News Early sample of AMD's Ryzen 5 9600X is only 12% quicker than Ryzen 5 7600X in CPU-Z benchmark

[Admin]

A CPU-Z benchmark has cropped up featuring the all-new Zen 5-powered Ryzen 5 9600X. However, the chip's CPU-Z performance appears to be weaker than its IPC performance suggests.

Early sample of AMD's Ryzen 5 9600X is only 12% quicker than Ryzen 5 7600X in CPU-Z benchmark : Read more

 

[Deleted member 2731765]

A bit lower than the 16% IPC improvements AMD touted at Computex

Regardless, the CPU-Z numbers suggest that the new Ryzen 5 9600X won't be that much faster than the Ryzen 5 7600X. This makes sense, considering both CPUs share very very similar specifications

However, the chip's CPU-Z performance appears to be weaker than its IPC performance suggests.

Nonsense. Actually, we should not trust CPU-Z scores though, at least for AMD Ryzen CPU benchmarks, for the Zen architecture.

Although, the chip might be having a 12.21% advantage in the single-core test and a 12.30% advantage in the multi-core test over the Ryzen 5 7600X, as per your math, but I wouldn't directly compare it with any IPC gain.

Because the CPU-Z app never took full advantage of Zen4's improvements to the arch like, micro-op cache, branch prediction, L2 cache capacity etc, but other apps did. I expect the same with ZEN 5.

CPU-Z score is a bad metric for comparison, at least for AMD Ryzen CPUs. So it's not worth the time comparing these 1T and NT scores. We also don't know anything about the test environment being used here.

The factors that limit performance in CPU-Z are very different from those in typical real-life workloads. From AMD’s own slides, Zen 4 barely improved over Zen 3 in CPU-Z app benchmark. Zen 4 received improvements like a larger micro-op cache, better branch prediction, and doubled L2 cache capacity.

Those would help a lot of other applications, but not CPU-Z. CPU-Z’s benchmark ended up being of little use to both CPU designers/testers and end consumers.

+1% gain !! Check this IPC chart.

 

[Deleted member 2731765]

The Ryzen 5 9600X scored 871.4 points in CPU-Z's single-core test and 7,096.6 points in the application's multi-core benchmark.

Wrong !

You actually need to make some correction in your entire article though. Those CPU-Z scores are from an overclocked chip. Kindly check the Tweet again. The normal CPU-Z scores are 776 Single, and 6201 MT.

So you need to edit the article. The same sample chip was also tested in CPU-z with an overclock of 5.7 GHz, or a +300 MHz boost over its stock boost clock of 5.4 GHz.

Btw, this OC/overclock was done across ALL cores, while the chip's 5.4 GHz frequency is a single-core boost with the average all-core boost rounding up around 5.0-5.1 GHz.

Here are the non-OC CPU-Z scores.

 

[MWeiss89]

871.4 / 765 = 13.91% increase
7,096.6 / 6221 = 14.07% increase
Where do 12%, 12.21%, and 12.30% come from?

 

[jackt]

I'm disappointed, 4nm ? why not 3nm ? isnt 3nm ready ?
I hope that a 3nm refresh comes soon.

 

[blargh4]

These clickbaity articles about benchmark leaks have got to be my least favorite genre of tech site filler. We already have a good idea of the IPC gains because AMD gave us the numbers - and as the author himself is clearly aware, AMD's 16% number is based on a geometric mean of numerous benchmarks, where Zen 5 exhibits between a 10% and 35% improvement over Zen 4. CPU-Z is not among that set, probably because it's a useless synthetic benchmark.

 

[Makaveli]

lol this article is a mess.

Why zero mention of overclocked scores posted!

 

[Gas Racing]

The performance is closer to 14% difference... just as AMD claimed.
Did the author really write a whole article based on the fact he did math wrong?
How embarrassing.

 

[NeoMorpheus]

Hit piece sponsored by Intel?

 

[usertests]

I'm disappointed, 4nm ? why not 3nm ? isnt 3nm ready ?
I hope that a 3nm refresh comes soon.

The rumor mill said AMD hedged its bets because of TSMC N3 uncertainty. Remember these decisions are made years in advance. I think they will end up with the Zen 5 chiplets on N4 and Zen 5c chiplets on N3. And there's talk of other chiplet variants like a subtly different one for Strix Halo.

The node really doesn't matter. It's a decent enough increase, but it's the same basic design carried over from Zen 2. Zen 6 is set to be a lot more interesting if AMD switches to something like a die-to-die interface (Infinity Links) like High Yield predicts. That's also when we can expect core count increases, and maybe different ways of using L3 cache.

