[Blameless]

7900X3D (adv: higher cache per core)

Each and every core on a CCX has access to the entire local L3. It makes no sense to divide the cache per-core, nor combine the cache between CCXes, unless the data being worked on by each core has little relation to what the others are doing. There is virtually no scenario where six cores working on the same data is going to have an advantage over eight, all other things being equal, when attached to the same total last level cache.

The only advantage the 7900X3D might have is lower CCD thermal density.

 

[Evelyn2023]

One thing I have not seen covered anywhere.

How does everyone's Zen 4's do with Idle at the desktop doing nothing.

I ask because it just seems that even when at 0-1% usage, the task manager is always showing ~ 4.6-5.2+ Ghz and HWInfo always shows a core or two at those speeds with the others maybe chilling in the 3Ghz range. Is there no idle for these CPU's that chills in the sub GHZ range or at least parks like 12 cores when nothing is going on to save even more power than the idle usage it current has going on? I know the CPU is efficient, but It puzzles me it's always clocked like it's ready to go rather than allowing itself to downclock like all my previous CPU's.

I can switch to Low Power Windows Power Setting and see it clock down into the 2Ghz range, but it still seems to hang out in the 3Ghz range more than not.

In HWiNFO you want to look at the "Effective Clocks", this takes sleep states into account. The CPU is running at 5ghz but it's asleep for the vast majority of those 5 billion cycles. A CPU running at 5ghz but sleeping half of those cycles has an effective clock of 2.5ghz, it's only able to do work for 2.5 billion of the 5 billion cycles because it's asleep for the rest of them. My effective clocks can be as low as 1mhz: 1mhz divided by 5ghz = 0.0002, meaning it's awake for 0.002% of the time and asleep 99.998% of the time, so the power consumption should be very low. Even an "idle" system is making thousands of system calls every second, windows doles out work to each of your processors to handle all this work, most of it doesn't need very many cycles to complete but regardless there's constant work needing to be done so the CPU's are awake a fraction of a percent every second doing some random crap.

 

[Alemancio]

One thing still not clear to me:

Specifically in v-cache reliant games, who wins in performance? 7950X3D (adv: higher clock speed) with Processor Lasso set to only use CCD0 with those specific games, 7900X3D (adv: higher cache per core) set the same way, or 7800X3D (adv: 100% dedicated hardware/software to v-cache efficiency)?

On paper, the 7950x3D should win, but there are other software (scheduler?) things happening that might have the 7800x3D winning. The 7900x3D should always lose to the 7950x3D

 

[dansi]

2 chips 1 hour prime. Only thing we really care about in these screens are avg clocks fclk ability and imc ability. View attachment 2609157 View attachment 2609158

can we say the voltages that needs tweaking for ryzen 7000 are vsoc vdd and vddq? what's your vddio mems3 for these chips?

for asus if you set expo, the vsoc remains as auto but the vdd, vddq and vddiomem are manually set by the profiles

 

[HappySheep]

some interesting videos.

 

[Larkonian]

some interesting videos.

The last video is obviously experiencing the frame generation bug in Hogwarts Legacy. That reviewer is incompetent.

 

[harderthanfire]

One thing still not clear to me:

Specifically in v-cache reliant games, who wins in performance? 7950X3D (adv: higher clock speed) with Processor Lasso set to only use CCD0 with those specific games, 7900X3D (adv: higher cache per core) set the same way, or 7800X3D (adv: 100% dedicated hardware/software to v-cache efficiency)?

Generally the 7950X3D should beat or tie the 7800X3D depending on workload but it will depend heavily on silicon quality and stuff like PBO/boost settings and your cooling setup. 7800X3D has the added benefit of less windows **** messing up your benchmarks. 7900X3D will never win because the cache is shared with the whole CCD so cache per core is irrelevant.

 

[LtMatt]

Generally the 7950X3D should beat or tie the 7800X3D depending on workload but it will depend heavily on silicon quality and stuff like PBO/boost settings and your cooling setup. 7800X3D has the added benefit of less windows **** messing up your benchmarks. 7900X3D will never win because the cache is shared with the whole CCD so cache per core is irrelevant.

Here's my 7950X3D vs my simulated 7800X3D running The Last Of Us, 4K Ultra Settings with DLSS Performance mode.

The Last of Us Part I | 7950X3D Stock + Both CCDs + SMT Enabled + 4090 Stock

The Last of Us Part I | 7950X3D Stock + Cache CCD Only + SMT Enabled + 4090 Stock

 

[Evelyn2023]

Here's my 7950X3D vs my simulated 7800X3D running The Last Of Us, 4K Ultra Settings with DLSS Performance mode.

The Last of Us Part I | 7950X3D Stock + Both CCDs + SMT Enabled + 4090 Stock

The Last of Us Part I | 7950X3D Stock + Cache CCD Only + SMT Enabled + 4090 Stock

From admittedly very little data, it's looking like the 7950's 3D ccd is going to boost much higher than the average 7800x3d. The silicon prediction on 7950's 3d ccds looks to be around 100-110, but people getting their 7800x3ds are seeing 85-95. Mine has 6 between 84 and 89, only the best two are 94 and 95. There's a few other people in this thread with similar numbers, our single core scores are on par with reviewers but multi is 5% worse. It's looking like AMD binned a bunch of 7800x3d's and purposefully sent the big reviewers the best ones. One of reviewers had an issue and even said "AMD told us to expect cinebench multi to be 18,200", but retail 7800x3d's are seeing 17,100 to 17,500 according to the few data points we have here. And yes the 18k number is stock, once the reviewers do PBO they see 18.5k+.

