News AMD Launches Zen 4 Ryzen 7000 CPUs, Launches September 27

[Amdlova]

isnt one of the nice things about amd, is that you dont need to buy a 1000 watt 80 plus titanium psu for your build?

Yeah you need a 1000w psu because you have a graphics card need it. Overclock its dead nowadays. My graphic card draw only 25w and cpu limited 35w but I have a psu with 1125w because it's lol

 

[TCA_ChinChin]

There are no bad products, only bad prices. What I see here from AMD is a product that is held back by pricing, whether that is due to the platform (AM5 mobos + DDR5) prices, the CPU prices, inflation, or price relative to AM4, Alder Lake, and Raptor Lake. Its gonna be a real tough sell for a lot of people I think.

 

[TerryLaze]

These will still use less absolute power than the Intel chips.

If reviewers will stick to default settings it is going to be 10-20W difference...230W vs 240-250W
Since AMD hides the extra TDP in the CPU socket it might not show up in reviews but it's still going to be that difference of 10-20W max.

The biggest question will be efficiency of that extra power. If we are only getting 5% better performance for an increase of 50% in power then that is worthless. If we are getting 50% for 50% then that is worth while.

The 12900k was scaling performance with power up to 200W, above that the difference is within margin of error...didn't stop anybody from reviewing it at up to 300W, and almost always only at 300W or at whatever max their mobo could handle.

If ZEN 4 can't go much above the limit that AMD put on it then it will max out at 230W while raptor might max out at 300W again or maybe even higher, but maxing out and making good use of are different things, as you already pointed at.

https://www.hardwareluxx.de/index.p...-desktop-cpus-alder-lake-im-test.html?start=8

 

[ohio_buckeye]

Your 5600x will likely run cooler (a luxury comparable to a quiet car imo), with less bugs (because the platform is already very mature) and you will still be gpu limited in games. Right now the 5600x is fast, efficient, reliable and cheap. Those are a lot of benefits for not chasing #1.

Considering how close to Alder this Zen4 seems in games I imagine 4800c40 cheapo DDR5 will be slower than Alder with DDR4. Fast ram with fast Zen may be faster in games. Reviews will say more. Which will come when?
Might be a bitter pill when sales open up one day before Intel gets to announce how much faster Raptor is than Alder.

For right now I don’t think I’ll need to limited. Yes pci e 3, but from the stuff I’ve seen at least in mid range parts like a 6600xt or 6700xt there’s very little loss. I imagine the 6700xt should hold for the next couple of years anyway.

After reading all the comments I may just pick up a nice 32gb ddr4 ram kit and later go Intel if they keep the ability for using ddr4 until ddr5 comes down. But I imagine the 5600x has at least a year or two of relevance left.

 

[hannibal]

So $300 7600X. Haha some people hope that would be 7700X.

Well 5600x was also $299 so the same price as previous generation... That is actually much, much cheaper than I was expecting.
Yeah, peoples hopes are quite often really unrealistic!

But the product seems to be considerably faster than previous generation and at the same price... So not a bad release at all!
Most people shooting for the best bang for the buck will stay with Zen3 and AM4, but that is normal in these situations. Two more years and DDR5 may be cheaper than DDR4 and we get low end 65w 7600 or 8600 in that case and we have new value option. Until then 5600 and 5700 are good options if you go for AMD and intel's older stuff with 6 power cores. For gaming 5800x3d is still good pick until 7800x3d comes to market...

 

[shady28]

I won't take some hypothetical numbers and paint a blue sky picture here. As a matter of fact, looking at Intel Alder Lake's power consumption numbers, I am not optimistic about the power consumption of Raptor Lake given that they are based on Intel's 10nm with minor refinements. Also, I see no point in comparing clockspeed because it is just 1 metric that influences performance, while there are other factors at play. Just because it is designed to pull higher power to run at higher clock, does not mean it is better designed.

Dude, anyone who can pull up HWMonitor while gaming should be able to tell you that CPU power draw is trivial. With my 10850K + 2060 I'll sit here in almost any game and CPU will bounce 25-45W the entire time while my GPU is constantly 150-200W.