Zen 5 can be cheaper on N4 than N3, so that's a potential point in its favor. We've also heard that AMD might try something different with 3D V-Cache this generation, but no word on what that actually means.

 

[Conor Stewart]

Oh wow, AMD must be lying because they didn't score exactly their average performance increase on some synthetic benchmark that they didn't include.

This is a synthetic benchmark that wasn't included in AMDs published results so it really means nothing. AMD claimed that their next generation chips managed anywhere from 10 - 35 % improvement over the previous generation with a 16 % geomean, so the increase of 12 % falls pretty well inside that range and not too far off of the geomean.

So what is the problem? Either the author doesn't know what they are writing about or they purposely ignored the facts in order to write a clickbait and wrong article. I would think it is a combination especially with all the other errors in the article and acknowledging that AMDs 16 % was average. I suppose we shouldn't expect much more from modern "journalists".

 

[cp0x]

871.4 / 765 = 13.91% increase
7,096.6 / 6221 = 14.07% increase
Where do 12%, 12.21%, and 12.30% come from?

I came here to say this.

Both the single- and the multi-threaded tests showed a 14% improvement.

Still less than the 16% expected, but it does seem reasonable that we'd at least use math or something to calculate the numbers, right?

 

[razor512]

It would be interesting to see if there are any tasks that will be especially benefited from the increased cache throughput.

 

[TerryLaze]

+1% gain !! Check this IPC chart.

This article doesn't talk about IPC improvements...
It's just compared to what the 7600x reaches, and they even compare the overclocked numbers from the 9600x to the not-overclocked numbers of the 7600x to make it look as good as possible.
"AMD's new Zen 5 mid-range chip in CPU-Z have cropped up, showing a 12% performance increase for the 9600X over its predecessor, the Ryzen 5 7600X. "

871.4 / 765 = 13.91% increase
7,096.6 / 6221 = 14.07% increase
Where do 12%, 12.21%, and 12.30% come from?

You compare the o/c of the 9600x to the default numbers of the 9600x , that's not the IPC increase.
The 12% numbers come from comparing the overclocked 9600x to the default 7600x
"AMD's new Zen 5 mid-range chip in CPU-Z have cropped up, showing a 12% performance increase for the 9600X over its predecessor, the Ryzen 5 7600X. "

This is AMDs new low right here, selling these as 65w TDP while you will have to heavily overclock them to get the actual performance out of them.

The improvement from overclocking is 14% while the difference from the 7600x is about 12% ......without overclocking they will only be 2% faster.........

 

[Flayed]

So you can have lower power consumption or higher performance. That doesn't sound unreasonable to me.

 

[je90s]

"CPU-Z's performance estimations suggest the 9600X's performance is weaker than what AMD claimed in its Computex announcement"

this sentence is either paid for, or the author doesn't understand simple interptaion of averages. even if it is 5% better in cpu-z it is not weaker than they claimed.

Also, between 12% and 16% are 4% and if you know one or two about benchmarks, that there is variation of performances between released cpus within 1-2% and test themselves have bigger variations.

We are basing all of this on 1 cpu that is an engineering sample from 1 test and comparing it to an avg of things that are not represented by the cpu-z test in question. so stupid, this is like 4 mistakes at once

 

[kaalus]

12% performance gain while lowering TDP from 105 to 65W? This is an amazing result. A few more like this and x86 might be able to catch up with ARM.

 

[Deleted member 2731765]

This article doesn't talk about IPC improvements...

Oh, of course I know that. That was not my point. I was referring to what the author claims in his own article, and what the IPC chart says. Because he is comparing the CPU-Z scores directly with IPC.

""However, the chip's CPU-Z performance appears to be weaker than its IPC performance suggests.""


PS: Wait, I will get back with more details to clarify this. Right now, I'm battling with my slow internet connection here at a cafe !

 

[TerryLaze]

12% performance gain while lowering TDP from 105 to 65W? This is an amazing result. A few more like this and x86 might be able to catch up with ARM.

NOT while!
This is the bad part that every review is going to try and lie about.
It's either or.
It's either 65W at the same performance OR 12-16% more performance at (my guess) the same PPT as the previous gen, maybe even higher.

Oh, of course I know that. That was not my point. I was referring to what the author claims in his own article, and what the IPC chart says. Because he is comparing the CPU-Z scores directly with IPC.

""However, the chip's CPU-Z performance appears to be weaker than its IPC performance suggests.""


PS: Wait, I will get back with more details to clarify this. Right now, I'm battling with my slow internet connection here at a cafe !

Well there is some wriggle room there, if the overclocked new CPU uses the same amount of power (PPT) as the old one at default, then the result would be the difference in IPC.
The thing is we don't know the power draw at overclock so we can only talk about total performance.

