We found the Missing Performance: Zen 5 Tested with SMT Disabled

[Metroid]

How much of the blame for the scheduler doing stupid things is on Microsoft, AMD, and application developers - proportionally?


Exploits exist where a process can read data in the SMT pipeline from the other process sharing the core, despite the two processes not supposed to have access to the data from the other.

SMT exploits where the security barrier fell down due to the shared nature of two instructions running through a core at the same time were the working principle behind Spectre and Meltdown.

I have a hunch that is the reason Intel dropped hyperthreading aka smt and I believe that is the path forward.

 

[Rubinhood]

Nice finding, well done staff!
Makes me wonder if the Windows scheduler can't properly allocate Zen5 logical CPU's yet.
That would also align with the more favorable Linux Zen5 benchmarks on Phoronix.

 

[R0H1T]

The 7800x3d shows SMT is not the (biggest) issue. The bottleneck is through memory/cache & sometimes clock speeds.

 

[Dr. Dro]

This small 3-5% improvement in games is about the minimum you can expect and in line with what you should get out of any CPU. Disabling this will dramatically lower power consumption and current requirements, so there CPU will boost harder and be colder while at it.

 

[mb194dc]

Should bin gaming tests below 1440p.

No one with a 4090 and Zen 5 will use 1080p or 720p. So what's the point of testing them ?

 

[R0H1T]

Should bin gaming tests for all CPU's then, games even at 720p are likely more affected by the GPU than CPU. Gaming by definition is a graphics intensive test & no amount of (multiple) GPU's on a system will change that!

 

[SL2]

This small 3-5% improvement in games is about the minimum you can expect and in line with what you should get out of any CPU.

Quite a difference from the 7700X tho.

Are the clock speeds shown anywhere in the review?

 

[sil3ntearth]

I love you guys, I really do, but this is some serious copium for a bad product launch.

as others pointed out long before this, disabling SMT has almost always resulted in better gaming performance.

 

[Markosz]

LOL 4.7% gaming improvement with SMT off and PBO MAX everything unleashed. Efficiency is out of window, this release is just bad.
They need to cut prices right now if they want to sell these.

 

[R0H1T]

They're doing fine with 5xxx & 7xxx still flying off the shelves, why should they reduce prices? Just so you can splurge 3k on the next RTX 5090 Super reality distortion card

 

[SL2]

They need to cut prices right now if they want to sell these.

No need, AMD have plenty of older chips to sell that are MUCH cheaper to make. Prices will go down eventually.

 

[izy]

Should bin gaming tests below 1440p.

No one with a 4090 and Zen 5 will use 1080p or 720p. So what's the point of testing them ?

How can you test the performance difference between CPUs if you are not using 720/1080p?

 

[P4-630]

So Intel was on to something for Arrow Lake, ditching HT...

 

[Dr. Dro]

Quite a difference from the 7700X tho.

Are the clock speeds shown anywhere in the review?

I don't think so, but in any case the range should not change. The sustainable minimum clock is increased because of how the boost algorithm works, the CPU temperature being lower. With my i9-13900KS there are situations where it can shave 30°C under load.

I read recently Intel claims that the Core Ultra (Arrow Lake) processor consumes 100W less under load... the biggest thing is that it has no SMT in its P-cores. I'd hazard a guess at least half of that power is saved because of that. SMT has become more of a burden than a boon in a modern high-performance processor, a secondary pathway to stalled busy cores is necessary when you have a handful of cores, but when you've got twenty plus, it's simply not needed beyond artificially inflating benchmarks.

 

[izy]

Would be interesting if the x3D part will keep the same clocks as 9700x considering its just an 65W part , maybe they will release an 9800x too @ 105W or something.

 

[mb194dc]

How can you test the performance difference between CPUs if you are not using 720/1080p?

What's the point of know the that difference if no one uses the hardware like that?

How about actually testing at settings that some (any?) people will use, so they can make an informed buying choice based on the review? Unless the actual purpose of the review is to mislead people, so they'll think it's better. But in the real world at their settings, it isn't.

People spending the money for a 4090 and Zen 5 are surely using at least 1440p, maybe in ultra wide, probably at high Hz, maybe 4K or even higher res than that once screens exist. That's what should be tested, obviously along with the CPU focussed stuff like rendering, encoding and the other stuff people would actually buy such kit for.

 

[sodium]

Should bin gaming tests below 1440p.

No one with a 4090 and Zen 5 will use 1080p or 720p. So what's the point of testing them ?

