AMD Ryzen 9 7950X Geekbenched, Crushes i9-12900K, in the League of the i9-13900K

[Valantar]

Yes very unfair of them to launch their product prior to Intel how dare them f*ck any competition Intel FTW and ALL HAIL QUAD CORES FROM NOW UHNTIL FOREVER AFTER!! I'll wait for the quad cores to launch...I'm sure Intel will have one as people complain 6C's isn't enough for AMD.

Hey now, we all know it's AMD's responsibility to not launch anything that might paint Intel in a bad light

 

[Makaveli]

That's an engineering sample - an OC'd 12900K can hit that score, so at 5.5ghz they would have to have an IPC regression across the product stack for that to hold up (they won't).

Nah there is enough people that have called this out in Reddit and in the comments section in videocards that score is from a cpu at 380 watts.

https://www.reddit.com/r/Amd/comments/x1qril

AMD Ryzen 9 7950X and ASUS ROG Crosshair X670E Extreme also tested in Geekbench - VideoCardz.com

AMD Ryzen 9 7950X on Geekbench A quick follow-up to the first Ryzen 5 7600X benchmarks on Geekbench, the flagship 16-core 7950X has now been tested as well. The AMD Ryzen 9 7950X was tested on ASUS ROG Crosshair X670E Extreme motherboard equipped with 32GB of DDR5-6000 memory. The CPU reached...

videocardz.com

 

[R0H1T]

Any run at or close to the limits of 12900k is hardly representative of anything, not to mention GB is pretty bad for MT scores. It gets over in what under a minute these days? Why don't we OC the 7950x manually & match Intel PL2 limits, with exotic cooling.

 

[Shatun_Bear]

isn't that stock configs? because you can OC the 13600 further then the 7600x

What? Neither are out yet, you wouldn't know. In any case, the 13600 will likely be pushed to the very limit of the aging 10nm node Intel is using, whereas the 7600X should have some breathing room on TSMC's N5 node.

 

[Bomby569]

No, it confirms that the issue is that you don't understand what you're talking about. You don't get a 13% IPC improvement from "some optimizations" that are small enough for it to be "the exact same CPU". That you're actually arguing this is downright absurd. A node shrink doesn't affect IPC, and 13% is a pretty significant improvement. AMD also detailed which specific parts of the cores had seen significant changes to effect these improvements. Is that so hard to grasp?

you have right there what causes the IPC uplift for god sake. Sure i don't understand what i'm talking about. Hopeless, move on.

 

[Valantar]

What? Neither are out yet, you wouldn't know. In any case, the 13600 will likely be pushed to the very limit of the aging 10nm node Intel is using, whereas the 7600X should have some breathing room on TSMC's N5 node.

That's true to some extent, but on the other hand AMD also has far more sophisticated boost control systems, allowing them to run closer to the chip's limits - in previous chips, that is. Whether and how this has changed for these new chips, none of us know. None of this is knowable at this point. Which is part of what makes it interesting IMO - but some people can't seem to handle the uncertainty and need to jump to conclusions rather than accept that life is full of unknowns.

you have right there what causes the IPC uplift for god sake. Sure i don't understand what i'm talking about. Hopeless, move on.

Yes, and ... you're still not managing to even present a coherent argument for how this is "some optimizations", so minor as to render it "the exact same CPU". I'm just asking you to actually back up your words with more than *pointing at thing, looking bewildered*.

 

[phanbuey]

Nah there is enough people that have called this out in Reddit and in the comments section in videocards that score is from a cpu at 380 watts.

https://www.reddit.com/r/Amd/comments/x1qril

AMD Ryzen 9 7950X and ASUS ROG Crosshair X670E Extreme also tested in Geekbench - VideoCardz.com

AMD Ryzen 9 7950X on Geekbench A quick follow-up to the first Ryzen 5 7600X benchmarks on Geekbench, the flagship 16-core 7950X has now been tested as well. The AMD Ryzen 9 7950X was tested on ASUS ROG Crosshair X670E Extreme motherboard equipped with 32GB of DDR5-6000 memory. The CPU reached...

videocardz.com

So what you're saying is... my i5 12600K is getting almost the same ST score as a Raptor Lake at 380W that's also clocked 100Mhz higher, because "people in the reddit comments said it's from a cpu at 380W"

I mean... if you believe that, then there's some stuff I want to sell you. That score is fake or from a ES that's not running properly, a 12900K can top that easily. Mine is not:
Micro-Star International Co., Ltd. MS-7D29 - Geekbench Browser

They're not going to release a product with negative IPC at 380W - also people speculating in reddit comments is probably one of the least reliable sources of information in existence.