 

[cp0x]

The 12% numbers come from comparing the overclocked 9600x to the default 7600x

14%. The numbers showed a 14% gain.

Also, between 12% and 16% are 4%

Between 14% and 16%. The numbers posted showed a 14% gain.

12% performance gain while lowering TDP

14%. Not 12% as the article incorrectly asserted.

It's either 65W at the same performance OR 12-16%

14%-16%. Nothing showed 12%. The article was simply wrong, because they did the math backwards.

This is normal behavior on Toms Hardware: AMD gets the short end of the stick, and Intel walks on water. It's been this way for 25+ years here. Still a useful site, but you need a heavy Intel BS filter.

 

[TechyIT223]

I'm disappointed, 4nm ? why not 3nm ? isnt 3nm ready ?
I hope that a 3nm refresh comes soon.

3nm not yet ready yet. I suppose both cost and yield issues, or something else entirely

 

[Cooe]

🤦😑 Can someone explain to me why Aaron is still employed here??? O__o

Basically every single article he writes is absolutely LOADED with inaccuracies, misinformation, and misinterpretation... It's like he doesn't fact check literally ANYTHING before he hits "post"!

Those CPU-Z results include overclocked chips AND CPU-Z has been a SHOCKINGLY terribly inaccurate benchmark for Zen CPU's since at least Zen 4! (It literally shows a 0% IPC gain from Zen 3 to 4, and has the R9 7950X like ≈-25-30% behind the i9-13900K in single-core perf when the actual gap is ≈-0-10% workload depending, w/ Zen 4 even beating Raptor Lake in ST perf on occasion). Not to mention this "R5 9600X" almost surely being an underclocked engineering sample with an undercooked BIOS and who knows whatever else issues.

Also, an average IPC gain figure is a GEOMEAN!!! 🤦 Aka, there are going to be specific examples with less than a +16% gain and other examples with more. Does Aaron SERIOUSLY not understand how averages work??? 😑

This article is a freaking embarrassment for Tom's Hardware as a long lived PC hardware enthusiast site. This crap showing up on an ALLLLL too often basis these days is legitimately reputation damaging. Aka, it's WELLLLL past time to call an all-hands-on-deck meeting to make some SERIOUS changes about how this site is run Paul, and I DESPERATELY hope that you do... It's time to clean house.

Otherwise, I seriously struggle to see a long-term future for this site outside of being yet another clickbait regurgitation farm for poorly researched trash "articles" like this flaming stinker. 🤷

 

[TerryLaze]

14%-16%. Nothing showed 12%. The article was simply wrong, because they did the math backwards.

Yup, you are right, going from the lower number to the higher number it's 14% increase.

 

[TerryLaze]

🤦😑 Can someone explain to me why Aaron is still employed here??? O__o

Basically every single article he writes is absolutely LOADED with inaccuracies, misinformation, and misinterpretation... It's like he doesn't fact check literally ANYTHING before he hits "post"!

Those CPU-Z results include overclocked chips AND CPU-Z has been a SHOCKINGLY terribly inaccurate benchmark for Zen CPU's since at least Zen 4! (It literally shows a 0% IPC gain from Zen 3 to 4, and has the R9 7950X like ≈-25-30% behind the i9-13900K in single-core perf when the actual gap is ≈-0-10% workload depending, w/ Zen 4 even beating Raptor Lake in ST perf on occasion). Not to mention this "R5 9600X" almost surely being an underclocked engineering sample with an undercooked BIOS and who knows whatever else issues.

Also, an average IPC gain figure is a GEOMEAN!!! 🤦 Aka, there are going to be specific examples with less than a +16% gain and other examples with more. Does Aaron SERIOUSLY not understand how averages work??? 😑

This article is a freaking embarrassment for Tom's Hardware as a long lived PC hardware enthusiast site. This crap showing up on an ALLLLL too often basis these days is legitimately reputation damaging. Aka, it's WELLLLL past time to call an all-hands-on-deck meeting to make some SERIOUS changes about how this site is run Paul, and I DESPERATELY hope that you do... It's time to clean house.

Otherwise, I seriously struggle to see a long-term future for this site outside of being yet another clickbait regurgitation farm for poorly researched trash "articles" like this flaming stinker. 🤷

Dude take a pill or something, relax and read the article again.
They say : " The new benchmark was discovered by HXL on X (formally Twitter). " so it's an article about a random benchmark that was found, it's not a review of the CPU it's not tom's benchmarking a CPU, it's a random benchmark that they just comment on.

 

[Deleted member 2731765]

Nice to see many math classes going on here, ! lol