Why everyone say Zen 5 is bad ?

[sLowEnd]

There does seem to be a pretty good game performance improvement over W11 23H2 with 24H2 (not limited to just Zen 5, the 7800X3D in the video gets pretty substantial improvements too)

 

[sethmatrix7]

This has happened in the past, MSFT screwing AMD. Their OS is too slow or bs etc etc, need more scheduler lol. I'm not in a hurry so waiting for the chips to land where they're supposed to with the incoming patches is ok for me.

While I agree some fault lies with Windows, AMD didn’t have to release the product in this state and could have waited for the issues to be fixed.

 

[A Computer Guy]

There does seem to be a pretty good game performance improvement over W11 23H2 with 24H2 (not limited to just Zen 5, the 7800X3D in the video gets pretty substantial improvements too)

So basically if you want to use your shiny new CPU to the fullest either convert to Windows 11 24H2 or Linux. RIP Windows 10.

 

[thesmokingman]

While I agree some fault lies with Windows, AMD didn’t have to release the product in this state and could have waited for the issues to be fixed.

MSFT has always been aligned with Intel. Wait for issues to be fixed? That's a good one. This has happened a few times before with MSFT. Chicken or egg...

 

[lexluthermiester]

Other posts have already refuted that point.

Ask me if I care, go ahead... The real benchmarks show the facts. Facts are facts and no one can refute them with any merit whatsoever.

 

[ARF]

Agree with the fellow users:

AMD Ryzen 9 9950X drops to $623, 9900X now $449 in the US, with prices slashed up to 12% in Germany - VideoCardz.com

Ryzen 9000 gets first price cuts Not official AMD price cuts, but retailer price cuts. The AMD Ryzen 9000 series has arrived, with at least four SKUs already announced and launched. However, a few things need to be finalized before the full transition to the new platform can be declared...

videocardz.com

 

[DemonicRyzen666]

Ask me if I care, go ahead... The real benchmarks show the facts. Facts are facts and no one can refute them with any merit whatsoever.

I seriously wonder if the 2 ahead branch prediction is someone getting picked up by the spectre patches & it's seeing the extra instructions as a "trojan" injecting extra instructions, causing a premature flush of both micro-op caches & needed instructions. Which then causes front-end to not bother using the predicted instructions. Wendal from level1techs mentioned that Zen 5 is showing Cache misses on things that are usually never cache misses.

 

[ARF]

I seriously wonder if the 2 ahead branch prediction is someone getting picked up by the spectre patches & it's seeing the extra instructions as a "trojan" injecting extra instructions, causing a premature flush of both micro-op caches & needed instructions. Which then causes front-end to not bother using the predicted instructions. Wendal from level1techs mentioned that Zen 5 is showing Cache misses on things that are usually never cache misses.

How can this be fixed? A firmware update? A BIOS update? Zen 6?

 

[R-T-B]

How can this be fixed? A firmware update? A BIOS update? Zen 6?

First you need to confirm it even exists. I have doubts frankly (if this was really going on performance should be lower than we are seeing in practice, as whats being described is not a cheap operation and should be happening always).

 

[ARF]

It is obvious that Zen 5 is borked. The question is what exactly causes the performance regression, and why it's allowed by AMD's design teams to even exist? Why didn't they cancel this launch, as it proves it is utter failure and disappointment on so many fronts?

AMD Ryzen 9 9950X Review - The New Flagship

With 16 cores and 32 threads, the Ryzen 9 9950X, powered by AMD’s Zen 5 architecture, is the fastest desktop processor we've ever tested. In our review, it breezed through application workloads and delivered high FPS rates in gaming. But at $650, it doesn’t come cheap.

www.techpowerup.com

 

[Lost_Troll]

First you need to confirm it even exists. I have doubts frankly (if this was really going on performance should be lower than we are seeing in practice, as whats being described is not a cheap operation and should be happening always).

Well, according to MSI, they tweaked their latest BIOS and you can now adjust it to get 15% more boost when gamming. Link: Boost the Performance of AMD Ryzen 9000 by 15%. How True it is, has yet to be determined and we will have to see as the updated BIOS's come out, but I do be leave that there will be some performance gains from future driver and firmware updates. Also, both Intel and AMD have done this in the past with firmware updates as well as Nvidia and AMD with their graphics drivers as well.