Now maybe with a 3090 the CPU will start pulling over 100W, I've seen that in tests, but then your GPU will be pulling 400W+. So you are going to be worrying about your stubbed toe while your leg is broken.

Unless you're trying to run a render farm, sweating power draw on you CPU for a gamer is just backwards red herring garbage.

 

[jeremyj_83]

If reviewers will stick to default settings it is going to be 10-20W difference...230W vs 240-250W
Since AMD hides the extra TDP in the CPU socket it might not show up in reviews but it's still going to be that difference of 10-20W max.

AMD isn't hiding extra TDP in the CPU socket. Due to Zen's design there is extra power going to things that are not CPU core specific, eg Infinity Fabric. Intel's design doesn't have that, as of yet, therefore more power budget goes to the CPU cores themself. That said Intel's default is still unlimited turbo which means power will spike very high and stay there, depending on work load. We know that AMD's 170W has a 220W PPT limit and therefore won't really go beyond that assuming things are similar from the last several Zen generations.

The 12900k was scaling performance with power up to 200W, above that the difference is within margin of error...didn't stop anybody from reviewing it at up to 300W, and almost always only at 300W or at whatever max their mobo could handle.

Yes up to 200W performance kept increasing on the 12900k, however, even going to 160W the scale was mediocre at best. 12% more performance for 28% more power, sure that is better than 18% more performance for 62% more power as going to 200W from 125W but even for 160W the 12900k is outside the sweet spot for the power curve. The issue though is the default behavior on Intel chips/motherboards. Since the default is unlimited boost, reviewers will go and review based on that default behavior. This has the advantage of pulling out all the stops to win benchmarks in reviews and that is good press. Most people don't care much if in order to get the most performance they are going to use 40% more power than the competition for 5% better performance.

 

[shady28]

I for one am glad AMD is unlocking power on Zen 4.

It will make the AMD fans look more closely at what power is really used during actual tasks.

They'll realize that yeah, their CPU that can draw 230W under a synthetic all core workload, doesn't draw it outside of those 99.9% meaningless benchmarks.

A 10850K is basically a less efficient 10900K, and outside the 11900K is one of the least efficient CPUs on the planet.

Mine draws an average of 35W across my use cases, gaming work and leisure.

GPUs, on the other hand, draw massive power on almost any game unless you cap the FPS. My 2060 will sit there at 165-200W while my CPU runs 30W.

Power arguments based on unrealistic benchmarks have always been dumb and I'm looking forward to them getting put in their place.

 

[TerryLaze]

AMD isn't hiding extra TDP in the CPU socket. Due to Zen's design there is extra power going to things that are not CPU core specific, eg Infinity Fabric. Intel's design doesn't have that, as of yet, therefore more power budget goes to the CPU cores themself. That said Intel's default is still unlimited turbo which means power will spike very high and stay there, depending on work load. We know that AMD's 170W has a 220W PPT limit and therefore won't really go beyond that assuming things are similar from the last several Zen generations.


Yes up to 200W performance kept increasing on the 12900k, however, even going to 160W the scale was mediocre at best. 12% more performance for 28% more power, sure that is better than 18% more performance for 62% more power as going to 200W from 125W but even for 160W the 12900k is outside the sweet spot for the power curve. The issue though is the default behavior on Intel chips/motherboards. Since the default is unlimited boost, reviewers will go and review based on that default behavior. This has the advantage of pulling out all the stops to win benchmarks in reviews and that is good press. Most people don't care much if in order to get the most performance they are going to use 40% more power than the competition for 5% better performance.

No unlocked/unlimited is not only not the default it also voids your warranty...
Don't confuse what reviewers do with what intel does.

 

[jeremyj_83]

No unlocked/unlimited is not only not the default it also voids your warranty...
Don't confuse what reviewers do with what intel does.

