Your new CPU not as fast as shown in the review you watched? Silicon lottery affects stock Ryzen performance

[Grim5]

Derbauer made this video comparing 13 7600x CPUs and what we found was very interesting. Even though all are stock and use in the same system - some of the 7600x would run 250mhz slower than the top chips and that would cost it 7% performance


But it gets even more interesting because not only is there a 10c difference in temperature between the top and bottom 7600x but also a difference in power and voltage leading to the worst 7600x having 41% less performance per watt than the top 7600x tested!

The variance here between Ryzens is huge.
To be fair however, it is likely that that there is somewhat less variance between top chips - like a 7950x, but as you go down the the stack the variances become larger

 

[newuser]

It's not a new thing, and there are a lot of variables. Bought a binned celeron 300a from here a long time ago and couldn't get close to the advertised speed with my setup.

My 10700K was tapped out at +100mhz, the 13700k I just bought is boosting to +500 Pcore and +200 Ecore on the same cooling setup. My 4090 falls over at more than +500 vram despite having a metric fornication tonne of cooling.

Problem is, unless you're as rich as creasote you only get one to play with.

 

[Cyber-Mav]

Thats a massive difference in binning quality control.

 

[Cyber-Mav]

My 10700K was tapped out at +100mhz, the 13700k I just bought is boosting to +500 Pcore and +200 Ecore on the same cooling setup. My 4090 falls over at more than +500 vram despite having a metric fornication tonne of cooling.

Problem is, unless you're as rich as creasote you only get one to play with.

Your on about over clocked performance, the video is on about stock out of the box experience

 

[Hotwired]

The out of the box experience will be identical to anyone using any of those cpus.

What did you take away from the video? The worst chip he had was 4% below the average of the 13 tested in Cinebench 23.

Might think you saw a larger gap, you did, Roman used several non-zeroed graphs to inflate the visual difference.

You never see a fps graph and we are only given data to work out two points which are not the best/worst performance but instead... the worst efficiency cpu was 247 and the best efficiency cpu was doing 259fps

 

[MartinPrince]

The out of the box experience will be identical to anyone using any of those cpus.

What did you take away from the video? The worst chip he had was 4% below the average of the 13 tested in Cinebench 23.

Might think you saw a larger gap, you did, Roman used several non-zeroed graphs to inflate the visual difference.

You never see a fps graph and we are only given data to work out two points which are not the best/worst performance but instead... the worst efficiency cpu was 247 and the best efficiency cpu was doing 259fps

He does say that all 13 perform within spec so there is nothing there to be alarmist about or get bent out of shape over.

The very surprising thing is the difference in stock power usage so that there is a 50% difference in the performance per watt amount! Now that was unexpected and something to be cognisant of, especially when comparing with other reviews or when some individuals proclaim the best feature of this AMD CPU is power efficiency. You could get a very good sample or a poor one and the difference is notable.

 

[Grim5]

That is correct; the alarming part is not performance variance between the 13 CPUs - what is alarming is the large variance in power draw and voltages while gaming

 

[keyser van someone]

Where do we stand with ordering 5 CPUs and picking the best one and returning the others? Not that we would do such a thing.

 

[Hotwired]

Where do we stand with ordering 5 CPUs and picking the best one and returning the others? Not that we would do such a thing.

By the return rules you'd be able to do that.

Don't think they're obliged to eat the delivery costs if the reason is no fault.

 

[Cyber-Mav]

Where do we stand with ordering 5 CPUs and picking the best one and returning the others? Not that we would do such a thing.

More reason to believe that reviewers received golden samples.

 

[shalke]

More reason to believe that reviewers received golden samples.

That can be always the case. But also if you look at other data about the reviews, they aren't the same.

Launch-Analyse AMD Ryzen 7 7800X3D (Seite 2) | 3DCenter.org

Montag, 10. April 2023 / von Leonidas Wie üblich geht es bei der Ermittlung der Spiele-Performance moderner Prozessoren nicht darum, deren "Unterstützungs-Leistung" für die Grafikkarten im Gaming-Alltag abzubilden –

www.3dcenter.org

 

[no_1_dave]

My 13900k performs better in Cinebench out of the box than any of the reviews I've watched
I guess that means some must go the other way... I'd be annoyed if I was one of those owners

 

[shalke]

Didn't we use to hunt for specific batch revisions in the past that are typically high silicon bin?

 

[keyser van someone]

Didn't we use to hunt for specific batch revisions in the past that are typically high silicon bin?

The stepping? I think. Good times.

I also checked and there was a website that used to provide such binned CPUs at a premium - I hadn't thought about for a while! But it closed ages ago.

I'm not sure I could be bothered to return a CPU if I ended up with a hot one tbh. Interesting stuff though.

 

[Tetras]

Didn't we use to hunt for specific batch revisions in the past that are typically high silicon bin?

Yeah, used to be a thing back in the days before the turbos that killed meaningful value overclocks.

 

[oldbanana]

i dont understand the drama all of the cpus worked in the advertised range

 

[Tetras]

i dont understand the drama all of the cpus worked in the advertised range

Statistically, the variance he talked about (though a little short on some of the details to help us take it in context) means that one CPU reviews (and their conclusions) are basically worthless, especially if you're trying to make decisions between CPUs or platforms based on efficiency & power consumption.

 

[newuser]

Your on about over clocked performance, the video is on about stock out of the box experience

I'd have thought that oc performance was a fair indicator of chip quality?

 

[Sythe79]

They aren't really stock, otherwise they would all be power limited to about 90w max and probably all perform about the same.

 

[q974739]

Statistically, the variance he talked about (though a little short on some of the details to help us take it in context) means that one CPU reviews (and their conclusions) are basically worthless, especially if you're trying to make decisions between CPUs or platforms based on efficiency & power consumption.

That's harsh. It means that differences of less than 5% probably aren't worth caring about. CPU xxxxx is 30% faster? That's still meaningful.