Were we wrong about Zen 5?

[Tantor]

Wow, this will make older AM4 and AM5 owners very happy. Gamers will just have to wait for the 9700x3d.

I'm still unclear about the 7700x vs 7700 power efficiency. I always assumed they had the same native efficiency. The difference was just due to the 7700x higher TDP.

 

[HardReset]

Steve tests some of the games AMD cherry picks to serve it's needs, they don't find any such gains. Next you'll claim they didn't lie about 5800/5900XT gaming performance. I've been defending AMD but wasn't aware of their performance claims which are in their official slides.

Steve is using DDR5 CL30 memory, AMD's own sweetspot memory. I've been critical of his headlines, especially seeing we don't have the full release of cpu's yet, but this is on AMD. Wrong names and wrong prices and they supply the reviewers with kits, so it's up to them to ensure they use the best settings.

Same games with same settings too? If something is different, then results may be different. Accusing AMD about lying needs proof. In this case, using exactly same equipment, exactly same software , exactly same settings etc. Unless one can do it, accusing AMD about lying is pointless. That is not opinion but fact. Like it or not.

 

[emmzo]

Wow, this has turned into an AMD fan forum very quickly! To summarize: why nobody wants to use tech that AMD believes in? PBO is great, but so what if it is not, this isn't a gaming processor. They didn't lie, unless you can reproduce every detail like the biased software of their benchmarks. The gains are more than enough, but nobody wants to pay what AMD is asking....

 

[HardReset]

Wow, this has turned into an AMD fan forum very quickly! To summarize: why nobody wants to use tech that AMD believes in? PBO is great, but so what if it is not, this isn't a gaming processor. They didn't lie, unless you can reproduce every detail like the biased software of their benchmarks. The gains are more than enough, but nobody wants to pay what AMD is asking....

Or perhaps forum writers actually know more about this CPU than Techspot staff? AMD said long time ago Zen5 should have much better AVX512 performance than Zen4. Now it's confirmed that is true. Techspot has written two articles where it failed to even mention AVX512! AMD "fanboys" keep facts together, like it or not.

 

[zx8086]

It's rather amusing to read about "Misleading comparisons and misconceptions" when basically the whole mainstream tech journalism testing methods are producing just that. I'm sorry that I have to flog this dead horse yet again, but using only the extreme high end of CPU/GPU for testing is what this sub header should refer to, not some hair splitting over PBO (ok, that's important too, or could be, if the main point of contention was fixed to start with).

I know that many people struggle with grasping this, so please consider this simple example: let's say Tom's Hardware et al found that 9600X is on average 21% faster in 1440p than some other comparable CPU (it's just an example so don't get your dander up if the figures are wrong). Well, cool, for somebody like me who's looking to upgrade from 5600X, that's a significant uplift. I go out and buy the whole spanking new AM5 rig, go home, stick my 4070 Super in and.... see 5% (or less) gains?

This is what using only 4090, an extreme outlier, does for testing. It could be similar for GPU, using only top CPU in test bed. It's misinformation, pure and simple. The fps do not necessarily scale across the hardware spectrum and not including a lower range CPU or GPU will very often produce misleading results.

In this case TS kinda gets away with it because it's plain to see that if 4090 only gets minimal gains in 1080p then clearly it won't improve in higher res. But I did not see it mentioned anywhere in the article and suppose that it's just the chosen new style for TS benchmarking. It's rather worrying for the future articles and makes quotes like these:

"we want to point out the obvious: TechSpot and Hardware Unboxed are aimed at enthusiasts and PC gamers. You won't find us running a single 40-application benchmark here, but you will find countless big 40+ gaming benchmarks comparing various CPUs and GPUs."

....kinda silly. Where are these amazing reviews comparing various CPUs and GPUs in the same article? I recall one (1) from a few months back, it was indeed very good and gave me hope, which faded fast after that. And, I'm sorry, but it doesn't matter if you include 4 or 40 games in your test when your testing method is botched to start with.

Though, like I said, it's obviously not only TS but almost everyone. The couple of outlets like AnandTech or PC Mag which used 4080/3080 for this test also inexplicably omitted 1440p (the former, but they included 720p, wtf) or used some old irrelevant games (the latter).

 

[emmzo]

Or perhaps forum writers actually know more about this CPU than Techspot staff? AMD said long time ago Zen5 should have much better AVX512 performance than Zen4. Now it's confirmed that is true. Techspot has written two articles where it failed to even mention AVX512! AMD "fanboys" keep facts together, like it or not.

Technicals aside, the question is: would you buy it and would you advise people to buy it? Because whatever mind-blowing tech and excellent gains you claim they have it has to be worth of their asking price. Ok, I may know your answer: I don't need it atm, maybe down the road, who knows, maybe never, but everyone else should buy it at this price, it's a good price for what they are offering. Is that correct? Because if it isn't your point is void.

