Why everyone say Zen 5 is bad ?

[SL2]

The 2700X was another case of misleading naming; it should be compared to the 1800X..

I did compare it to the 1800X.

 

[evernessince]

View attachment 358155
AMD Strikes Back: Zen 5 CPU Architecture Changes & Chipset Differences (X870E vs. X870, B850, B840) | GamersNexus

AMD can move sensors to get a more accurate read on specific components and avoid thermal bleeding from adjacent parts of the CPU.

 

[SL2]

People act as if up until Zen 5 both Intel and AMD managed to double performance every generation, and this is the first time we get small improvements.
People also seem to view Zen 5 exclusively from a gaming perspective as if gaming is the only thing people do on their computers.

It began with AMD overpromising, mind you, for the first time in many years when launching a new CPU generation.

This actually is the first time we get such small improvements from AMD in at least 8 years, that's no acting. (Can't say for sure that over 10 years is correct.)

I personally try to stay away from acting here, benchmarks are the way to go. The numbers speaks for themselves, outside gaming as well.

If you want to call looking at benchmarks acting, that's on you.

 

[phanbuey]

AMD can move sensors to get a more accurate read on specific components and avoid thermal bleeding from adjacent parts of the CPU.

"Thermal bleeding" is bit of a stretch - plus wouldn't "thermal bleeding" still cause the same parts of the cpu to run very hot, exactly as hot (if not even hotter) as if the sensor wasnt there? In the end they're getting higher boost by measuring at a cooler point on the CPU.

Im not saying it's right or wrong, just pointing out that the temps may not be actually lower - and that will get OC'ers in potential trouble.

 

[SL2]

In the end they're getting higher boost by measuring at a cooler point on the CPU.

Or in this case, lower boost because of lower TDP. That 2.6 C doesn't make much difference here.

 

[Zazigalka]

11 gen trouble was that in 10 gen i9-10900K was perfect i9 with 10 cores, but in 11 gen 11900K was 8 cores, so just OC i7

If you look at youtube side-by-side comparisons, 11700 loses to 10700 (I'm talking non-K) AND draws more power. RL was an absolute disaster, and it has like 10x higher failure rate than 10th gen too.

 

[marios15]

Everytime we get into IPC discussions, you can quickly see how Intel has ruined the term, by using it for final performance increase, after clockspeed increases
The IPC improvements are there but they were used for power efficiency, Pentium 4 -> Pentium M -> C2D was a similar situation.

It was always the lower power chip - 1700, 2700, 3700 that people bought and used PBO to get near the x800 level.
This looks like AMD went back to Ryzen 1000 product stack there wasn't a 9800X released yet but i bet they will release it. At least this time, there's a reason to buy the 8core at launch.

7700X -> 9700X (5nm to 4nm)
1x process shrink
+13% IPC
+10-15% OC headroom
+62% power/thermal headroom
-40W power consumption

AND higher RAM speed support? This looks amazing for what the 9800X/9800X3D can do!
(Not to mention the 16core will have crazy potential, when accounting for the thermal and IPC improvements)

People are used to always getting some performance increase, because they are used to changing their whole system with Intel on every upgrade
With AMD it does not matter since you can keep the MB, you can pick the 9xxx series and upgrade to a later X3D chip or the 7xxx and upgrade sooner.


I also expect some games that release patches to improve performance for Zen 5, due to the big architecture changes which require compiler support. (linux is showing performance increases for many apps)

 

[Ruru]

Yeah! Give us back pencil mods and BSEL mods!

Personally I saw the use of BSEL mods for only boards without proper BIOS settings. Though unlocking the multiplier from Athlons with pencil was dope.

 

[Zazigalka]

AND higher RAM speed support? This looks amazing for what the 9800X/9800X3D can do!

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.

 

[marios15]

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.

Is it possible to run 8000+ on 7000 with the new AGESA? Another Zen+ moment? It allows more flexibility to be able to run high frequencies, gains are always <5% from RAM tuning

 

[Zazigalka]

Is it possible to run 8000+ on 7000 with the new AGESA?

I think they got 8300 working. It's not worth it though. Just stay on 6400 1:1.

 

[phanbuey]

Or in this case, lower boost because of lower TDP. That 2.6 C doesn't make much difference here.

Im talking specifically for Overclockers -- these chips can actually clock, so I can see people (and motherboard vendors) push these - "Oh we're at 85C - all good keep going!"

meanwhile inside the chip

 

[AnotherReader]

I did compare it to the 1800X.

Yes, that's my mistake. I believe @kapone32 brought the 1700X into the discussion.

 

[evernessince]

"Thermal bleeding" is bit of a stretch - plus wouldn't "thermal bleeding" still cause the same parts of the cpu to run very hot, exactly as hot (if not even hotter) as if the sensor wasnt there? In the end they're getting higher boost by measuring at a cooler point on the CPU.