 

[R0H1T]

It won't pull 380W from an ST run, even in MT it won't get close to that but Intel's issue is that the same PL2 also allows their chips to boost insanely high for a good length of time. In one of the threads about ADL reviews I pointed out how over longer periods of time 5950x gets close to or beats it in many applications, obviously depending on application. Basically disregard GB for anything but a minor point of reference, it's notoriously bad for reporting clocks as well.

 

[phanbuey]

It won't pull 380W from an ST run, even in MT it won't get close to that but Intel's issue is that the same PL2 also allows their chips to boost insanely high for a good length of time. In one of the threads about ADL reviews I pointed out how over longer periods of time 5950x gets close to or beats it in many applications, obviously depending on application. Basically disregard GB for anything as a minor point of reference, it's notoriously bad for reporting clocks as well.

It's like any other software, it has it's limitations, but from what I see here - there is really a very slim margin between the processors. This competition is really close, hopefully we get some good prices out of it.

cinebench/geekbench/etc. I hope raptorlake will be competitive enough to force down x3d and zen 4 pricing, and it's looking like it will.

 

[R0H1T]

Yes hopefully they don't raise the prices this gen, I just need one more AMD CPU to fill the last AM4 socket I have lying idle.

 

[Makaveli]

So what you're saying is... my i5 12600K is getting almost the same ST score as a Raptor Lake at 380W that's also clocked 100Mhz higher, because "people in the reddit comments said it's from a cpu at 380W"
View attachment 260133

I mean... if you believe that, then there's some stuff I want to sell you. That score is fake or from a ES that's not running properly, a 12900K can top that easily. Mine is not:
Micro-Star International Co., Ltd. MS-7D29 - Geekbench Browser
View attachment 260134

They're not going to release a product with negative IPC at 380W - also people speculating in reddit comments is probably one of the least reliable sources of information in existence.

Yes your overclocked 12600k scores close to the stock raptor lake sample there since most of the improvement going into RPL is from the extra cores to boost multithread performance. Single thread performance is up slightly between them.

a Stock 12600k scores this below.

Its why these comparisons need to be done at stock clocks. So your CPU at stock clocks is not close to that Raptor lake sample at stock.

 

[CrAsHnBuRnXp]

IMO, none of these AMD numbers matter until Raptor Lake gets released and it is benched against the AMD 7000 series. Literally pointless to be comparing it to 12th gen right now.

 

[Punkenjoy]

The key thing for having good and affordable high speed/low latency DDR5 is when server CPU with DDR5 will start to chip. Memory manufacturer will have to pump out massive number of DDR5 chip. Server just use JEDEC spec so they won't need those chips and memory manufacturer will be able to bin the best chips for XMP/EXPO memory.

Genoa is scheduled to launch at the same time as Ryzen 7000. I expect the demand initially for DDR5 will be very high at first but few month after launch, there should be plenty of good memory kit for the platform.

Anyway, i still think people should at least wait for Raptor lake official independent benchmark before buying. And if we don't have a crypto surge and a Global pandemic, it's best to wait a bit after launch for price to settle.

 

[phanbuey]

Yes your overclocked 12600k scores close to the stock raptor lake sample there since most of the improvement going into RPL is from the extra cores to boost multithread performance. Single thread performance is up slightly between them.

a Stock 12600k scores this below.

Its why these comparisons need to be done at stock clocks. So your CPU at stock clocks is not close to that Raptor lake sample at stock.


View attachment 260138

Raptor lake is running at 5.5Ghz my OC is 100Mhz below that... so yes in order for that raptor lake to be that close in score they would have to regress the IPC. Do the math. Stock or not stock doesn't matter. RP @ 5.5Ghz -> 2133 ST vs ADL-S @5.4-> 2121... ADL-S @ 5.5Ghz would then get 2160... so clock per clock you're saying Raptor Lake is SLOWER than cache-starved i5 ADL-S core at the same clocks.

You're talking about a core with more cache a higher ipc being slower in ST at the same clocks -- think for a second why that woudn't make any sense.