 

[Porkous]

The results are cherry picked on purpose.

Because internal CPU optimisation designs are small, the broad list of benchmarks had to be small. If there is no improvement in any benchmark how could it be on the CPU itself, inside it's blueprints?

Now the article has to prove those CPU inner changes are not doing anymore for performance...

 

[HardReset]

Technicals aside, the question is: would you buy it and would you advise people to buy it? Because whatever mind-blowing tech and excellent gains you claim they have it has to be worth of their asking price. Ok, I may know your answer: I don't need it atm, maybe down the road, who knows, maybe never, but everyone else should buy it at this price, it's a good price for what they are offering. Is that correct? Because if it isn't your point is void.

CPU technical improvement and "who should buy it" are totally different things. CPU may be "bad buy" for some people but for some it may be only viable option.

Also price is something that will definitely change when time goes on. CPU stays basically same (later ones are better binned probably as manufacturing process matures) but price will very probably come down. Zen5 may be bad buy right now but it could be excellent buy few months later. Pricing change , CPU don't (at least, a lot).

As I said elsewhere (this thread or other), Zen5 competes with Zen4 since both fit AM5 socket. It makes zero sense for AMD to make Zen5 both faster and cheaper, then nobody would buy Zen4. That is cost of having same socket for long time. Price you just have to accept. Because only solution for problem would be new socket for Zen5...

 

[emmzo]

CPU technical improvement and "who should buy it" are totally different things. CPU may be "bad buy" for some people but for some it may be only viable option.

Also price is something that will definitely change when time goes on. CPU stays basically same (later ones are better binned probably as manufacturing process matures) but price will very probably come down. Zen5 may be bad buy right now but it could be excellent buy few months later. Pricing change , CPU don't (at least, a lot).

As I said elsewhere (this thread or other), Zen5 competes with Zen4 since both fit AM5 socket. It makes zero sense for AMD to make Zen5 both faster and cheaper, then nobody would buy Zen4. That is cost of having same socket for long time. Price you just have to accept. Because only solution for problem would be new socket for Zen5...

I think you understood what I implied, but choose to ignore it. I was referring if it is a good buy from last gen. That is the question Techspot has asked. And if you say, well, at a certain time maybe, then it is not. You also argue, well, only the tech improvement matters, but being worth in real life depends on personal preference. Really? Also, your argument for not making Zen 5 faster and better is just rubbish. If it competes with Zen 4 with little to no advantage then it's pointless.

 

[HardReset]

I think you understood what I implied, but choose to ignore it. I was referring if it is a good buy from last gen. That is the question Techspot has asked. And if you say, well, at a certain time maybe, then it is not. You also argue, well, only the tech improvement matters, but being worth in real life depends on personal preference. Really? Also, your argument for not making Zen 5 faster and better is just rubbish. If it competes with Zen 4 with little to no advantage then it's pointless.

Tbh, I didn't understand fully. But OK, is Zen5 good buy vs Zen4? Right now, probably not. But again, prices will very probably go down and then situation is different. That's why basically no-one is interested about this "good buy or not" -thing because it depends on price, that again, will change.

As for Zen5 advantage vs Zen4, that is easy one. Fact is, with just x86 and/or SSE based improvements, very small double digit performance increase can be expected. If something like 50% is needed, then using new or "new" instruction set is a must. Now, AVX512 have been around for about a decade. It's not "new" anymore. Using AVX512 even triple digit performance increases are possible with Zen5 vs Zen4. Based on that, cannot understand this Zen5 is not faster" BS. Again, if you use old software, then accept that performance increase is also small. There are no free lunches. Like said already, Steve didn't even Mention AVX512 single time on these two articles!

 

[Che Cazzo]

Were we wrong about Zen 5?
Yes, we was wrong and we is still wrong.

Intel fanboism is oozing all over this PR stunt , I mean so-called "tech article".

 

[EdmondRC]

It's rather amusing to read about "Misleading comparisons and misconceptions" when basically the whole mainstream tech journalism testing methods are producing just that. I'm sorry that I have to flog this dead horse yet again, but using only the extreme high end of CPU/GPU for testing is what this sub header should refer to, not some hair splitting over PBO (ok, that's important too, or could be, if the main point of contention was fixed to start with).

I know that many people struggle with grasping this, so please consider this simple example: let's say Tom's Hardware et al found that 9600X is on average 21% faster in 1440p than some other comparable CPU (it's just an example so don't get your dander up if the figures are wrong). Well, cool, for somebody like me who's looking to upgrade from 5600X, that's a significant uplift. I go out and buy the whole spanking new AM5 rig, go home, stick my 4070 Super in and.... see 5% (or less) gains?

