Why everyone say Zen 5 is bad ?

[azrael]

Sure, but you can do the same by putting a custom TDP (ECO mode) on the 7700X, or by getting a 7700 non-X, and end up paying way less.

Like I said above, a 65 W TDP (88 W PPT) gives my 7700X a 10% MT performance loss with a 61% reduction in power usage.

Doesn't the 9700X give you 7700X performance at 7700 efficiency?

 

[AusWolf]

Doesn't the 9700X give you 7700X performance at 7700 efficiency?

Yes, but the 7700 also gives you (very close to) 7700X performance with 7700 efficiency at a much lower price than that of the 9700X. Here lies the problem. You have two CPUs with very similar performance and efficiency (7700 and 9700X), but one is more expensive because... AVX-512, maybe?

 

[azrael]

Yes, but the 7700 also gives you (very close to) 7700X performance with 7700 efficiency at a much lower price than that of the 9700X. Here lies the problem. You have two CPUs with very similar performance and efficiency (7700 and 9700X), but one is more expensive because... AVX-512, maybe?

I didn't realize the 7700 and the 7700X were that close with regard to performance. As for the price difference, I guess it all comes down to that the 9000 series is new and the 7000 series isn't. It's just how these things are.

 

[HTC]

This a PBO approach on the right direction...
Not just enabling it and let the board apply stupid high limits (The Intel approach )
-15 steps on CurveOptimizer (45~75mV UV) and a 120W PPT

View attachment 358683

1st review i've seen where they compare PBO results of the CPU being tested, to PBO results of previous CPUs, instead of to stock of previous CPUs.

Extra credit for not pushing PBO to it's limit, thus showing a realistic compromise between better performance and worse efficiency.

 

[Zach_01]

1st review i've seen where they compare PBO results of the CPU being tested, to PBO results of previous CPUs, instead of to stock of previous CPUs.

Extra credit for not pushing PBO to it's limit, thus showing a realistic compromise between better performance and worse efficiency.

Right!

To be fair and clear about his charts... on the 7700X he chose to go with a static OC because he said that performance was the same with PBO, but the static was better in terms of power/thermals.
He really test them thoroughly and not just click 1 button (all auto).
Although most users dont even know XMP/DOCP/EXPO exits... let alone about PBO and V/F curves...

 

[SL2]

This a PBO approach on the right direction...

Thanks for the link.

Also, 10 % improvement at stock in 13 games average is more than acceptable, much more than I wished for a couple of pages back.

I didn't realize the 7700 and the 7700X were that close with regard to performance. As for the price difference, I guess it all comes down to that the 9000 series is new and the 7000 series isn't. It's just how these things are.

Then we have pricing that isn't always where you'd expect. Last time I checked pcpartpicker, the 7700X was cheaper than the 7700.

 

[Zach_01]

Thanks for the link.

Also, 10 % improvement at stock in 13 games average is more than acceptable, much more than I wished for a couple of pages back.

What I'm not sure about is, if he used PBO on the stock/6000 configuration or on top of the "DDR5 tweaked" configuration.
Because this has real difference in value of the whole experiment.

 

[dirtyferret]

Pretty good article on techspot on the whole subject even mentions tpu

Were we wrong about Zen 5?

Zen 5 reviews that focused primarily on server workloads, development software, and AI-type benchmarks found Zen 5 to be generally excellent, often offering good performance gains over...

www.techspot.com

 

[Zazigalka]

This a PBO approach on the right direction...
Not just enabling it and let the board apply stupid high limits (The Intel approach )
-15 steps on CurveOptimizer (45~75mV UV) and a 120W PPT

View attachment 358683

5800x3d on ddr4 matching 9700x w. 6000 ddr5, what a beast it still is.

 

[Zach_01]

Pretty good article on techspot on the whole subject even mentions tpu

Were we wrong about Zen 5?

Zen 5 reviews that focused primarily on server workloads, development software, and AI-type benchmarks found Zen 5 to be generally excellent, often offering good performance gains over...

www.techspot.com

Yes I did saw their vid, same thing.

At this point they have to stick with their initial opinion I think, at least for these 2 first samples of Zen5

No doubt that AMD has launched them in not the greatest way and with not the best value.
But I think there's more than meets the eye. I could be totally wrong.
Time will tell

5800x3d on ddr4 matching 9700x w. 6000 ddr5, what a beast it still is.

Who would of thought a few years back that a CPU with significantly dropped clocks and power against the equivalent non3D part would be so much faster in gaming... when the mentality of higher clocks = higher fps was dominating us all for so many years.

Who to blame?

 

[HTC]

Right!

To be fair and clear about his charts... on the 7700X he chose to go with a static OC because he said that performance was the same with PBO, but the static was better in terms of power/thermals.
He really test them thoroughly and not just click 1 button (all auto).
Although most users dont even know XMP/DOCP/EXPO exits... let alone about PBO and V/F curves...

It's puzzling why AMD opted for the 9700X to be a 65W part, instead of ... say ... 80W: performance "@ stock" would be better and the efficiency wouldn't suffer as much, but IT WOULD make the gains over the 7700X "more palpable", instead of "rather small".

 

[Hyderz]

It's puzzling why AMD opted for the 9700X to be a 65W part, instead of ... say ... 80W: performance "@ stock" would be better and the efficiency wouldn't suffer as much, but IT WOULD make the gains over the 7700X "more palpable", instead of "rather small".

im guessing amd knows where the limit is and the cpu might degrade faster if pushed too hard... kinda like what intel is in atm

 

[SL2]

I'm still so curious about variance in reviews.

I'm not looking for vague explanations or guesses like sample variance, different methods/motherboards/benchmark runs. I wan't to hear reviewers discuss their results with each other, and get to some kind of better conclusion than that.