Another thing that I have pondered (speculated) over, is latest rumor about AMD, is them releasing a 7600x3D CPU in September and the reason they are releasing their 9000x3D chips next year, is that they are splitting the two into their own families of processors. One for the average PC workload and the other for performance use like gamming, witch kinds of makes sense in a way. So, you would have the work CPU family, 9600X, 9700X, 9900X, 9950X and the performance CPU family, 9600x3D, 9700x3D, 9900x3D, 9950x3D with SMT working differently between the families.

 

[RUSerious]

You don't know if calling the chips rejects is accurate.

Rejects was too strong a word, hence the reason I put it in quotes. All the CCDs are tested b/4 assembly. The CCDs with the best electrostatic properties are reserved for the best EPYC CPUs. The CCDs we get for desktop, particularly this early on, have higher leakage CCDs. They perform well at high frequencies, but at a higher energy cost. The Zen5 CCDs, so far, appear to have had a major focus on power efficiency. This will work out very well for Hyperscalers using EPYC CPU as they will give higher perf/watt. The highest operational cost for these huge server farms is electricity; both the power and to cool the server blades.

As far as desktop goes, we will see, purportedly by the time the X3D CPUs come out, how well Win 2H24 plus AMD driver/BIOS updates work out.

 

[Athlonite]

This will work out very well for Hyperscalers using EPYC CPU as they will give higher perf/watt

That pretty much because they are Epyc cores in in a desktop CPU

 

[A Computer Guy]

It is obvious that Zen 5 is borked. The question is what exactly causes the performance regression, and why it's allowed by AMD's design teams to even exist? Why didn't they cancel this launch, as it proves it is utter failure and disappointment on so many fronts?

View attachment 360438

AMD Ryzen 9 9950X Review - The New Flagship

With 16 cores and 32 threads, the Ryzen 9 9950X, powered by AMD’s Zen 5 architecture, is the fastest desktop processor we've ever tested. In our review, it breezed through application workloads and delivered high FPS rates in gaming. But at $650, it doesn’t come cheap.

www.techpowerup.com

Design changes can have unintended consequences and/or tradeoffs since it's impossible to optimize everything all the time past and future. I do have very similar questions on how some performance regressions have manifested too compared to last gen but I'll have to wait for those answers.

So far as I have been following the news and reviews here is the list I've been building in understand the "disappointing" aspect of the 9000 series launch.
- Core design changes favoring server workloads
- Core to core latency increases
- Lack of IMC improvements
- Windows updates/tweaks needed
- UEFI/BIOS updates/tweaks needed
- Market saturation (well performing and affordable 5000 and 7000 parts have already saturated the market and people are remembering the GPU is the most important part for gaming when it comes to the overall cost of your system)
- AMD decided they wanted to be a Software Company ( LOL I almost forgot about that )

From an upgraders perspective this graph below makes more sense. Now imagine what this graph would look like if you were coming from 3950x and you had actually paid the $750-ish MRSP during that time. Initial launch price for 9550x is $650-ish and it seems to me might fall a bit sooner than expected by all the "disappointing" press around the release and reports of initial poor sales. In an optimistic scenario lets say by Christmas legitimate specials popup at $550 this CPU will be totally worth it as an upgrade from 3950x/5950x even taking into account the platform switch from AM4 to AM5. ( Massive price drops are already showing up from questionable vendors on Amazon so beware. )

With the right part choices you could get a very effective 30% upgrade uplift well under $1000 for very similar prices you may have paid for 4-6 years ago. There are no bad CPU's just bad pricing (assuming the CPU isn't actually defective). Zen5 isn't bad it's just not everything everyone wants to to be from the often over hyped expectations constantly surrounding Ryzen. There is no reason to have sour grapes because AMD couldn't reach impossible expectations in a single generational release on a planned long term socket. I can't help but think if AMD launched 9000 series with a more uniform and significant uplift on a new socket people would just be complaining about that like the world was ending and AMD abandoned them.

 

[RUSerious]

That pretty much because they are Epyc cores in in a desktop CPU

Yes, OFC. That is the most important market for AMD.

 

[ARF]

From an upgraders perspective this graph below makes more sense.

It doesn't. Look, the AM5 platform is utter garbage:

AMD Ryzen 9 7950X Review - Impressive 16-core Powerhouse

The Ryzen 9 7950X is a monster CPU. When paired with the right workload it will eat even the 12900K for breakfast. As our review shows, the performance uplifts can be massive: +30-50% gen-over-gen is totally possible. What makes things complicated though, is that keeping the beast cool is almost...

www.techpowerup.com

 

[R-T-B]

It is obvious that Zen 5 is borked. The question is what exactly causes the performance regression, and why it's allowed by AMD's design teams to even exist? Why didn't they cancel this launch, as it proves it is utter failure and disappointment on so many fronts?