So in short I suspect your simulated 7800x3d is realistically going to be more like "simulated best bin one in a hundred 7800x3d".

 

[chewonthis]

can we say the voltages that needs tweaking for ryzen 7000 are vsoc vdd and vddq? what's your vddio mems3 for these chips?

for asus if you set expo, the vsoc remains as auto but the vdd, vddq and vddiomem are manually set by the profiles

Generally speaking the sweet spot I've found after binning multiple CPUs (at min 10 chips). 1.4soc 1.25 vddio.1.25 vddq 1.4-1.45 vdd (depending on sticks)1.1 misc 1.15 cldo_vddp vddg range 950-1100 and usually independent vddg on multi ccd CPUs.

The auto 1.4 that some of the boards I have used apply with expo/xmp have actually pissed off a few CPUs and caused graphic corruption. Considering mem vddio is actually apu vddio in Ryzen master....this is not a surprise. Vdd is actually mem vddio in Ryzen master.

 

[Jalkion]

Hey friends,

I found a Z790 apex but I'm having regrets I think, I have no issues having to disable e-cores, OC to 5.8GHz and 7600mhz but does it even match a 7800x3d at that point?

I don't mind paying for the best gaming fps, I know 1% and 0.1% lows are slightly better on Intel but that does even matter if you're not a hyper competitive gamer.

I went all out because I can for the first time in years and it's fun to OC. But ehhh

 

[yzonker]

Hey friends,

I found a Z790 apex but I'm having regrets I think, I have no issues having to disable e-cores, OC to 5.8GHz and 7600mhz but does it even match a 7800x3d at that point?

I don't mind paying for the best gaming fps, I know 1% and 0.1% lows are slightly better on Intel but that does even matter if you're not a hyper competitive gamer.

I went all out because I can for the first time in years and it's fun to OC. But ehhh

No, I think AMD has the advantage right now for pure gaming use, even over an Intel rig like you have. I have Intel, but that's due to other reasons, mainly I like dabbling in benchmarking, particularly in 3DMark where Intel dominates.

 

[Nizzen]

Hey friends,

I found a Z790 apex but I'm having regrets I think, I have no issues having to disable e-cores, OC to 5.8GHz and 7600mhz but does it even match a 7800x3d at that point?

I don't mind paying for the best gaming fps, I know 1% and 0.1% lows are slightly better on Intel but that does even matter if you're not a hyper competitive gamer.

I went all out because I can for the first time in years and it's fun to OC. But ehhh

What game do you play?
There is no "best" in every game

The "best" is to have many computers like me. Both AMD and Intel. Then you have the "best" all the time

I'm waiting for the 7800x3d to test. 7950x wasn't good enough in many scenarioes for me.

 

[yzonker]

What game do you play?
There is no "best" in every game

Yea, this is true too. I was speaking in general in my post, but there are certainly exceptions. Like I tell people, buy the platform that best suits your use case.

 

[This is a hat.]

The number of units sold by Mind Factory in the first week of sales of the 7800X3D (1860) is quite remarkable.
(The 7950X3D's first two weeks of sales were 690)

With its high power-to-performance ratio and low heat, we expect to see a lot of high-end gamers switching from competitors this summer.

 

[Jalkion]

What game do you play?
There is no "best" in every game

MMOs(Guild Wars2), Planetside 2, Path of Exile, BF2042.

The first 2 are super single core limited but idk is it's the kind cache has an impact on, it probably does, a lot.

 

[harderthanfire]

View attachment 2609272

View attachment 2609273

View attachment 2609274

The number of units sold by Mind Factory in the first week of sales of the 7800X3D (1860) is quite remarkable.
(The 7950X3D's first two weeks of sales were 690)

With its high power-to-performance ratio and low heat, we expect to see a lot of high-end gamers switching from competitors this summer.

Really shows you the difference in sales volume for the two market segments nice.

 

[Nizzen]

View attachment 2609272

View attachment 2609273

View attachment 2609274

The number of units sold by Mind Factory in the first week of sales of the 7800X3D (1860) is quite remarkable.
(The 7950X3D's first two weeks of sales were 690)

With its high power-to-performance ratio and low heat, we expect to see a lot of high-end gamers switching from competitors this summer.

Germany got almost all 7800x3d 😕

 

[LtMatt]

From admittedly very little data, it's looking like the 7950's 3D ccd is going to boost much higher than the average 7800x3d. The silicon prediction on 7950's 3d ccds looks to be around 100-110, but people getting their 7800x3ds are seeing 85-95. Mine has 6 between 84 and 89, only the best two are 94 and 95. There's a few other people in this thread with similar numbers, our single core scores are on par with reviewers but multi is 5% worse. It's looking like AMD binned a bunch of 7800x3d's and purposefully sent the big reviewers the best ones. One of reviewers had an issue and even said "AMD told us to expect cinebench multi to be 18,200", but retail 7800x3d's are seeing 17,100 to 17,500 according to the few data points we have here. And yes the 18k number is stock, once the reviewers do PBO they see 18.5k+.

So in short I suspect your simulated 7800x3d is realistically going to be more like "simulated best bin one in a hundred 7800x3d".

Did any reviewers show the SP rating to prove that? Otherwise I doubt that is true. However, are 7950 X3D’s using binned chips that will be better quality than 7800X3D? 100%, that’s what you pay the extra money for. 5950X was the same too.

 

[Nizzen]

Did any reviewers show the SP rating to prove that? Otherwise I doubt that is true. However, are 7950 X3D’s using binned chips that will be better quality than 7800X3D? 100%, that’s what you pay the extra money for. 5950X was the same too.

Yes, all the "trash" cores is for 7800x3d, that's for shure It's WAY better binned for 7950x3d.