Ummmmmmm bull that it's not default. I don't know how many times WE, as is those here on Tomshardware, have told you that Intel Motherboards come with unlimited tau set BY DEFAULT. Intel knows of this practice and DOES NOT say anything to them about this. IF Intel were not wanting that to be the DEFAULT setting THEN they would inform the manufactures that this cannot be set by default. Also since this is in essence turbo YOU are not altering the state or voltages yourself. Intel's warranty says: "any product which has been modified or operated outside of Intel’s publicly available specifications, including where clock frequencies or voltages have been altered, or where the original identification markings have been removed, altered or obliterated." Again since this is the DEFAULT SETTING on the motherboard and 99% of users never do much if anything in the BIOS Intel WILL cover the unlimited tau.

 

[TerryLaze]

Ummmmmmm bull that it's not default. I don't know how many times WE, as is those here on Tomshardware, have told you that Intel Motherboards come with unlimited tau set BY DEFAULT. Intel knows of this practice and DOES NOT say anything to them about this. IF Intel were not wanting that to be the DEFAULT setting THEN they would inform the manufactures that this cannot be set by default. Also since this is in essence turbo YOU are not altering the state or voltages yourself. Intel's warranty says: "any product which has been modified or operated outside of Intel’s publicly available specifications, including where clock frequencies or voltages have been altered, or where the original identification markings have been removed, altered or obliterated." Again since this is the DEFAULT SETTING on the motherboard and 99% of users never do much if anything in the BIOS Intel WILL cover the unlimited tau.

CPU and MOBO are not the same thing, no matter how much YOU (all of you) want to pretend that it is.
Going above PL2, no matter for what reason, voids your warranty, you can't go and tell intel that it's their fault that you have no idea what a mobo is and what it does.

 

[PCWarrior]

CPU and MOBO are not the same thing, no matter how much YOU (all of you) want to pretend that it is.
Going above PL2, no matter for what reason, voids your warranty, you can't go and tell intel that it's their fault that you have no idea what a mobo is and what it does.

Unlocking power limits does not void your warranty. You won't go above PL2 if you only turbo at the stock all-core turbo frequency which is what unlocking power limits means. To make things clearer below is everything anyone needs to know.

Old multicore enhancement (used by virtually all motherboards up to 8th gen and many motherboards up to 9th gen)

1. The cpu is allowed to draw unlimited power (or something ridiculous like 4096W) for indefinite period of time (disregard τ). The way is typically achieved is by setting PL1=PL2 and PL2 to a ridiculous value (e.g. 4096W).
2. Allow boosting of all cores to the top single-threaded boost frequency. E.g. The 9900K had a 5GHz single core boost and 4.7GHz all core boost. Multicore enhancement would push all cores to boost to 5GHz in an all-core workload.
3. That was considered an overclock by Intel and voided your warranty.

Classic unlocking of power limits = New multicore enhancement (valid since 10th gen):

1. The cpu is allowed to draw unlimited power (or something ridiculous like 4096W) for indefinite period of time (disregard τ). The way is typically achieved is by setting PL1=PL2 and PL2 to a ridiculous value (e.g. 4096W).
2. It follows the stock turbo frequency table. That means that the all-core turbo remains equal to the stock all-core turbo. For example the 10900K has a single turbo boost of 5.3GHz with TVB (5.1GHz TB2.0, 5.2GHz TB3.0) and an all-core of 4.9Ghz with TVB (4.8GHz TB2.0). Under the old multicore enhancement it would be pushed to turbo to 5.3Ghz all-core. But with new multicore enhancement it boosts to the stock all-core turbo of 4.8GHz (or 4.9GHz if thermal velocity boost kicks in if the thermals allow it).
3. This is not considered an overclock by Intel and does not void your warranty.

Note: From 12th gen onwards Intel are themselves setting PL1=PL2 and PL2 is set to a value equal to the highest possible stock all-core frequency power draw (something like running Prime95 AVX with small FFTs). For the 12900K that was 241W. Classically unlocking power limits will remove that limit but it is unnecessary unless of course you want to overclock and push for higher power draw. With 13th gen there will be two PL2 values. One is PL2=350W (serving the same purpose as with 12th gen i.e. equal to the highest possible stock all-core frequency power draw), and one tamer at 228W. And of course you can completely unlock power limits too.

Intel Adaptive Boost (ABT) + unlocking power limits (valid for 11th gen i9 11900K/KS only).

