data the more the relative segmentation by both AMD and intel becomes apparent.

But something else has to be going on, because RPL is also faster than ADL when comparing the 13700K to the 12900KS, with the same amount of cores there's still a double digit percentage difference and the clock speed difference isn't enough to explain it. That's what got me looking at the cache bandwidth differences, and as I said before, RPL has double the L1 and nearly double the L3 cache bandwidth for ADL.

That's what got me thinking that it must be due to RPL's ability to sustain the high performance in vectorized workloads due to superior cache bandwidth.

Case in point. Now of course there are version differences between the software, but no way on Earth that could account for a near doubling of bandwidth for L1 in particular.

13700K:

And 12900K, 11900K and 5950x:

 

[Carfax83]

And another one showing RPL way out in front of ADL:

 

[lobz]

The jump in performance from Bulldozer to Zen 1 was between 50-60% IPC gain. That was how bad AMD stuff was before Ryzen. Big jumps were Zen 2 and Zen 3. I am not knocking Zen 4, but the jump from Zen 3 to Zen 4 is not as significant as people would have hoped it would be. When 3D v-cache arrives for Zen 4. People will be happy with Zen 4.

but it (Zen 4) finally brought that real freq jump that everybody was hoping for about the 2 other launches, we shouldn't exclude that from the equasion

 

[maddie]

TLDW - Price drop made Linus change his pick from Raptor to Ryzen.

GN continues to prove that they are out of touch with the budget market. I will mention potential bias, but I personally think being out of touch is the reality. This is the holiday shopping CPU buyer guide I will dissect. They must use PCPARTSPICKER or something for pricing. Because some of the recommendations and conclusions are poor.

Let's start with the fact you can still get the 3D for $329.99. Making it the choice he wanted it to be, at the price he wanted. Therefore should have been the de facto recommendation. Also, no inclusion of the actual prices after promo from Newegg at the time of his intro and outro even though they are right there on screen is sus. You can't wish away the real price for Holiday shoppers, it fosters accusations of bias.

He takes shots at AM4 being abandoned. Yet somehow does not do the same for Alder or Raptor. More fuel for the bias bandwagon.

Now let's really get that bias bandwagon moving. Choosing the 12100f and 10100 series for the best budget CPUs is what happens when you are either biased, or out of touch with the current market. The Ryzen 5500 is clearly the best sub $100 choice given it's extra threads, overclocking capability, and superior upgrade path. It absolutely bodies the 10100 series. The Ryzen 5600 is clearly the best budget CPU and most agree with me, as the sales stats reflect. It is currently #1 on Newegg, #2 on Amazon. At this tier, Ryzen again has more threads and overclocking in its favor.

BTW, today is the first time since launch day, that I have seen Zen 4 in the top 10 on either Amazon or Newegg. The value of the 7950X at current pricing is winning over holiday builders.

One more comment on the GN vid. He closes with Intel putting pressure on AMD. I agree, if we look at newest gen v. newest gen. However, Amazon is the biggest U.S. vendor, and AMD occupies all 10 top sellers spots right now. Influencers are trying to control the narrative, but it only carries so much weight. DIY builders on a tight budget know AM4 is where it's at.

Hopefully the accounting & marketing dept staff are being trained to the new reality.

 

[Schmide]

But something else has to be going on, because RPL is also faster than ADL when comparing the 13700K to the 12900KS, with the same amount of cores there's still a double digit percentage difference and the clock speed difference isn't enough to explain it. That's what got me looking at the cache bandwidth differences, and as I said before, RPL has double the L1 and nearly double the L3 cache bandwidth for ADL.

That's what got me thinking that it must be due to RPL's ability to sustain the high performance in vectorized workloads due to superior cache bandwidth.

Case in point. Now of course there are version differences between the software, but no way on Earth that could account for a near doubling of bandwidth for L1 in particular.

13700K:

And 12900K, 11900K and 5950x:

Bandwidth, size, policy, a lot changed for both intel and AMD. They both basically doubled the L2 and intel did so on both p(1.25->2) and e. Intel also made changes to their ring bus and L2/L3 retirement policy. (I think they claimed 16% for that inclusive/exclusive trick).

In the handbrake image you posted, you see similar gains for both AMD and intel both generationally and core and cache relative. (15-20fps).

I'm sure the extra cache bandwidth is responsible for some of these gains on both the p and e cores. I don't know if they enhanced the e cores bandwidth. (maybe size is everything)

 

[Wolverine2349]

If the E cores help out a lot, then shouldn't the 13900K be well ahead of the 13700K, and how is the 13700K which is technically a 16 core CPU but with 8 efficiency cores so close to the 7950x which has 16 big cores in the Computerbase.de benchmark?

The E cores are nowhere near as powerful as a Zen 4 core and by all appearances, don't really help out that much.

I agree. The e-cores do not help that much. They are there and help some in multi threaded workloads. But the P cores are just that good they add additional mediocre e-cores to make the Alder Lake and Raptor Lake CPUs compete with Intel 12 and 16 core Zen 3 and Zen 4 parts respectively in workloads that scale to infinite CPU cores.