I don't know if I've seen such variance in results from reviewers before.

 

[HTC]

im guessing amd knows where the limit is and the cpu might degrade faster if pushed too hard... kinda like what intel is in atm

Are you saying pushing 15W more @ stock is too much, when with PBO it gets close to TWICE of that?

 

[SL2]

Are you saying pushing 15W more @ stock is too much, when with PBO it gets close to TWICE of that?

Exactly, and how can the 9950X have a 170 W TDP if that were true, wouldn't it also degrade?

 

[authorized]

I'm still so curious about variance in reviews.

I'm not looking for vague explanations or guesses like sample variance, different methods/motherboards/benchmark runs. I wan't to hear reviewers discuss their results with each other, and get to some kind of better conclusion than that.

I don't know if I've seen such variance in results from reviewers before.

Is there much difference between reviews though? I think the data is more or less consistent. Maybe the tone and opinions differ more, as in some say it's "a flop" while others call it "decent" or "ok".

 

[SL2]

Is there much difference between reviews though?

I think so, more than I'm used to.

 

[Zach_01]

What I'm getting out mostly (except AMD's lousy launch choices) is that these chips can benefit more than a bit when some one goes outside AMD's review guidelines, spend some time, and test them with higher than DDR5 6000MT/s + tweaked timings + tweaked PBO (=doesn't do it with 1button click)
Unappealing to the avg user but I'm very interested on the subject.

So seeing how their behavior changes quite a bit with better memory subsystem all around, I cant help but think of what the X3D will bring on those cores performance wise.
Dont want to be too excited about it tho...

 

[rv8000]

computerbase has ddr5 5600-6000-7200-8000 scaling comparisons. It's the same as 7000 series almost to a single percent. 6000 1:1 still is faster than 7000 1:2 anyway.

I wish they would disclose secondary and tertiary timings when they do tests like these. If they're letting the bios optimize after setting primaries and/or running expo/xmp, the secondary/tertiary timings are likely horrendously loose. The best things you could do for gaming on Zen 4 was adjust trfc and trefi. That accompanied with tightening primaries at DDR5 6000 could cut 4-6ns off of latency.

I shake my head everytime I see some generic ram scaling benchmarks from reviewers. So much more can be done.

 

[Caring1]

I'm still so curious about variance in reviews.

I'm not looking for vague explanations or guesses like sample variance, different methods/motherboards/benchmark runs. I wan't to hear reviewers discuss their results with each other, and get to some kind of better conclusion than that.

I don't know if I've seen such variance in results from reviewers before.

This is one answer.
Reviewers possibly swapping CPUs without a clean install.

 

[Lost_Troll]

Reviewers possibly swapping CPUs without a clean install.

He also showed that if you left the BIOS setting in the default mode, it may not work correctly. I also think this is part of the problem when you install the driver without setting the CPPC to driver mode, the driver does not install amdedvcashsvc driver as well.

I think in time AMD will have the 9X Ryzen chips running faster once the issues with the BIOS and microcode are cleaned up and optimized to get things like SMT to work correctly. This launch reminds me of when AM5 was released and there were issues with DRAM optimization and stabilization, but in the end, AMD got straightened out.

 

[SL2]

This is one answer.
Reviewers possibly swapping CPUs without a clean install.

Does this apply to single CCD, non 3D models? I never heard him say so.

I mean I've thought about how reusing Windows might impact reviews, but that's just my guess.

Why everyone say Zen 5 is bad ?​

Well, they're talking about two out of seven models, or maybe even just the 9700X.

It's not the first time one model in a new CPU series gets bad reviews, but that doesn't mean they're all bad.

 

[AusWolf]

Does this apply to single CCD, non 3D models? I never heard him say so.

I mean I've thought about how reusing Windows might impact reviews, but that's just my guess.

I've been reusing my Windows 10 installation since 2-3 total system upgrades and several CPU swaps, and never had any issue. I've had AM4, Intel Rocket Lake, and now AM5 with it, including both a 7700X and 7800X3D, although no dual CCD models.

 

[Assimilator]

Because AMD lied about performance, then had the additional nerve to price these chips higher than their Zen 4 predecessors. Given Intel's issues with their own CPUs this launch should've been a triumph for AMD, but in typical AMD fashion they've managed to snatch defeat from the jaws of victory.

 

[yfn_ratchet]

What I'm getting out mostly (except AMD's lousy launch choices) is that these chips can benefit more than a bit when some one goes outside AMD's review guidelines, spend some time, and test them with higher than DDR5 6000MT/s + tweaked timings + tweaked PBO (=doesn't do it with 1button click)
Unappealing to the avg user but I'm very interested on the subject.

So seeing how their behavior changes quite a bit with better memory subsystem all around, I cant help but think of what the X3D will bring on those cores performance wise.
Dont want to be too excited about it tho...

He also showed that if you left the BIOS setting in the default mode, it may not work correctly. I also think this is part of the problem when you install the driver without setting the CPPC to driver mode, the driver does not install amdedvcashsvc driver as well.

I think in time AMD will have the 9X Ryzen chips running faster once the issues with the BIOS and microcode are cleaned up and optimized to get things like SMT to work correctly. This launch reminds me of when AM5 was released and there were issues with DRAM optimization and stabilization, but in the end, AMD got straightened out.

Hmh. It might be that AMD was in a bit of a rush to get these out the door right near the end, especially with the misprint whoopsie and the scattered performance characteristics. By all means, it could be that over time and with a lot of tweaking these CPUs will begin to form more of a gap with Zen 4. Fingers crossed that the Zen 5 go-around (I doubt they'll call it Ryzen 10k, maybe A-100 or something) comes with an updated IOD along with the whole iGPU thing they do.