1. The cpu is allowed to draw unlimited power (or something ridiculous like 4096W) for indefinite period of time (disregard τ).
2. It allows all cores to boost to the frequency of TB2.0 for 1-2cores active. The all-core TB2.0 of the 11900K is 4.7Ghz but with adaptive boost the cpu boosts to 5.1GHz i.e. equal to the TB2.0 for 1-2 cores active.
3. This is a lot like the old multicore enhancement, but is not considered an overclock by Intel and does not void your warranty.

 

[TJ Hooker]

It's time to add ecc on pci nvme ssd.

Luckily CRCs are used at the PCIe layer, so you get some error/integrity checking by default with PCIe NVMe drives. PCIe 6.0 improves on this by adding forward error correction.

 

[jeremyj_83]

CPU and MOBO are not the same thing, no matter how much YOU (all of you) want to pretend that it is.
Going above PL2, no matter for what reason, voids your warranty, you can't go and tell intel that it's their fault that you have no idea what a mobo is and what it does.

Considering aren't going above PL2 using your motherboard in default settings your point is moot.

 

[escksu]

CPU and MOBO are not the same thing, no matter how much YOU (all of you) want to pretend that it is.
Going above PL2, no matter for what reason, voids your warranty, you can't go and tell intel that it's their fault that you have no idea what a mobo is and what it does.

Not quite. Because PL2 is simply just power limit. There are still voltage, thermal and clock speed limits (assuming one did not manually increase the voltage and clocks).

There are max boost speeds when using 1 core and it will drop as more cores are utilized.. This will not change regardless of PL2 settings.

 

[KyaraM]

1. AMD included gaming benchmarks in their IPC calculation to inflate the IPC uplift. And the gaming uplift is skewed by the RAM used and its overclocking/tuning as they used AMD EXPO Technology to achieve 6000C30 to get the results they did. They also handpicked some of the synthetic benchmarks including outliers of course. And the number of benchmarks was also such that the effect of the outliers was significant. Without the gaming benchmarks and the two 30+% outliers the geomean is 9.4% which is in line to what AMD said the IPC uplift was going to be back in June. Now they just included a bunch of gaming benchmarks and outliers to skew the results and present the result as the outcome of optimisations. What a scam!

2. AMD conveniently compared their overpriced 7600X versus the 12900K in gaming (and who knows the bias and the meddling in the testing). They did so to justify the pricing of $300 for a pure 6-core in 2022-2023. To say “hey look it is better by 5% than a $600 cpu, so you get something better at half the price, what a steal!”. But in gaming their $700 7950X is only 5.7% faster than the $300 7600X (and 11% faster than the 12900K). So if pricing is to be extrapolated from gaming performance then the 7950X should only cost $317. Make up your mind AMD. Either the 7600X is overpriced or the 7950X. You can’t have it both ways.

3. What I see is AMD being very eager to announce-present Zen 4 as early as they could even though they aren’t ready for launch and even in one month when they will be launching it will be a soft/paper launch. There is a reason for that. They had to announce it before Intel announces Raptor lake. That way they only had to compare Zen 4 against Alderlake and be able to present some big wins. They know fully well that against Raptor lake their big wins will be converted to losses or very narrow unimpressive wins. Reminds me when Intel launched 10000 series HEDT to avoid being compared to 3000 series Threadrippers. Back then the AMD fboys were attacking Intel. Now of course its fair play. Total hypocrites.

I mean. When aren't they? XD

 

[KyaraM]

CPU and MOBO are not the same thing, no matter how much YOU (all of you) want to pretend that it is.
Going above PL2, no matter for what reason, voids your warranty, you can't go and tell intel that it's their fault that you have no idea what a mobo is and what it does.

"I didn't know stealing was a crime, it's not fair to arrest me! I shouldn't need a law degree just to live my life!"
They feel like this tbh...