The P cores on other hand just straight up defeat AMD Zen 3 cores easily and even beat AMD Zen 4 cores by a little at same clock speed and core count though gap with Zen 4 is very close at same clock speed and much more dependent on type of workload.

 

[Hulk]

I agree. The e-cores do not help that much. They are there and help some in multi threaded workloads. But the P cores are just that good they add additional mediocre e-cores to make the Alder Lake and Raptor Lake CPUs compete with Intel 12 and 16 core Zen 3 and Zen 4 parts respectively in workloads that scale to infinite CPU cores.

The P cores on other hand just straight up defeat AMD Zen 3 cores easily and even beat AMD Zen 4 cores by a little at same clock speed and core count though gap with Zen 4 is very close at same clock speed and much more dependent on type of workload.

With the 13900K running 5.5GHz P cores and 4.3GHz E cores and scoring 40k in Cinebench R23 MT...

The P cores contribute to 56% of that score and the E's 44%.

I think based on those numbers in well-optimized MT workloads the E's contribute quite a lot.

In fact those 16 Skylake level E's running 4.3GHz contribute over 18,000 points to that score. A not insignificant amount of compute they are generating.

In other terms, approximately 13.1 P's could fit on the die if there were no E's and the resulting score would be 33,450 IF the thermals would allow all of those E's to run at 5.5GHz. By my calculations it would draw about 350W.

The E's are there for area efficiency for MT workloads. The fact that the 13900K is competitive with the 7950X is proof that is works.

I know what many of you are thinking. "Yeah, but if they just put 16 P's on the 13900K it would be better." And you are right it would be a more performant part, scoring over 44,000 in Cinebench MT. But it would draw about 425W and more importantly would have a die so large that is wouldn't be cost effective for Intel. Given the constraints Intel had (mainly being behind a node in transistor density and no chiplets as yet) they came up with the best part possible. By "best" I mean in terms of pricing, performance, and manufacturing (yields, etc.)

 

[Carfax83]

With the 13900K running 5.5GHz P cores and 4.3GHz E cores and scoring 40k in Cinebench R23 MT...

The P cores contribute to 56% of that score and the E's 44%.

I think based on those number in well-optimized MT workloads the E's contribute quite a lot.

I agree, but rendering is the most scalable workload for multiple cores. The encoding example that I used isn't nearly as scalable, as seen by the smaller margin between the 13700K and the 13900K in AV1 encoding compared to rendering.

Raw compute performance is a heavier influence in AV1 encoding I believe than in rendering because a lot of encoding relies on single threaded performance as well as multi threaded performance due to the nature of how encoding works.

 

[Wolverine2349]

With the 13900K running 5.5GHz P cores and 4.3GHz E cores and scoring 40k in Cinebench R23 MT...

The P cores contribute to 56% of that score and the E's 44%.

I think based on those numbers in well-optimized MT workloads the E's contribute quite a lot.

In fact those 16 Skylake level E's running 4.3GHz contribute over 18,000 points to that score. A not insignificant amount of compute they are generating.

In other terms, approximately 13.1 P's could fit on the die if there were no E's and the resulting score would be 33,450 IF the thermals would allow all of those E's to run at 5.5GHz. By my calculations it would draw about 350W.

The E's are there for area efficiency for MT workloads. The fact that the 13900K is competitive with the 7950X is proof that is works.

I know what many of you are thinking. "Yeah, but if they just put 16 P's on the 13900K it would be better." And you are right it would be a more performant part, scoring over 44,000 in Cinebench MT. But it would draw about 425W and more importantly would have a die so large that is wouldn't be cost effective for Intel. Given the constraints Intel had (mainly being behind a node in transistor density and no chiplets as yet) they came up with the best part possible. By "best" I mean in terms of pricing, performance, and manufacturing (yields, etc.)

Are the4 e-cores really Skylake like performance. Needs 16 of them to be competitive with 7950X combined with 8 super P cores.

Though yes they do provide raw compute power to really help the multi core load in apps that scale to infinite CPU cores.

Though e-cores like Skylake performance, isn't that somewhat true, but with severely crippled latency which makes them much worse especially if they were to be used by themselves without any P core assistance meaning Skylake derivatives like 9700K 98 cores at same clock would slaughter the Gracemont cores because despite IPC being similar, 9700K has faster better latency and more instruction sets as well.

 

[Thunder 57]

Are the4 e-cores really Skylake like performance. Needs 16 of them to be competitive with 7950X combined with 8 super P cores.

Amdhal's law. Plus you think those P cores are better than they really are. Zen 4 competes well against them.

Though e-cores like Skylake performance, isn't that somewhat true, but with severely crippled latency which makes them much worse especially if they were to be used by themselves without any P core assistance meaning Skylake derivatives like 9700K 98 cores at same clock would slaughter the Gracemont cores because despite IPC being similar, 9700K has faster better latency and more instruction sets as well.

What instruction sets? RPL has no AVX-512. The 9700k clocked a bit higher. That's probably the difference.