View attachment 360438

AMD Ryzen 9 9950X Review - The New Flagship

With 16 cores and 32 threads, the Ryzen 9 9950X, powered by AMD’s Zen 5 architecture, is the fastest desktop processor we've ever tested. In our review, it breezed through application workloads and delivered high FPS rates in gaming. But at $650, it doesn’t come cheap.

www.techpowerup.com

Nice misleading graph crop dude.

Well, according to MSI, they tweaked their latest BIOS and you can now adjust it to get 15% more boost when gamming. Link: Boost the Performance of AMD Ryzen 9000 by 15%. How True it is, has yet to be determined and we will have to see as the updated BIOS's come out, but I do be leave that there will be some performance gains from future driver and firmware updates. Also, both Intel and AMD have done this in the past with firmware updates as well as Nvidia and AMD with their graphics drivers as well.

Another thing that I have pondered (speculated) over, is latest rumor about AMD, is them releasing a 7600x3D CPU in September and the reason they are releasing their 9000x3D chips next year, is that they are splitting the two into their own families of processors. One for the average PC workload and the other for performance use like gamming, witch kinds of makes sense in a way. So, you would have the work CPU family, 9600X, 9700X, 9900X, 9950X and the performance CPU family, 9600x3D, 9700x3D, 9900x3D, 9950x3D with SMT working differently between the families.

Fixing it by firmware I can believe. It being related to spectre fixes interfering I cannot.

 

[igormp]

It doesn't. Look, the AM5 platform is utter garbage:

View attachment 360459
View attachment 360461
View attachment 360458

AMD Ryzen 9 7950X Review - Impressive 16-core Powerhouse

The Ryzen 9 7950X is a monster CPU. When paired with the right workload it will eat even the 12900K for breakfast. As our review shows, the performance uplifts can be massive: +30-50% gen-over-gen is totally possible. What makes things complicated though, is that keeping the beast cool is almost...

www.techpowerup.com

As someone on a 5950x, I'll gladly take a 25W increase in consumption (which is really insignificant compared to the 500W+ of both my GPUs) for a ~87% perf uplift in average (and even higher for some daily tasks of mine that I already mentioned before)

AMD Ryzen 9 9950X & Ryzen 9 9900X Deliver Excellent Linux Performance Review - Phoronix

www.phoronix.com

 

[R-T-B]

It doesn't. Look, the AM5 platform is utter garbage:

View attachment 360459
View attachment 360461
View attachment 360458

AMD Ryzen 9 7950X Review - Impressive 16-core Powerhouse

The Ryzen 9 7950X is a monster CPU. When paired with the right workload it will eat even the 12900K for breakfast. As our review shows, the performance uplifts can be massive: +30-50% gen-over-gen is totally possible. What makes things complicated though, is that keeping the beast cool is almost...

www.techpowerup.com

Just ignore the performance per watt right below being massively better why don't ya.

 

[ShrimpBrime]

In order for reviewers to take accurate data measurements, HPET needs to be enabled in W8 and up. This will help prevent skewed benchmark results.

In competitive benchmarking, any results with HPET disabled are actually invalid. Windows HPET is disabled by default.

Gotta open CMD as admin. Type in.
Bcdedit /set useplatformclock true.
Hit enter.
Restart PC.

If your scores seem different, at least you know they are accurate. There won't be any clock skewing going on.

 

[A Computer Guy]

It doesn't. Look, the AM5 platform is utter garbage:

View attachment 360459
View attachment 360461
View attachment 360458

AMD Ryzen 9 7950X Review - Impressive 16-core Powerhouse

The Ryzen 9 7950X is a monster CPU. When paired with the right workload it will eat even the 12900K for breakfast. As our review shows, the performance uplifts can be massive: +30-50% gen-over-gen is totally possible. What makes things complicated though, is that keeping the beast cool is almost...

www.techpowerup.com

I don't see how you can make that assertion rationally based on a single metric.
Was power management not a thing back then or now? Ok let me get back on point.

I'm pretty sure your trolling me but for the sake of argument.

