AMD Ryzen 9000 Zen 5 Single Thread Performance at 5.80 GHz Found 19% Over Zen 4

[JWNoctis]

Nothing stops me or you from buying an RTX 4090 and running it at half power limit - at 225W that will become a very efficient GPU with surprising bit of its performance intact. The problem - this will bring its performance down to lets say RTX 4080 level. RTX 4080 would be much cheaper to buy.

Although if I remember correctly 4090 is most efficient somewhere around 300W where it does not lose as much performance and would still be faster and more effcient than RTX4080. More costly, still.

Off topic, but specifically for this card, its professional equivalent based around the same GPU, RTX 6000 Ada, has a power limit of 300W. Going by that logic, the best efficiency range for the consumer-grade RTX 4090 could be slightly lower than 300W, since it had less of the GPU chip active and only half the VRAM. It stands to reason that professional cards would be rated close to best efficiency for their expected usage pattern, or slightly above to account for overhead from the rest of the system.

In my own experience, I could run a 4070 Ti Super at 200W instead of the rated 285W, and lose maybe 10-15% framerate or compute throughput doing so. The only significantly affected benchmark I've seen so far was FurMark. It also appeared that significant power - on the scale of 50W for this card - would be consumed by the memory bus and VRAM when they run at rated frequency, a consumption that won't be significantly reduced by reducing board power limit, amplifying actual GPU power reduction by percentage in such scheme.

So...yes, it aligns with expectations.

 

[Pavlinius]

AMD fanboys chilax. 14900K at 6Ghz scores way higher at about 980. Mine with no extra voltage and just +1 divider to 61 for 3 cores(6.1Ghz at up to 3 cores) scores 1000. I'd like to see next gen AMD CPU scoring 1000. Also Performance per clock is still worse for AMD. Again 14900K at 6Ghz scores 980. If it scores 910 then it's not reaching this clock. Then per clock performance is better than AMD's 910@5.8Ghz.

 

[stimpy88]

AMD fanboys chilax. 14900K at 6Ghz scores way higher at about 980. Mine with no extra voltage and just +1 divider to 61 for 3 cores(6.1Ghz at up to 3 cores) scores 1000. I'd like to see next gen AMD CPU scoring 1000. Also Performance per clock is still worse for AMD. Again 14900K at 6Ghz scores 980. If it scores 910 then it's not reaching this clock. Then per clock performance is better than AMD's 910@5.8Ghz.

I don't think I've seen one post that fits your description. But good for you, son! Enjoy the chip, while it lasts...

 

[A Computer Guy]

Note 1: High power consumption is an industry-wide trend, in both desktop CPUs and desktop GPUs, enabled by advances in chip manufacturing and by larger&heavier coolers.

Also an industry trend are Laptops, Tablets, and NUC like boxes. Never before have they been so prevalent with a dizzying array of low power multi-core devices.

 

[RandallFlagg]

I guess I never realized what a turd Zen 4 was on 1T.

So now Zen 5 = 14900K and people claiming an AMD win. I think not.

 

[ARF]

I guess I never realized what a turd Zen 4 was on 1T.

Zen 4 is a 2022 technology, it's normal that AMD lags behind. They have to release a refresh or a new generation.
But still, 50,000 Cinebench R23 points is a result which no other CPU has achieved.

 

[Konceptz]

Ya'll love b*tching about power consumption, even though its high on both sides...

I don't think I've seen one post that fits your description. But good for you, son! Enjoy the chip, while it lasts...

The raptorlake14900k/ks failures are highly exaggerated son!

 

[ARF]

Ya'll love b*tching about power consumption, even though its high on both sides...
The raptorlake14900k/ks failures are highly exaggerated son!

Everything is written in the internet, but there is no one to read it

 

[Makaveli]

AMD fanboys chilax. 14900K at 6Ghz scores way higher at about 980. Mine with no extra voltage and just +1 divider to 61 for 3 cores(6.1Ghz at up to 3 cores) scores 1000. I'd like to see next gen AMD CPU scoring 1000. Also Performance per clock is still worse for AMD. Again 14900K at 6Ghz scores 980. If it scores 910 then it's not reaching this clock. Then per clock performance is better than AMD's 910@5.8Ghz.

so you created an account to troll?