 

[Punkenjoy]

Raptor lake is running at 5.5Ghz my OC is 100Mhz below that... so yes in order for that raptor lake to be that close in score they would have to regress the IPC. Do the math. Stock or not stock doesn't matter. RP @ 5.5Ghz -> 2133 ST vs ADL-S @5.4-> 2121... ADL-S @ 5.5Ghz would then get 2160... so clock per clock you're saying Raptor Lake is SLOWER than ADL.

There are some rumors that need to be taken with a grain of salt, but that most of the ST perf gain of Raptor lake will come from a frequency increase and the actual IPC gain is very small. We will see. The rumors state that Intel worked mostly on MT perf and a lot of the design choice went into that direction instead.

We will see, they already had a very strong ST perf so that is not impossible at all. but that may just come from some bench that aren't really taking advantage of the new architecture. Like if someone would say the Raphael IPC gain is minimal because of the CPU-Z score by example

 

[GhostRyder]

I mean, it all sounds promising. I have been debating my rebuild with what I will end up buying. Personally of the AMD lineup, I think the 7900X sounds the most interesting for me because of its higher base clocks and still very high boost of 5.6 (With 12 cores and 24 threads).

 

[AdmiralThrawn]

With these numbers, why would anybody want to buy low end ryzen? You can buy a 14core i5 for 299 that wins at everything over the 8 core 7600x.

 

[InVasMani]

Gaming?

 

[Divide Overflow]

IMO, none of these AMD numbers matter until Raptor Lake gets released and it is benched against the AMD 7000 series. Literally pointless to be comparing it to 12th gen right now.

And then AMD 7000 X3D gets released in Q1 23, so this is all "literally pointless" for some time, right?

 

[Makaveli]

Raptor lake is running at 5.5Ghz my OC is 100Mhz below that... so yes in order for that raptor lake to be that close in score they would have to regress the IPC. Do the math. Stock or not stock doesn't matter. RP @ 5.5Ghz -> 2133 ST vs ADL-S @5.4-> 2121... ADL-S @ 5.5Ghz would then get 2160... so clock per clock you're saying Raptor Lake is SLOWER than cache-starved i5 ADL-S core at the same clocks.

You're talking about a core with more cache a higher ipc being slower in ST at the same clocks -- think for a second why that woudn't make any sense.

Ok then go on that reddit thread and the videocardz comments and tell all those people they are wrong. Its your opinion vs 20 other people saying the same thing.

 

[phanbuey]

Ok then go on that reddit thread and the videocardz comments and tell all those people they are wrong. Its your opinion vs 20 other people saying the same thing.

Here we have actual numbers. I think analyzing real world numbers vs arguing with 20 jamokes on reddit is a better use of my/our time.

 

[Valantar]

With these numbers, why would anybody want to buy low end ryzen? You can buy a 14core i5 for 299 that wins at everything over the 8 core 7600x.

Except in gaming? Gaming doesn't see meaningful gains past 6c12t, so if the 7600X is faster at that, that makes it an attractive proposition even if it's clearly weaker in nT productivity.

 

[Oberon]

Geekbench is highly memory sensitive. My stock 12900k gets 2050 point in st so... Yeah.

Not in ST is isn't.

 

[Sabotaged_Enigma]

The power of all "P-core"s!
13900K has 16 more E-cores running but gets that score, lol.

 

[Tomorrow]

Just 1 million Ageesa updates to worry about, they still release them now for AM3 even after all this time.

Is someone forcing you to update your BIOS?

To clarify i never had the AM4 USB bug that some people struggled with. And since i dont use fTPM i never experienced any stutter either.
The only bug with my AM4 BIOS that i can think of is that it's laggy to navigate with CSM disabled for some reason and this has never been resolved. And im not sure if this is AMD's or Gigabyte's fault (X570 Master v1).

Not to mention some of them will break bios settings, even from backup, so you'll have to photo your entire tune and reenter them.

Earlier ones occasionally did. Thankfully i have dual BIOS so i can switch back with a flip of a switch.

No one will buy the poor 6-core instead of the far superior Core i5-13600K.
You are looking at performance difference +50% and lower cost!

Yeah im sure no one bough 5600X either and they all bought 12600K instead /s
Lower cost? We dont know 13th gen prices yet but you already know that 13600K will be cheaper than 7600X and will OC better?

It has been discussed already that the intel platform cost will be lower..

For now and that is only if you use DDR4. Going forward Intel will lose that advantage vs AM5.
What about "Intel platform cost" requiring a new socket every two generations? Im sure consumers LOVE that.