AMD Ryzen 9 7950X3D

[Bagerklestyne]

That's the benefit of having your own fabs too. You can optimize the process to fit your products. The 13900k scales better with more power after about 140 W. TSMC's biggest customers are focused on low power devices. The table below is condensed from the Computerbase.de review of the 13900k. It shows multi-core performance normalized to Core i7-13700K at a PL1 and PL2 of 65 W.

CPU	45 W	65 W	88 W	125 W	142 W	181 W	230 W	241/253 W	Unlimited
Core i9-13900K	92%	117%	135%	153%	159%	–	–	184%	186%
Ryzen 9 7950X	84%	130%	154%	–	180%	–	189%	–	–

Like you, I prefer decent performance at sane power levels, but the superior power scaling of Intel's process at high power levels is undeniable.

Did I misread that, but isn't the 7950X putting in 180% vs 159% at 142w ? How is higher power scaling favouring Intel ? I see it favouring Intel at the 45w mark (speculatively i'd say crossing over at about 52-54w)

@fevgatos I'm not trolling or being rude here, do you have any tables (or reviews) to substantiate your version that Intel is the most efficient out of the box in different skus ?

 

[Chrispy_]

@fevgatos I'm not trolling or being rude here, do you have any tables (or reviews) to substantiate your version that Intel is the most efficient out of the box in different skus ?

The wording you just used is all the explanation you need: Out-of-the-box.

Ryzen out-of-the-box is 230W at 189%
Intel out-of-the-box is 253W at 184%

Sure, at 142W the 13900K is probably operating near it's efficiency sweet spot, whilst 142W is quite a long way beyond the efficiency sweet-spot for a 7950X.
YMMV depending on silicon lottery, but I'm finding the 7900X I have at work to run exceptionally well at a 128W PPT which is equivalent to a 95W traditional "TDP". There's still more perfomance on tap but at 128W PPT I'm getting around 90% of the benchmark results that I get with a 250W PPT.

 

[AnotherReader]

Did I misread that, but isn't the 7950X putting in 180% vs 159% at 142w ? How is higher power scaling favouring Intel ? I see it favouring Intel at the 45w mark (speculatively i'd say crossing over at about 52-54w)

The table shows that the 7950X scales better at power levels lower than or equal to 142 W. After that point, it gains only 5% additional performance (189/180) for a 62% increase in power draw. The 13900k gains about 15% more performance going from 142 W to 253 W which is a 78% increase in power draw.

 

[JustBenching]

@fevgatos I'm not trolling or being rude here, do you have any tables (or reviews) to substantiate your version that Intel is the most efficient out of the box in different skus ?

Intel has skus that run at 65w or 35w out of the box, like the 13900 or the 13900t (or 13700 / 12900 / 13700t / 12900t etc.). Those are much more power efficient than eg a 7950x.

 

[Lew Zealand]

Intel has skus that run at 65w or 35w out of the box, like the 13900 or the 13900t (or 13700 / 12900 / 13700t / 12900t etc.). Those are much more power efficient than eg a 7950x.

Except that 65W is not enforced on any Intel Mobos better than a few bottom-end B660s from Asrock and one from MSI. Out of the box, the 13900 and 13900T will run at full, unrestricted wattage and be far less efficient in a realistic Mobo for an i9 (or i7) processor.

 

[JustBenching]

Except that 65W is not enforced on any Intel Mobos better than a few bottom-end B660s from Asrock and one from MSI. Out of the box, the 13900 and 13900T will run at full, unrestricted wattage and be far less efficient in a realistic Mobo for an i9 (or i7) processor.

I do not accept that premise. Have you actually tried? What do you mean out of the box? Like no xmp either? My mobo asks me what cooler I have the first time I try to get in the bios, and unless I choose watercooling or something it sets the default power limits

 

[Chrispy_]

I do not accept that premise. Have you actually tried? What do you mean out of the box? Like no xmp either? My mobo asks me what cooler I have the first time I try to get in the bios, and unless I choose watercooling or something it sets the default power limits

I'm sure it varies by motherboard, but we know without question that motherboard vendors largely ignore Intel's TDP, PL1, and PL2 limits unless you manually enforce them.

Perhaps that's not true on ultra-basic or OEM boards from Dell/HP/Acer etc, but we're talking about the retail boards for the DIY PC market here, I assume?

 

[Bagerklestyne]

The wording you just used is all the explanation you need: Out-of-the-box.