 

[TerryLaze]

Classic unlocking of power limits = New multicore enhancement (valid since 10th gen):

1. The cpu is allowed to draw unlimited power (or something ridiculous like 4096W) for indefinite period of time (disregard τ). The way is typically achieved is by setting PL1=PL2 and PL2 to a ridiculous value (e.g. 4096W).
2. It follows the stock turbo frequency table. That means that the all-core turbo remains equal to the stock all-core turbo. For example the 10900K has a single turbo boost of 5.3GHz with TVB (5.1GHz TB2.0, 5.2GHz TB3.0) and an all-core of 4.9Ghz with TVB (4.8GHz TB2.0). Under the old multicore enhancement it would be pushed to turbo to 5.3Ghz all-core. But with new multicore enhancement it boosts to the stock all-core turbo of 4.8GHz (or 4.9GHz if thermal velocity boost kicks in if the thermals allow it).
3. This is not considered an overclock by Intel and does not void your warranty.

Fine, even if you think that going above PL2 doesn't void your warranty.

That has nothing to do with "default" settings of motherboards.
Z boards be made for overclocking and most of them will come with the old style MCE enabled as well as just having PL2 unlimited, just like xmp or anything else that can give a boost. Z boards are as far away from a "default" setting as humanly possible.

 

[jeremyj_83]

Fine, even if you think that going above PL2 doesn't void your warranty.

That has nothing to do with "default" settings of motherboards.
Z boards be made for overclocking and most of them will come with the old style MCE enabled as well as just having PL2 unlimited, just like xmp or anything else that can give a boost. Z boards are as far away from a "default" setting as humanly possible.

PL2 is a POWER LIMIT as in the power that a CPU can draw and has nothing to do with time. There is also the TAU which is for HOW LONG you can be at a certain power limit. The only way to go ABOVE PL2 is via manual overclocking. Motherboard manufacturers default the TAU on their boards, especially Z series, to unlimited. The only way to change this is via the BIOS and 99% of people don't go into the BIOS or only set the boot order and RAM and thats's it. Intel knows of this practice and DOES NOT say that they cannot do this. Therefore your CPU runs at the MAXIMUM ALLOWED PL2 for an UNLIMITED time, assuming your cooling is able to handle this amount of waste heat. This IS NOT going above PL2.

Going beyond this lets look at this from a legal perspective. There is ample documentation out there that Intel motherboards default to unlimited Tau and Intel looks the other way. You get an Intel CPU and you use it in the stock setting on motherboard but Tau is unlimited since that was the default setting, you are a new builder and don't know these things. Your CPU blows up for some reason and you make a warranty claim. With Intel knowing of the practice by motherboard manufacturers and not telling them to stop that is LEGALLY allowing the practice and admitting that this is normal behavior and settings. If Intel were to deny your warranty claims there would be a class action lawsuit and they WOULD LOSE.

 

[-Fran-]

PL2 is a POWER LIMIT as in the power that a CPU can draw and has nothing to do with time. There is also the TAU which is for HOW LONG you can be at a certain power limit. The only way to go ABOVE PL2 is via manual overclocking. Motherboard manufacturers default the TAU on their boards, especially Z series, to unlimited. The only way to change this is via the BIOS and 99% of people don't go into the BIOS or only set the boot order and RAM and thats's it. Intel knows of this practice and DOES NOT say that they cannot do this. Therefore your CPU runs at the MAXIMUM ALLOWED PL2 for an UNLIMITED time, assuming your cooling is able to handle this amount of waste heat. This IS NOT going above PL2.

Going beyond this lets look at this from a legal perspective. There is ample documentation out there that Intel motherboards default to unlimited Tau and Intel looks the other way. You get an Intel CPU and you use it in the stock setting on motherboard but Tau is unlimited since that was the default setting, you are a new builder and don't know these things. Your CPU blows up for some reason and you make a warranty claim. With Intel knowing of the practice by motherboard manufacturers and not telling them to stop that is LEGALLY allowing the practice and admitting that this is normal behavior and settings. If Intel were to deny your warranty claims there would be a class action lawsuit and they WOULD LOSE.

Just to add a tiny bit to your point: low end Z690s (I think) and B660s (AsRock low end ones do) limit the power delivery to 150W and 100W max. Some don't even advertise it and still list "i9 12900/K/KS support" and power-throttle them. I'm surprised no one has sued them.