In a CPU comparison you need to remove as much GPU bound situations as possible so your data isn't ruined. And your opinion about what you think someone else will do with their system is irrelevant considering there's 1080p 480Hz oled monitors out and you're still going to need fast hardware to push those numbers.

 

[izy]

What's the point of know the that difference if no one uses the hardware like that?

How about actually testing at settings that some (any?) people will use, so they can make an informed buying choice based on the review? Unless the actual purpose of the review is to mislead people, so they'll think it's better. But in the real world at their settings, it isn't.

People spending the money for a 4090 and Zen 5 are surely using at least 1440p, maybe in ultra wide, probably at high Hz, maybe 4K or even higher res than that once screens exist. That's what should be tested, obviously along with the CPU focussed stuff like rendering, encoding and the other stuff people would actually buy such kit for.

You have the 1440p/4k benchmarks as well and how sodium said there are people that are still playing at 1080p (many e-sports players) plus many of us are interested in the performance uplifts generation by generation and so on, you cant see the CPU improvements if you are GPU limited and as the GPUs improve over time having the max CPU performance makes real sense (i can make you a big list of why and so on but thats not the point of this topic)

 

[blu3dragon]

What's the point of know the that difference if no one uses the hardware like that?

How about actually testing at settings that some (any?) people will use, so they can make an informed buying choice based on the review? Unless the actual purpose of the review is to mislead people, so they'll think it's better. But in the real world at their settings, it isn't.

People spending the money for a 4090 and Zen 5 are surely using at least 1440p, maybe in ultra wide, probably at high Hz, maybe 4K or even higher res than that once screens exist. That's what should be tested, obviously along with the CPU focussed stuff like rendering, encoding and the other stuff people would actually buy such kit for.

The reason for doing this is to get a comparison of relative CPU performance in games when they are not limited by the gpu.

The data for higher resolutions is provided so you can make a decision on an appropriate CPU for a certain GPU based on that.

 

[Wirko]

Some games runs better without SMT
Always true for both AMD and Intel CPUs for a while.

Zen5 is no different.

Yes but this is extreme. To me it looks like broken Windows scheduler, which applies a power saving plan when it's supposed to maximise performance.
I expect Phoronix to jump on this issue and maybe prove it's not an issue in Linux.

 

[Dr. Dro]

Yes but this is extreme. To me it looks like broken Windows scheduler, which applies a power saving plan when it's supposed to maximise performance.
I expect Phoronix to jump on this issue and maybe prove it's not an issue in Linux.

It's not, this problem has been going on for a long time.

 

[Onasi]

It's not, this problem has been going on for a long time.

It really has. To the point where I distinctly remember in early 2010s recommendations to turn off HT on Sandy i7 for games turning them, effectively, into i5s. Of course, over time games evolved a bit and those virtual threads came in a bit useful. But yeah, overall HT/SMT was hit or miss with games for as long as I can remember the technology exists. You’d think that theoretically modern engines should work better with it seeing as how consoles now do use SMT, but seems to be arguable if they do.

 

[SL2]

Would be interesting if the x3D part will keep the same clocks as 9700x considering its just an 65W part , maybe they will release an 9800x too @ 105W or something.

Games benefit from clock speed so I think it will be higher, and I think AMD set this one to 65 W just to make the next 3D look good outside games., which is more expensive anyway so it's reasonable.


A 9800X wouldn't surprise me, but then again, who's it for? Not overclockers (unless heavily binned), not gamers (low cache), not MT (few cores)..

 

[sdedalus83]

Nice finding, well done staff!
Makes me wonder if the Windows scheduler can't properly allocate Zen5 logical CPU's yet.
That would also align with the more favorable Linux Zen5 benchmarks on Phoronix.

The Phoronix results suggest a lot of Windows software needs to be recompiled with a Zen 5 target. The most obvious are the AES and encryption results where Windows tests are showing Zen 5 to be slower than Zen 4, while in Linux Zen 5 is considerably faster, by more than 100% in some cases.

 

[phanbuey]

It really has. To the point where I distinctly remember in early 2010s recommendations to turn off HT on Sandy i7 for games turning them, effectively, into i5s. Of course, over time games evolved a bit and those virtual threads came in a bit useful. But yeah, overall HT/SMT was hit or miss with games for as long as I can remember the technology exists. You’d think that theoretically modern engines should work better with it seeing as how consoles now do use SMT, but seems to be arguable if they do.

Disabling HT back then was a double-digit % boost on the lows. Far Cry 4 went from choppy mess to perfectly playable for me.