Ryzen out-of-the-box is 230W at 189%
Intel out-of-the-box is 253W at 184%

So out of the box, Ryzen has more compute at less wattage, that very sentence says Ryzen is more efficient (producing more results for less power)
189/230 > 184/253

The table shows that the 7950X scales better at power levels lower than or equal to 142 W. After that point, it gains only 5% additional performance (189/180) for a 62% increase in power draw. The 13900k gains about 15% more performance going from 142 W to 253 W which is a 78% increase in power draw.

No, that shows that Intel gains more ramping up the power, but even at the end of the table, it requires more power for less compute. Looking at the table, any result beyond 45w (65+) shows the Ryzen with more compute power at the same wattage (so more efficient.) Nowhere on the table other than the 45w column does the Intel chip produce more cmpute results for the same power draw as the Ryzen.

Intel has skus that run at 65w or 35w out of the box, like the 13900 or the 13900t (or 13700 / 12900 / 13700t / 12900t etc.). Those are much more power efficient than eg a 7950x.

At that's what I asked for, anything to support that, other than citing model numbers with wattages. You've given me 6 models numbers, 2 wattages but not a single compute result for any of those models and at what wattage thoses numbers were derived at.

Without trying to sound rude, you're given me the equivalent of "my car is faster because I said it was" rather than provide a time slip from a track or drag race.

 

[Chrispy_]

So out of the box, Ryzen has more compute at less wattage, that very sentence says Ryzen is more efficient (producing more results for less power)
189/230 > 184/253

Uh yeah, I mixed up you and @fevgatos
Ryzen mainstream parts are more efficient than Intel's mainstream parts out of the box.
Intel has 65W-specific parts, but there's no need for that on AMD because EVERY motherboard uses AMD AGESA and has a 65W Eco-mode built right into the BIOS by AMD without fail.

Off the top of my head, I don't actually know which CPUs are more efficient at 65W. I'd say it's pretty close but the edge might go to Intel because MCM's like AM4 Ryzens actually have a non-trivial power baseline for the infinity fabric that connects the IO die to the CCD(s). It's one reason why laptops mostly use monolithic dies. I guess the upcoming Zen4 APUs to replace Cezanne parts like the 5700G will be vastly more efficient than Intel - simply because they're monolithic and TSMC N5 is vastly superior to Intel's 10nm (yes, I know Intel rebranded it as Intel7 but that doesn't change what it is!)

 

[AnotherReader]

No, that shows that Intel gains more ramping up the power, but even at the end of the table, it requires more power for less compute. Looking at the table, any result beyond 45w (65+) shows the Ryzen with more compute power at the same wattage (so more efficient.) Nowhere on the table other than the 45w column does the Intel chip produce more cmpute results for the same power draw as the Ryzen.

I agree with you. That's why I said that the 7950X scales better at power levels lower than or equal to 142 W. After that, there's less gain than Intel for additional power. Nevertheless, at almost all power levels between 65 W and 230 W, the 7950X is more efficient.

 

[Bagerklestyne]

Uh yeah, I mixed up you and @fevgatos

All good, I figured it might have been a mistake.

I agree with you. That's why I said that the 7950X scales better at power levels lower than or equal to 142 W. After that, there's less gain than Intel for additional power. Nevertheless, at almost all power levels between 65 W and 230 W, the 7950X is more efficient.

Yeah exactly. I feel the fanboi's could in theory say the efficiency gains from 142w and up are better for Intel than Ryzen, but the overall efficiency still falls to Ryzen. But you know what fanbois are like, cherry picking to support the tin foil hat theories they have.

 

[Lew Zealand]

I do not accept that premise. Have you actually tried? What do you mean out of the box? Like no xmp either? My mobo asks me what cooler I have the first time I try to get in the bios, and unless I choose watercooling or something it sets the default power limits

I posted answers to this elsewhere when you've made the same 65W claim. See Techspot's testing of entry-Level and midrange B660 Motherboards. All midrange B660s run all CPUs at unlocked wattages except for one Asus that topped out at 165W because it can't do any more. Most entry B660s run all CPUs at 125+W except ones from Asrock and a single MSI board.

Intel B660 Motherboard VRM, Mid-Range Roundup

In this roundup we have 8 new Intel B660 motherboards that go for from $150 to $250. Spoiler alert: We're looking at a hefty 70% difference between...

www.techspot.com

Best Budget Intel B660 Motherboards, VRM Thermal Test

Our first look at VRM thermal performance of entry-level Intel B660 motherboards. On hand we have models boards from Asus, Gigabyte, MSI, Asrock, Soyo and Maxsun. Pricing...

www.techspot.com

 

[JustBenching]

I posted answers to this elsewhere when you've made the same 65W claim. See Techspot's testing of entry-Level and midrange B660 Motherboards. All midrange B660s run all CPUs at unlocked wattages except for one Asus that topped out at 165W because it can't do any more. Most entry B660s run all CPUs at 125+W except ones from Asrock and a single MSI board.