This is what using only 4090, an extreme outlier, does for testing. It could be similar for GPU, using only top CPU in test bed. It's misinformation, pure and simple. The fps do not necessarily scale across the hardware spectrum and not including a lower range CPU or GPU will very often produce misleading results.

In this case TS kinda gets away with it because it's plain to see that if 4090 only gets minimal gains in 1080p then clearly it won't improve in higher res. But I did not see it mentioned anywhere in the article and suppose that it's just the chosen new style for TS benchmarking. It's rather worrying for the future articles and makes quotes like these:

"we want to point out the obvious: TechSpot and Hardware Unboxed are aimed at enthusiasts and PC gamers. You won't find us running a single 40-application benchmark here, but you will find countless big 40+ gaming benchmarks comparing various CPUs and GPUs."

....kinda silly. Where are these amazing reviews comparing various CPUs and GPUs in the same article? I recall one (1) from a few months back, it was indeed very good and gave me hope, which faded fast after that. And, I'm sorry, but it doesn't matter if you include 4 or 40 games in your test when your testing method is botched to start with.

Though, like I said, it's obviously not only TS but almost everyone. The couple of outlets like AnandTech or PC Mag which used 4080/3080 for this test also inexplicably omitted 1440p (the former, but they included 720p, wtf) or used some old irrelevant games (the latter).

They are testing the processor not the graphics card. In order to fully give you the absolute best frames the processor can push at 1080p, they need to use the best GPU. If your GPU limited at 1080p on your current processor, then a new processor won't do you much good. It is your responsibility to do the research.

Example 1, you have a 4070 and looking for new CPU - You check benchmarks for your games that use a 4070 and 7800X3D, you find that your processor is delivering the same frames for 4070 in the games you care about - conclusion, you don't need a new processor

Example 2, you have a 5800X with a 4070, you check benchmarks for 4070 for your games where reviewer is using 4070 and 7800X3D. You find that the games you play could be 10% faster at 1080p, however you play at 1440p and the uplift at 1440p is only 2% - conclusion, you don't need a new processor

Example 3, you have a 5800X with a 4080, you check benchmarks for the 4080 that is using a 7800X3D, you see that even at 4K you are losing 5-20% performance in the games you're playing. Some games you are even getting a whole tier less performance out of your GPU. You want to see which processor is right for the 4080, so you then check CPU benchmarks to determine the best processor for you, you notice that the 9700X cost as much as the 7800X3D, but delivers less performance in games, but better productivity. You then have to make a choice, best gaming processor, or better productivity and decent gaming processor? --- You choose 7800X3D as your computer is 90% for gaming...

 

[Yppah]

Well, guess I'll wait to see if tom's hardware prints a retraction of their 21% FPS gains average at 1080P for the 9700X PBO over 7700X. https://www.tomshardware.com/pc-components/cpus/amd-ryzen-5-9600x-cpu-review/2

You having a giggle? Toms Hardware are about as renowned as LTT. As in both are a joke. It'd make as much sense to get your tech news from the Daily Mail.

 

[EdmondRC]

Well, guess I'll wait to see if tom's hardware prints a retraction of their 21% FPS gains average at 1080P for the 9700X PBO over 7700X. https://www.tomshardware.com/pc-components/cpus/amd-ryzen-5-9600x-cpu-review/2

For example TS shows a virtual tie between 9700X and 7700X in Watch Dogs at 1080P Ultra/4090:

https://imgur.com/CtyCbIc

TH shows a 19% FPS gain with same settings with PBO (which TS states shows no gains), also with 4090:

https://imgur.com/XDt2ieb

Could Techspot post your test system config for the 9600X and 9700X? This should be standard practice for benchmarking. A simple explanation could be the motherboard used isn't updated correctly for PBO support.

It seems like Tom's Hardware also set an overclock and CO Curve of -20 to get those results.

I noticed you gave kudos to TechPowerUp's review and discussed their findings. Don't suppose you caught this article on there: https://www.techpowerup.com/review/amd-ryzen-9-9700x-performance-smt-disabled/ ? There might be more left in the tank on the 9700x after all. Wouldn't be the first time the microcode was the problem.

This is interesting, at least moving toward an explanation.

 

[hwertz]

1) I'm not going to hate on Techspot for not reviewing AVX512 performance. They reviewed gaming performance. If games are not using AVX512, they aren't using AVX512. That tells me, if you're going to be doing video encodes, get a Zen 5 ASAP!

2) PBO -- that's the thing, the power use is a pretty close to exponential at the high end. This has gone on for a LONG time -- I had a Motorola Droid 2 Global, the first thing I did is get an "overclocking" utility to underclock it. They factory overclocked the CPU from 1.1ghz to 1.2ghz -- dropping it from 1.2ghz back down to 1.1ghz cut power use in half, and it cut it by 2/3rds to go down to 900mhz (which, back then, didn't affect anything but 1 or 2 games -- so I'd let it clock up for those 1 or 2 games.)