 

[Konceptz]

Everything is written in the internet, but there is no one to read it

View attachment 349237

Yeah....in testing....in real world usage with a power limit (news flash, you should be setting a power limit on AMD side as well) they dont come anywhere close...but go on...

 

[ARF]

Yeah....in testing....in real world usage with a power limit (news flash, you should be setting a power limit on AMD side as well) they dont come anywhere close...but go on...

When you underclock it, it gets slower. AMD still wins the performance per watt graphs.

Intel Core i9-14900K Raptor Lake Tested at Power Limits Down to 35 W

There's no denying that Intel's Core i9-14900K is a power hog. This article examines the power, performance and thermals effects of limiting its power consumption, with very interesting results. Additionally, we investigate undervolting to determine if efficiency can be optimized even further.

www.techpowerup.com

 

[Konceptz]

When you underclock it, it gets slower. AMD still wins the performance per watt graphs.

View attachment 349238

Intel Core i9-14900K Raptor Lake Tested at Power Limits Down to 35 W

There's no denying that Intel's Core i9-14900K is a power hog. This article examines the power, performance and thermals effects of limiting its power consumption, with very interesting results. Additionally, we investigate undervolting to determine if efficiency can be optimized even further.

www.techpowerup.com

Yeah, those tests were done before bios updates allowed you to disable CEP. I know you're AMD biased but facts are facts..

Also...gaming isn't the only thing you can do on a PC....

https://www.msi.com/blog/lowering-cpu-voltage-and-temperature-without-compromising-performance-disabling-cep-on-intel-14th-gen-non-k-CPUs

 

[Kapone33]

Yeah, those tests were done before bios updates allowed you to disable CEP. I know you're AMD biased but facts are facts..

Also...gaming isn't the only thing you can do on a PC....

https://www.msi.com/blog/lowering-cpu-voltage-and-temperature-without-compromising-performance-disabling-cep-on-intel-14th-gen-non-k-CPUs

MSI have something with Intel that we are not aware of. They are the only user selling an Intel based handheld ($899) so expect them to wax towards Intel. When the mobile version of this launches it will be in the next Steam deck which outsells everything else in that space. We are in a CPU War but we have sites like TPU that prove the truth. The narrative is powerful though. There was enough noise about instability that Nvidia blamed Intel who then blamed the MB makers and we go on but Intel chips are still more expensive anyway, the hubris.

 

[Shatun_Bear]

AMD fanboys chilax. 14900K at 6Ghz scores way higher at about 980. Mine with no extra voltage and just +1 divider to 61 for 3 cores(6.1Ghz at up to 3 cores) scores 1000. I'd like to see next gen AMD CPU scoring 1000. Also Performance per clock is still worse for AMD. Again 14900K at 6Ghz scores 980. If it scores 910 then it's not reaching this clock. Then per clock performance is better than AMD's 910@5.8Ghz.

What's your power draw to achieve 980 score? 500 watts from the CPU alone?

 

[RandallFlagg]

What's your power draw to achieve 980 score? 500 watts from the CPU alone?

Single thread?

 

[londiste]

What's your power draw to achieve 980 score? 500 watts from the CPU alone?

Not the same test but good enough for reference:
(as an offtopic sidenote - I seriously doubt the 7800X3D's 17W number)

 

[mkppo]

I guess I never realized what a turd Zen 4 was on 1T.

So now Zen 5 = 14900K and people claiming an AMD win. I think not.

In the CPU-z bench, sure. But that bench is a pretty terrible bench as it has no connection to any real world workload and tests a very small subset of the CPU itself. Real world workloads? Zen 4 is pretty damn close to 14th gen, even in their unstably boosted state. Anandtech did a lot of SPEC comparisons back then and this is what they had to say with regard to 13900k vs 7950x's IPC:

"With Raptor Lake being more of a transitional and enhanced core design that Intel's worked with before (Alder Lake), it remains to be seen what the future of 2023 holds for Intel's advancement in IPC and single-threaded performance. Right now, however SPEC paints a picture where it's pretty much neck and neck between Raptor Cove and Zen 4"

Zen 5 won't be equal in IPC to 14900K, it'll be comfortably ahead.