Intel B660 Motherboard VRM, Mid-Range Roundup

In this roundup we have 8 new Intel B660 motherboards that go for from $150 to $250. Spoiler alert: We're looking at a hefty 70% difference between...

www.techspot.com

Best Budget Intel B660 Motherboards, VRM Thermal Test

Our first look at VRM thermal performance of entry-level Intel B660 motherboards. On hand we have models boards from Asus, Gigabyte, MSI, Asrock, Soyo and Maxsun. Pricing...

www.techspot.com

Does he actually run them literally out of the box? So he doesn't get into the bios,, at all. Not even for xmp?

At that's what I asked for, anything to support that, other than citing model numbers with wattages. You've given me 6 models numbers, 2 wattages but not a single compute result for any of those models and at what wattage thoses numbers were derived at.

Without trying to sound rude, you're given me the equivalent of "my car is faster because I said it was" rather than provide a time slip from a track or drag race.

Isn't the 13900k at let's say 100w more efficient than the 7950x at 230w?

 

[Panther_Seraphin]

Does he actually run them literally out of the box? So he doesn't get into the bios,, at all. Not even for xmp?

Actually it really does vary by mobo by the looks of it:

Best Budget Intel B660 Motherboards, VRM Thermal Test

Our first look at VRM thermal performance of entry-level Intel B660 motherboards. On hand we have models boards from Asus, Gigabyte, MSI, Asrock, Soyo and Maxsun. Pricing...

www.techspot.com

Most entry-level B660 boards limit the 12900K to a PL1 of 125W
with the only exceptions being the Asus Prime B660M-A which allows for up to 200W, the Asrock B660M-HDV which limits power to just 65W and the MSI Pro B660M-A which by default runs K-SKU parts without any power limits in place, so the CPU just uses as much power as it needs and in the case of the 12900K that's around 240 watts.

Isn't the 13900k at let's say 100w more efficient than the 7950x at 230w?

Where are the numbers you are pulling this from? TPU, Gamers Nexus, Linus etc etc etc all show 7950x is more efficent all over the curve apart from PURELY single thread scenario let alone the 7950x3d or plain 7950.

The one thing Intel still beats AMD hands down on is pure single threaded/gaming kingdom also to see those benefits you need a 7900xtx/4090 as a minimum. What you sacrifice for that in the long run is excessive IMO and probably a big decision factor to most other people.

 

[Chrispy_]

I'm not really sure why this discussion is still going? You had a chart with computerbase.de numbers from @AnotherReader last week.

AMD is more efficient between 65W and 230W.
Intel can be pushed further than 250W to get more performance.
Intel wins at 45W because the infinity fabric connecting a multi-chip design has some overhead that Intel doesn't have.

Whatever your choice, Intel can be pushed harder than AMD at the top end but both are close enough in efficiency for the typical TDPs that you shouldn't really worry too much about it either way; Standardised ATX desktop parts are awful for efficiency, there are easily 75W of completely unavoidable losses in your typical desktop before you even choose your brand of CPU, so worrying about whether your CPU is meeting a certain performance level at 88W or 142W isn't really that important. AMD are technically more efficient but that victory is mostly irrelevant for the big ATX desktops with hungry dGPUs and 25W of RGBLED fans running all the time anyway!

 

[zinidoto]

When I tried installing the chipset drivers I only got 7 options. I have gaming mode on and xbox game bar on. Any fix to this? I'm missing PCI Device Manager

 

[Bagerklestyne]

Isn't the 13900k at let's say 100w more efficient than the 7950x at 230w?

The table I read above says at 230w the 7950x is producing 1.89x the performance of the 13700k and the 13900k is producing 1.84x.
So I would say it's most definitely NOT 100w more efficient @ 230w, in fact it's actually less efficient on a watt to watt basis until you force both cpu's down below 45w which isn't an out of the box configuration

 

[JustBenching]

The table I read above says at 230w the 7950x is producing 1.89x the performance of the 13700k and the 13900k is producing 1.84x.
So I would say it's most definitely NOT 100w more efficient @ 230w, in fact it's actually less efficient on a watt to watt basis until you force both cpu's down below 45w which isn't an out of the box configuration

Re read my post, cause that's not what I said

 

[Bagerklestyne]

Re read my post, cause that's not what I said

I did, I misread it. However

Also according to the table at 88w, 13900k is 1.35x a 13700k, at 125w it's 1.53x at 13700k
but
at 88w the 7950x is 1.54x a 13700k

This tells us that no matter what wattage you give a 13900k between 88w and 125w the 7950x is more efficient, it puts out more compute at 88w than the 13900k does at 125w

So at 100w the Ryzen still trumps the Intel chip for efficiency.