I imagine that AMD is still measuring those higher temps via another sensor. I believe the goal of sensor relocation in this example is to provide more accurate temp readings to portions of the die that are more temp sensitive and / or more directly correlate to stability / increased clocks. Of course I don't work at AMD so I cannot guarantee this is what they are doing but making the temperature data more relevant to end goal makes sense to me.

 

[DropToasterInTub]

Awful marketing via cherry-picked gaming benchmarks. AMD was marketing this as a huge win for gamers, when in reality, many titles didn't really see much performance improvements, if any at all in many games. Zen 5's IPC is higher, and AVX-512 when not memory bottlenecked can see nearly a 2x performance increase over Zen 4. Scalar integer performance was also up to 35% faster than Zen 4.

Games still predominantly use SSE for calculations, and AMD didn't make any improvements there over Zen 4, hence the disappointing gaming performance. Zen 5 is impressively quick, given the correct workload, and not being memory bottlenecked. AMD really needed to improve the IO die for Zen 5 as the memory bandwidth is quite atrocious; even Zen 4 was already bottlenecked.

This article basically sums up Zen 5's major improvements (straight from the y-cruncher dev): http://www.numberworld.org/blogs/2024_8_7_zen5_avx512_teardown/#ipc_inconsistency

 

[NRANM]

The numbers speaks for themselves, outside gaming as well.

Indeed they do, and they look fine. Basically the same performance overall but with a significantly lower power consumption and thermal output when it comes to all-core workload. That is a decent win in my book. Could it have been better? Sure. But this is still alright. To me at least.
If you don't like it, you are not forced to upgrade.

 

[remixedcat]

It's because it doesn't leave Raptor Lake in the dust and the gains over Zen 4 tend to range from non-existent to minor. People are way too caught up with raw performance numbers and seem to forget what's imo most important - this little chip performs amazing while keeping the watts very low. It gives me Core 2 vibes. I'm very impressed.

and intel wanted to outdo this in benchmarks and look what we got....

 

[oxrufiioxo]

Im talking specifically for Overclockers -- these chips can actually clock, so I can see people (and motherboard vendors) push these - "Oh we're at 85C - all good keep going!"

meanwhile inside the chip

Honestly it's probably just more accurate now on the 7950X3D the cores will be in the 45-55c range and the Tctl will be 75c lol I think this is mostly a case of them playing it too safe with the offset with the older generations....


These CPUs the 9600/9700X just suck by any metric a normal user would use them for at their current pricing people need to stop with all the cope no need to defend AMD they had a good run with 3000-5000-7000 for the most part, they are allowed to miss the mark one generation now if Zen6 is meh and also requires a new socket shame of them though and nobody will trust their we will support X platform for X number of years again at least nobody should.

Just compare it to the 7700 its 3-5% better at gaming depending on who's reviews you like best and up 5-10% faster in applications.... So net gains at a similar wattage are at best in the 5-10% range this is after two whole years.

Even efficiency really isn't that impressive unless you compare it to the very inefficient Zen4 options.....

 

[agent_x007]

You have no point, as there's not point in comparing it with a 7700X. I don't think you're trolling, but I'm not sure.

The 9700X uses 71 W in games.
The 7700 uses 60 W in games.
View attachment 358133

The 9700X is 3% faster and uses 18% more power than a 7700 in games.
View attachment 358134
Is that impressive?

It doesn't deliver the performance AMD promised. AMD's keynotes are usually reliable, but not this time. (No, forget about 5000XT)


Back to the 7700X comparison. Tell me, why should we defend such a product? Exactly half the improvement compared to the 2700X (over the 1800X) which was considered small back in the day, how's that a good thing?
View attachment 358126
A 2% uplift in games, and all you have to say is that we shouldn't expect 25%? How about 6%, is that too much to ask for?

If anyone wants lower power, well that's easily fixed.


It's not a person we're questioning, it's a lifeless commercial product so no need to feel sorry for anyone. Don't treat it, or our intellects, like a baby.

I see this differently : What if future 9700 non-X (IF it exists), is 45W or 50W TDP part instead of 65W and offers similar performance to 7700 non-X ?
I compare "X" model to "X" model, because that's how AMD named them.
Going down on TDP instead of up, for "X" parts, is a trend I'm 100% for.
("max. performance is something you get through OC" - like back in ol' days, yes I'm old )
I just hope we see that 45/50W TDP part at non "X" mark.
But not going to lie here - I don't know AMD plans, if they keep same TDP on non-X parts as last gen - I'm going to be very disappointed.