In reality if you are being productive and/or making money the increase in watts can be trivial/acceptable/not-a-problem compared to the gains. Let's say I have a workload of 7zip compression that I do at least 52 weeks a year for about an hour. ( I actually do weekly backups of my virtual machines. )

Multi-core Power Consumption: 5950x 118w
Multi-core Power Consumption: 7950x 235w
Workload: 7zip compression (7950x is about 27% better)
References
- https://www.techpowerup.com/review/amd-ryzen-9-7950x/15.html
- https://www.techpowerup.com/review/amd-ryzen-9-7950x/24.html

118w * 1.0hr = 118.0w
235w * 0.7hr = 164.5w
0.3hrs of my time weekly = not trivial
15.6 hours of my time per year = priceless
( 7950x just paid for itself assuming I didn't screw up my math )

 

[lexluthermiester]

I seriously wonder if the 2 ahead branch prediction is someone getting picked up by the spectre patches & it's seeing the extra instructions as a "trojan" injecting extra instructions, causing a premature flush of both micro-op caches & needed instructions. Which then causes front-end to not bother using the predicted instructions. Wendal from level1techs mentioned that Zen 5 is showing Cache misses on things that are usually never cache misses.

First you need to confirm it even exists. I have doubts frankly (if this was really going on performance should be lower than we are seeing in practice, as whats being described is not a cheap operation and should be happening always).

I'm not convinced either. At the very least, we need to see a functional(not theoretical) proof of concept.

 

[eidairaman1]

In order for reviewers to take accurate data measurements, HPET needs to be enabled in W8 and up. This will help prevent skewed benchmark results.

In competitive benchmarking, any results with HPET disabled are actually invalid. Windows HPET is disabled by default.

Gotta open CMD as admin. Type in.
Bcdedit /set useplatformclock true.
Hit enter.
Restart PC.

If your scores seem different, at least you know they are accurate. There won't be any clock skewing going on.

I have HPET enabled on my Rig

I don't see how you can make that assertion rationally based on a single metric.
Was power management not a thing back then or now? Ok let me get back on point.

I'm pretty sure your trolling me but for the sake of argument.

In reality if you are being productive and/or making money the increase in watts can be trivial/acceptable/not-a-problem compared to the gains. Let's say I have a workload of 7zip compression that I do at least 52 weeks a year for about an hour. ( I actually do weekly backups of my virtual machines. )

Multi-core Power Consumption: 5950x 118w
Multi-core Power Consumption: 7950x 235w
Workload: 7zip compression (7950x is about 27% better)
References
- https://www.techpowerup.com/review/amd-ryzen-9-7950x/15.html
- https://www.techpowerup.com/review/amd-ryzen-9-7950x/24.html

118w * 1.0hr = 118.0w
235w * 0.7hr = 164.5w
0.3hrs of my time weekly = not trivial
15.6 hours of my time per year = priceless
( 7950x just paid for itself assuming I didn't screw up my math )

Don't argue with igorance man, they prove it by 0.000000000000000000001 second

 

[AusWolf]

Just ignore the performance per watt right below being massively better why don't ya.

How good or bad a platform is is measured in Watts on a randomly selected CPU in a randomly selected group of applications without any consideration of performance, don't ya' know?

 

[dj-electric]

Agree with the fellow users:

View attachment 360425

View attachment 360423

View attachment 360424

AMD Ryzen 9 9950X drops to $623, 9900X now $449 in the US, with prices slashed up to 12% in Germany - VideoCardz.com

Ryzen 9000 gets first price cuts Not official AMD price cuts, but retailer price cuts. The AMD Ryzen 9000 series has arrived, with at least four SKUs already announced and launched. However, a few things need to be finalized before the full transition to the new platform can be declared...

videocardz.com

I honestly have no idea how we got to this degree of armchair silicon engineering on the internet.
You know what? I sort of do. This must have some connection to one of those horrible speculation galore, information deprived tech channels.

People aren't being told why sockets are designed the way they are. They are not being told how pinout and die constraints work with latency, signal integrity and future-proofing. Some people seem to think you can just slap whatever socket size you want, with however many pins you want and launch it as a consumer platform without regard to any other aspect of compatibility or manufacturing.

This has to be one of the largest armchairs of armchair engineers. Go ahead random user, please diagnose the 30,000,000 engineer-hour challenge of transistor fabrication in the world of high speed computing. Just make sure to make quick stops at TSMC and Samsung to tell them how to improve their multi-year outlook and plans of fabrication methodology alongside ASML, Zeiss, Canon and other lithography tool making vendors to make everything sing in complete harmony so you could get your 30% IPC improvement as quickly as you desire. This is all lego-magic anyway.

Never visit unmoderated comment sections of rumor-heavy computer websites, kids.