Again, I'm not being rude, but I am just posting the numbers I see.
13900k is a great chip for the top gamers, but efficiency is most definitely not it's forte compared to other offerings, even out of the box.

 

[SL2]

Anyone who expected like a >10 % lead over the competition is delsuional and/or isn't up to date with CPUs. Both brands have great CPUs IMO, that works well in both desktop and laptop.

Spoiler: When AMD or Intel have a separate archiecture for laptop CPUs...

...it usually means that their desktop CPUs aren't the best:

2005 - PD + PM
2011 - Bulldozer + Llano
2021 - Rocket + Tiger

You can easily see it in the market share history.

Not really comparable in terms of success, as Bulldozer was bad, and extra bad at launch. The other two desktop CPU's weren't that bad, but paled in comparison to their respective successors.
That kind of performance jump wasn't feasible this time IMO. The 14.5 % for the 7950X3D over the -X is fantastic tho.. same CPU, slower speed, more cache.

Nice summary here, although I've just looked at the numbers, my german isn't very good.

Launch-Analyse AMD Ryzen 9 7900X3D & 7950X3D (Seite 3) | 3DCenter.org

Sonntag, 6. März 2022 / von Leonidas Wieder einmal herausragend ist dann die Energieeffizienz, mit welcher AMD die dargebrachte Performance erreicht: Diesbezüglich sind schon die regulären Ryzen 7000 Prozessoren nicht schlecht, der

www.3dcenter.org

 

[AnotherReader]

Phoronix recently wrote an article focusing on the non gaming workloads where the 7950X3D excels. Most of these are unlikely to be encountered by the average forum member, but it's interesting to see the tremendous benefits in some software such as GROMACS and PENNANT (link to PDF).

 

[Earthrise]

I was hoping to play Zelda far enough to go to places that have meaningful load, but not even time to start playing yet :/ Maybe over the summer for the Fall 2023 CPU Rebench

Zelda BotW would be really interesting! I'm getting barely 60 FPS with my 5800X3D, right from the beginning. Would be great to know how the 7800X3D compares to the 13900K in that game.

 

[desiprofessor]

PBO in all fields 222 because it's quick to type and high enough, auto oc max, scalar max, CO -15 or -20, nothing special on memory

Hey W1zzard, need your help please. I want to identify if there is something wrong with my chip, cooling or bios settings?

I'm using PBO, (Have it set between -38 and -24 on cores 0-7). I have Boost set to +200. I am using an h150i elite LCD 360 aio corsair cooler.

However I am not able to get boost speeds constantly at 5.25 for CCD 0. I see tons of people getting that so I'm wondering what's going on with mine. I get between 4.9 to 5.1 ghz on games, and it goes down to like 4.8 on occasion. on HWinfo I see it gets to 5.24 but it's not sustained.

I'm using Gskill cl30 6000mhz ddr5. I have used expo timings, buildzoids timings, and even tweaked it further to 6200mhz cl28, and no luck.

Given that you go those boosts on a noctua I'm worried that this is an issue with one of my components.

Any help here would be greatly appreciated. If it's an issue with the CPU, I can RMA it, but I'm hoping it's just some sort of setting.

Also when you say set scalar to max, do you mean 10x?

Is that safe? I have heard some issues with setting scalar past 1x, so just wondering, thanks!

Thanks so much for the help man!

 

[W1zzard]

Also when you say set scalar to max, do you mean 10x?

Yes

to get boost speeds constantly at 5.25 for CCD 0

I'm not even sure if this is possible. I remember seeing around 5.0 in games, like you do

 

[desiprofessor]

Yes


I'm not even sure if this is possible. I remember seeing around 5.0 in games, like you do

Thanks for clarifying man, reason I'm asking about 5.25 in games is because of videos like this, and threads like this.

(359) R9-7950X3D | Red Dead Redemption 2 | RTX 4090 - YouTube

(359) Hogwarts Legacy @4K NATIVE | ULTRA - RT ON | RTX 4090 + R9 7950X3D - YouTube

(359) Warzone 2 Ashika Island Benchmark | 7950X3D +7900 XTX Nitro | 1440P Competitive Settings - YouTube

(359) Dead Island 2 Benchmark | 7950X3D +7900 XTX Nitro | 1080P Ultra Settings - YouTube

(359) Dead Island 2 Benchmark | 7950X3D +7900 XTX Nitro | 2160P Ultra Settings - YouTube

Reddit - Dive into anything

Also, you don't think there are any concerns with setting scalar to 10x?