 

[RandallFlagg]

In the CPU-z bench, sure. But that bench is a pretty terrible bench as it has no connection to any real world workload and tests a very small subset of the CPU itself. Real world workloads?

Have you ever looked at what it tests? Never mind, I already know the answer.

Zen 4 is pretty damn close to 14th gen, even in their unstably boosted state. Anandtech did a lot of SPEC comparisons back then and this is what they had to say with regard to 13900k vs 7950x's IPC:

"With Raptor Lake being more of a transitional and enhanced core design that Intel's worked with before (Alder Lake), it remains to be seen what the future of 2023 holds for Intel's advancement in IPC and single-threaded performance. Right now, however SPEC paints a picture where it's pretty much neck and neck between Raptor Cove and Zen 4"

Zen 5 won't be equal in IPC to 14900K, it'll be comfortably ahead.

AT opinion is garbage, has been for years.

 

[mkppo]

Have you ever looked at what it tests? Never mind, I already know the answer.



AT opinion is garbage, has been for years.

Their opinion is based on a whole suite of SPEC tests. Granted this was after Ian Cuttress left and Gavin picked it up, but Ian in his own channel had the same opinion. Now what are you going to say, Ian's opinion is garbage too? Surely the CPU-z test matters more because it's so relevant to real world workloads eh?

Also if you don't agree with their opinion, make up your own after seeing the numbers. I don't think it'll be any different. edit: numbers for you https://www.anandtech.com/show/17601/intel-core-i9-13900k-and-i5-13600k-review/6

Edit: Oh sorry, you asked me what the CPU-z bench tests and then go on to say never mind you already know the answer. Strange. Well, instead of explaining it, i'll just paste the source of where I based my statement from: https://chipsandcheese.com/2023/11/03/cpu-zs-inadequate-benchmark/

Just don't say chips and cheese is garbage too lol

 

[ARF]

Not the same test but good enough for reference:
(as an offtopic sidenote - I seriously doubt the 7800X3D's 17W number)

Do you think it is in fact less?

AT opinion is garbage, has been for years.

Err, then look at the DIY market, and actual sales figures, so you will know the opinion of the society as a whole.

x.com

x.com

If you have updated data, share it.

 

[RandallFlagg]

Their opinion is based on a whole suite of SPEC tests. Granted this was after Ian Cuttress left and Gavin picked it up, but Ian in his own channel had the same opinion. Now what are you going to say, Ian's opinion is garbage too? Surely the CPU-z test matters more because it's so relevant to real world workloads eh?

Also if you don't agree with their opinion, make up your own after seeing the numbers. I don't think it'll be any different.

Edit: Oh sorry, you asked me what the CPU-z bench tests and then go on to say never mind you already know the answer. Strange. Well, instead of explaining it, i'll just paste the source of where I based my statement from: https://chipsandcheese.com/2023/11/03/cpu-zs-inadequate-benchmark/

Just don't say chips and cheese is garbage too lol

Ian's opinion is most certainly garbage. He has always been more concerned with the esoteric architecture and changes to architecture than the useable performance of systems. He's a PHD that works alongside college dropouts.

As far as CPU-Z, your own link explains that due to its size it is a fairly pure IPC benchmark. Data mostly fits in L1/L2 cache. So, what we're saying here is that the actual core of Zen 5, ignoring memory and IO performance, is no faster than Raptor Lake.

That's a big miss for AMD. And that is probably why they are rushing to market first. They need to sell what they can, before Arrow Lake smashes them.

 

[ARF]

They need to sell what they can, before Arrow Lake smashes them.

Without HT, and before AMD even caring to increase the core count from the old 16-core figure to at least 24-core figure?
Err, good luck waiting that.

 

[londiste]

offtopic:

Do you think it is in fact less?

If it is CPU package, it is definitely too little. If it is cores only, feels like too much (and way WAY more than I expected on other Zen4 CPUs).

Wanted to test myself but it looks like something strange is going on with my 7800X3D. It does show somewhat low numbers on various single-core workloads but it also never goes beyond 4.2GHz for some reason...

 

[mkppo]

Ian's opinion is most certainly garbage. He has always been more concerned with the esoteric architecture and changes to architecture than the useable performance of systems. He's a PHD that works alongside college dropouts.