Marketing is marketing, overhyping is "the usual" (at least to me), and why you always wait for 3-rd party to verify any claims. Getting mad by being "tricked" by it's promises ?
I think of it's like "old man shouts at cloud" thing.
Before launch marketing being correct is such rare situation I find it odd when it occurs (happy to be wrong in such cases ). But back to main thing : Can we force company responsible for "false marketing before launch" to not do it again in future ?
Not really. Unless you don't buy their products ? But how do you want to do that when they sell so many of them?
On top of that you properly have to show them it's marketing fault you aren't buying it vs. chips being plain bad (if they are otherwise good, duh).
There is also a possibility of manufacturer pushing price higher to reduce strain on logistics (by launching at higher price and then reducing it after supply can cope with increased demand). However, I doubt it's the case in Zen 5 for example.

EDIT : I DO NOT want to color Zen 5 as next Sandy Bridge or Zen 2, just a bit reduce negativity surrounding it currently (to show a very positive trend I see).
To enforce this point some more :
DO NOT buy Zen 5 for current (launch) prices, they are NOT worth it.
(at least from what launch up to this day, so 9600X/9700X models)

 

[oxrufiioxo]

EDIT : I DO NOT want to color Zen 5 as next Sandy Bridge or Zen 2, just a bit reduce negativity surrounding it currently (to show a very positive trend I see).
To enforce this point some more :
DO NOT buy Zen 5 for current (launch) prices, they are NOT worth it.
(at least from what launch up to this day, so 9600X/9700X models)

 

[windwhirl]

and intel wanted to outdo this in benchmarks and look what we got....

Speaking of which, I heard the new microcode came out and performance suffered a small hit in a few tests (though, for the rest the performance impact is negligible or non-existent)

Intel Publishes First Microcode Update for Raptor Lake Stability Issue, BIOSes Going Out Now

www.anandtech.com

If one of the comments there is right, people should get it through Windows Update soon-ish (a couple months)

 

[chrcoluk]

Comparing benchmarks with the previous model in TPU reviews, well except for the first one.

                   1800X    2700X    3700X    5800X    7700X    9700X
Cinebench R15 MT   224%*    10%
Cinebench R20 MT   -                 23%      24%
Cinebench R23 MT   -                                   45%
Cinebench 24 MT    -                                            5%
Games 1080p        -        4%       7%       5%       11%      2%

*AT reviews shows that the 95W 1800X is 224% faster than the 220 W FX-9590 in R15. I just had to add it.


I'm not sure the 65 W TDP makes such a difference after comparing with the 7700.

                7700    9700X
Cinebench 24 MT -       9%
Games 1080p     -       3%

Those gen on gen gains werent sustainable for such frequent releases, it was going to come back down to earth at some point.

I do think the FX to Zen1 gain is worth a smile though. FX were dark days for AMD, great recovery from them.

 

[jaszy]

It's not bad. People just have the wrong expectations and AMD wants to compete with itself due to 3D cache.

Gamers won't see any benefits in regards to FPS vs a legacy X3D chip of previous gen... It's been that way since 5800X3D.. which matched 7700X, but at less power.

I can understand why a 6/8 core single CCD chip is a letdown though (gamers), but AMD did this to themselves. Zen 5 has good efficiency and single core performance based on synthetics.

Just wait for X3D. 9800X3D will be the better alternative to 7800X3D.

 

[AusWolf]

You know, I don't see what the fuss is all about. With 9600X and more importantly 9700X we all get:

• A chip on a new production process (4 nm vs Zen 4's 5 nm);

Who cares?

• Cooler processors with lower power consumption compared to Zen 4 equivalents;

Look at the 7700 non-X.

• A brand new architecture;

Who cares?

• 9700X shows roughly the same performance as 7700X or is even faster while working at a base clock speed of 3800 MHz vs 7700X's base speed of 4500 MHz. This is a difference of 700 MHz. Who says then there are no IPC improvements?

Look at the 7700 non-X.

• You can use the new processors in your AM5 board with a BIOS update pretty much on day 1 unlike Intel where almost every new chip requires you to buy a new motherboard.

That's awesome, except that Zen 5 isn't meant for owners of Zen 4 systems.

• You even have a choice. You don't like Zen 5? Then stay on AM4, keep your Zen 4 chip or buy a faster one or wait for Zen 6.

We're still allowed to have an opinion and discuss it with others, right?

And all this is before 9950X and the Zen 5 3D V-Cache models that will be released for sure. In addition I believe AMD will release TDP-relaxed Zen 5 models where the new architecture can really spread its wings.

Zen 4 works equally well with a reduced TDP, just look at the 7700 non-X.

 

[BoggledBeagle]

I kind of frustrated with this.

I get that for example 9700X is on par with performance of 7700X, however is it REALLY enough to claim Zen 5 is "DOA" or "Bad" ?
...
What is wrong with everyone dumping on CPU that is clearly better than previous one ?
...
So, WTF reviewers ?

I would say WTF AMD, when they cannot provide a competitive solution in this price bracket to even a cheaper 13600K, which is now over 2 years old.