 

[wizardB]

Intel really needed a break! Thanks AMD!

Intel is already broken it needs no help, it's doing perfectly blowing itself up.

 

[atomsymbol]

Example 1 ... You check benchmarks for your games that use a 4070 and 7800X3D, you find that your processor is delivering the same frames for 4070 in the games you care about ...

The set of games selected by reviewers (1) is just a small sample from the set of games people actually play, (2) is heavily influenced by what the reviewer believes is a game "worth playing" or "worth to be benchmarked", (3) isn't based on gaming statistics pulled from Epic/GOG/Steam/etc, and (4) isn't based on the game's reviews score on Metacritic/Steam/etc. None of TechSpot/TomsHardware/etc adheres to scientific principles.

 

[SAYEDMOHD]

I don't know why you are 100% sure that new 800 series motherboard will not enhance the performance of R7 9700x? Did you see how MSI boost the performance of R7 9700x by applying new bios and you need to use DDR5 6400 CL32 which is priced approximately $120 and classified as affordable

 

[trieste1s]

I think we need to strongly remind all marketing teams that marketing doesn't mean lying.
Zen 5 IS phenomenal for workstation and server and Linux, so just market in that direction. No need to lie about the lack of gaming improvements this round. @AMD

 

[Toliandar]

Makes sense. Only cpu a gamer should get should be an X3D cpu. It makes more sense for AMD to focus on how productivity usage for gains. Performance for general use has been strong enough to be irrelevant for a long time. Gaming performance improvement has been significant, but at best, moderate. Main CPU gaming gains has come from the X3D chips.

The flaw in AMD's plan is not releasing the 9800X3D with the main release of Zen 5.

 

[i like foxes]

Steven , bro , Ryzen 9 9900X and 9950X are a flop indeed , not the ones in question . AMD , what are you doing !? The only consolation - i9 14900K
has worse efficiency , and they are suitable for productivity .

 

[Angelscry]

Long story short, AMD failed to deliver this time. After all the hype, disappointing really.

 

[HardReset]

Long story short, AMD failed to deliver this time. After all the hype, disappointing really.

Long story short: Intel fanboys cannot accept they are wrong: https://openbenchmarking.org/vs/Processor/AMD+Ryzen+9+9950X+16-Core,AMD+Ryzen+9+7950X+16-Core

Failed to deliver my ***.

 

[zx8086]

They are testing the processor not the graphics card. In order to fully give you the absolute best frames the processor can push at 1080p, they need to use the best GPU. If your GPU limited at 1080p on your current processor, then a new processor won't do you much good. It is your responsibility to do the research.

Example 1, you have a 4070 and looking for new CPU - You check benchmarks for your games that use a 4070 and 7800X3D, you find that your processor is delivering the same frames for 4070 in the games you care about - conclusion, you don't need a new processor

Example 2, you have a 5800X with a 4070, you check benchmarks for 4070 for your games where reviewer is using 4070 and 7800X3D. You find that the games you play could be 10% faster at 1080p, however you play at 1440p and the uplift at 1440p is only 2% - conclusion, you don't need a new processor

Example 3, you have a 5800X with a 4080, you check benchmarks for the 4080 that is using a 7800X3D, you see that even at 4K you are losing 5-20% performance in the games you're playing. Some games you are even getting a whole tier less performance out of your GPU. You want to see which processor is right for the 4080, so you then check CPU benchmarks to determine the best processor for you, you notice that the 9700X cost as much as the 7800X3D, but delivers less performance in games, but better productivity. You then have to make a choice, best gaming processor, or better productivity and decent gaming processor? --- You choose 7800X3D as your computer is 90% for gaming...

Excuse me, but where do I find these amazing benchmarks for new processors, seeing as they mostly use different criteria, and so the examples you quote here are mostly as useful as reading tea leaves?
Furthermore, how did you miss the simple fact that the lack of these kind of benchmarks anywhere one the most popular tech sites is the very point I'm making?
To cap it off, no, they're not "testing the processor not the graphics card". Trying to pretend these are scientific articles that need to absolutely maximize te results is a lame excuse (just like saying somebody to "do their own research") because they are not. These are reviews written for consumers, with the main aim of advising people whether to buy or not to buy something. And so it's the reviewers job to conduct and present proper research which is relevant to most people (which can be done alongside the maxing-out method, btw, just needs a bit more resources) and not make them scramble around the web trying to piece together different benchmarks.

 

[WhiteLeaff]

https://www.phoronix.com/review/ryzen-9950x-windows11-ubuntu/7

Windows or perhaps the software version is underutilizing both zen4 and zen4, 20% extra performance is not something I would want to leave on the table.