Intel needs to sweep the house a bit there.

Regards.

 

[TerryLaze]

PL2 is a POWER LIMIT as in the power that a CPU can draw and has nothing to do with time. There is also the TAU which is for HOW LONG you can be at a certain power limit. The only way to go ABOVE PL2 is via manual overclocking. Motherboard manufacturers default the TAU on their boards, especially Z series, to unlimited. The only way to change this is via the BIOS and 99% of people don't go into the BIOS or only set the boot order and RAM and thats's it. Intel knows of this practice and DOES NOT say that they cannot do this. Therefore your CPU runs at the MAXIMUM ALLOWED PL2 for an UNLIMITED time, assuming your cooling is able to handle this amount of waste heat. This IS NOT going above PL2.

Going beyond this lets look at this from a legal perspective. There is ample documentation out there that Intel motherboards default to unlimited Tau and Intel looks the other way. You get an Intel CPU and you use it in the stock setting on motherboard but Tau is unlimited since that was the default setting, you are a new builder and don't know these things. Your CPU blows up for some reason and you make a warranty claim. With Intel knowing of the practice by motherboard manufacturers and not telling them to stop that is LEGALLY allowing the practice and admitting that this is normal behavior and settings. If Intel were to deny your warranty claims there would be a class action lawsuit and they WOULD LOSE.

TAU=unlimited is what intel does allow, PL1=PL2=241=forever is allowed by intel.
Allowed and default are different things though.
It's one of two basic settings you can use which means that neither one of those settings is the default, you can use either.

Intel allows people to use base power with a defined TAU interval

And intel also allows people to use turbo power with no defined TAU interval

Neither of them is more default or less default than the other.

 

[TerryLaze]

Just to add a tiny bit to your point: low end Z690s (I think) and B660s (AsRock low end ones do) limit the power delivery to 150W and 100W max. Some don't even advertise it and still list "i9 12900/K/KS support" and power-throttle them. I'm surprised no one has sued them.

Intel needs to sweep the house a bit there.

Regards.

Yet more proof that unlimited power isn't the default...

For them to get sued the CPU would have to not work at all on the mobo.
If they disclaim on their webpage how much power their mobo can deliver to the CPU then that's all they have to do.

 

[jeremyj_83]

TAU=unlimited is what intel does allow, PL1=PL2=241=forever is allowed by intel.
Allowed and default are different things though.
It's one of two basic settings you can use which means that neither one of those settings is the default, you can use either.

Intel allows people to use base power with a defined TAU interval

And intel also allows people to use turbo power with no defined TAU interval

Neither of them is more default or less default than the other.

What do you not understand about the word default in this regard? Tell ya what here is the definition from Merriam-Webster's dictionary for default in regards to computers: "to make a selection automatically in the absence of a choice made by the user"

If a motherboard manufacturer enables unlimited Tau right away in the BIOS that means it is the DEFAULT SETTING!!! Intel HAS NOT said that they do not want the manufacturers to discontinue this practice. Therefore even INTEL views this as normal and therefore would be the DEFAULT.

 

[TerryLaze]

What do you not understand about the word default in this regard? Tell ya what here is the definition from Merriam-Webster's dictionary for default in regards to computers: "to make a selection automatically in the absence of a choice made by the user"

If a motherboard manufacturer enables unlimited Tau right away in the BIOS that means it is the DEFAULT SETTING!!!

For that one mobo, even for every mobo that has it set at unlimited...
Not in general and not for every mobo in existence.
What is default for overclocking is not "the" default, it just is one default.

Intel HAS NOT said that they do not want the manufacturers to discontinue this practice. Therefore even INTEL views this as normal and therefore would be the DEFAULT.

I already showed proof that intel also allows manufacturers to use other settings.

 

[digitalgriffin]

It's time to add ecc on pci nvme ssd.

At the transport level or the drive level? In both cases you are covered. Drive level has multiple redundant layers to ensure the cell data has integrity and data correction. It became necessary with TLC/QLC. The high noise of PCIe 6 necessitated built in error correction protocol at the transport level.