As far as CPU-Z, your own link explains that due to its size it is a fairly pure IPC benchmark. Data mostly fits in L1/L2 cache. So, what we're saying here is that the actual core of Zen 5, ignoring memory and IO performance, is no faster than Raptor Lake.

That's a big miss for AMD. And that is probably why they are rushing to market first. They need to sell what they can, before Arrow Lake smashes them.

Ian has always been great at explaining what architectural changes there are and what workloads will benefit from it. And guess what, he's usually right. What more, he actually uses these systems for real world workloads in the work that he does. Not sure what your last sentence is supposed to mean, what does it matter who he works with? Surely you're not questioning his credentials? I mean, considering Jim Keller has a pretty high opinion of him I don't think we should really go there.

Also, I have absolutely no idea how you read my link and came up with the conclusion that it's a 'pure' IPC benchmark. The article literally states the opposite. I mean, it clearly explains that Zen 3 to Zen 4 had no change in IPC in that bench simply because Zen 4 runs into FP register file capacity which this bench is most sensitive to. Let me quote the conclusion of the article just for reference:

"AMD’s architects likely saw changes that could benefit CPU-Z wouldn’t pay off in other applications. Zen 4 received improvements like a larger micro-op cache, better branch prediction, and doubled L2 cache capacity. Those would help a lot of applications, but not CPU-Z. Thus, CPU-Z’s benchmark ends up being useless to both CPU designers and end users"

To clarify, those changes are what results in the 13% IPC gain from Zen 3 to Zen 4. But guess what the IPC gain in the CPU-z bench is? Zero.

So yeah, the bench is pretty useless as it has no real world relevance and tests a small subset of the CPU. Fitting into the L1 cache doesn't make anything a 'pure IPC' benchmark. No, we're not saying the actual core of Zen 5 is no faster than RPL, it doesn't even test 3/4 of the core lol. The test doesn't test the branch predictors, basically doesn't even touch the front end of the CPU and I can go on. At the least, that would make it a very incomplete IPC bench. Nothing pure about that.

 

[RandallFlagg]

Ian has always been great at explaining what architectural changes there are and what workloads will benefit from it. And guess what, he's usually right. What more, he actually uses these systems for real world workloads in the work that he does. Not sure what your last sentence is supposed to mean, what does it matter who he works with? Surely you're not questioning his credentials? I mean, considering Jim Keller has a pretty high opinion of him I don't think we should really go there.

Also, I have absolutely no idea how you read my link and came up with the conclusion that it's a 'pure' IPC benchmark. The article literally states the opposite. I mean, it clearly explains that Zen 3 to Zen 4 had no change in IPC in that bench simply because Zen 4 runs into FP register file capacity which this bench is most sensitive to. Let me quote the conclusion of the article just for reference:

That might be because, Zen 3 and Zen 4, are not much different at the core level. They mostly changed the IO / memory / cache.

"AMD’s architects likely saw changes that could benefit CPU-Z wouldn’t pay off in other applications. Zen 4 received improvements like a larger micro-op cache, better branch prediction, and doubled L2 cache capacity. Those would help a lot of applications, but not CPU-Z. Thus, CPU-Z’s benchmark ends up being useless to both CPU designers and end users"

So he thinks AMD was looking at CPU-Z, or anyone was? That's absurd.

To clarify, those changes are what results in the 13% IPC gain from Zen 3 to Zen 4. But guess what the IPC gain in the CPU-z bench is? Zero.

So yeah, the bench is pretty useless as it has no real world relevance and tests a small subset of the CPU. Fitting into the L1 cache doesn't make anything a 'pure IPC' benchmark.

Of course it does. If it does not fit in L1, what exactly are you testing?

L1<->L2 latency and bandwidth? L2<->L3 Latency and bandwidth? L3<-> main memory latency and bandwidth? Yes on all counts.

No, we're not saying the actual core of Zen 5 is no faster than RPL, it doesn't even test 3/4 of the core lol. The test doesn't test the branch predictors, basically doesn't even touch the front end of the CPU and I can go on. At the least, that would make it a very incomplete IPC bench. Nothing pure about that.

Work on that reading comprehension thing. It does test branch predictors, just not to the degree that "chipsandcheese" wants.