Ok guys, I've done
two new benchmarks...
BUT...
Please read the following main points before looking at the scores, because there's specific things to consider (after realizing a few things were going on).
===================================================
#1
I realized quickly enough (after checking Resource Monitor as I was testing various games, including VAM too) that my games (and VAM as well) were - in fact - NOT running on the CCD with the 3D vCache. Now, that one shocked me, because I DO use the XBOX Game Bar (as specified in all the guides I've seen so far about how to 'properly' use the X3D lines of processors for gaming purposes, to take full advantage of the extra vCache), the games ARE considered 'as a game' in the Game Bar (VAM was, as well, which I applied manually) AND... on top of that, all my drivers are up to date (Game Bar itself is the latest version, including the Game Bar service version); Chipset (latest for Windows 11, for my specific motherboard model), Audio, Ethernet and of course Graphics; and obviously Windows 11 itself is all up to date.
Which leads me to conclude that - at least as of now - relying on just Windows scheduling in "combination" with the XBOX Game Bar 'solution' does NOT work as described in most of the guides I've seen out there about "how to" make sure that Windows throws Apps at the CCD with the extra vCache. It is _supposed_ to work that way, yes. And I'm 100% sure that most people would tell me I did something wrong but NOPE. I triple-checked everything, nothing is 'wrong', it's just that for some reason everything I was doing was non-3D vCache'ed, going to the CPU portion that is running at a higher Frequency instead (I mean it, everything... Browsing, checking movies on VLC Player, Gaming, using VAM, toying around in Blender... ALL of it was thrown at my
Non-3D vCached CCD).
#2
Due to the 'problem' as stated above, I decided to - instead - 'force' the whole "Apps should run on the extra 3D vCache CCD by priority" thing via the BIOS (thankfully, some guides around do recommend to do that, instead of just 'letting' Windows doing it; which obviously didn't work in my case). There's that option in the BIOS, the CPPC or CPCC or whatever it is, can't recall the name right now. But basically, you can go either Auto, Driver (*Auto is just like Driver from what I've seen online, but by default it was Auto for me but... it doesn't seem to work anyways in Windows afterwards, neither would the 'Driver' option I assume), then you can also go with Frequency, or Cache.
Going with Frequency = basically what was happening to me in Windows by 'default', I.E. the games, and any other Apps would first use the Non-3D vCache CCD, favoring the higher frequency Cores (those are non 3D vCached). But, going with the Cache option, however, is what changed everything. It worked, that way. And that way only. Going with Cache meant that everything I did suddenly was thrown at the 3D vCache Cores by priority; and that's what I am sticking with for now.
#3
Which means, to reiterate, that my previous Benchmark results above were indeed NOT representing the extra 3D vCache thing of the 7900X3D.
#4
I decided to re-run the Benchmarks on my current 'bloated' VAM installation still, because I wanted to compare to my previous Benchmark as a baseline point of reference.
#5
With all this said, there might still be a slight "problem" though.
Despite the fact that I did then run the new Benchmarks below on the 3D vCache CCD, it seems like my actual overall CPU frequencies were lower-than-should-have-been nonetheless. By around 300 to 350Mhz, no less (although it's not the end of the world).
So here's the thing.
1) The Non-3D CCD portion of the CPU (at least the 7900X3D model that I know of) "should", normally (under best conditions) boost itself "Up To" 5.6 Ghz (give or take a few dozen points up or down from that of course, it would fluctuate). But, according to Hardware Info (I checked on my 2nd monitor while doing the tests, not just in VAM mind you but in other games as well) the highest-observed boost clock reached 'only' 5.2 Ghz.
2) The extra 3D vCache CCD portion of the CPU "should" normally boost itself "Up to" 5.0Ghz. But, according to Hardware Info, it stayed at around 4.6 Ghz, occasionally reaching 4.75 Ghz 'only'.
So there you have it, there seems to be a 'lack' of around 300 Mhz of boost potential on both CCDs of the CPU.
However, THAT part of the whole situation to me is beyond my capacity to fix. I don't know why that would be. Maybe the vCore voltage in the BIOS isn't high enough? The temperatures were fine during all my tests (using a Noctua cooler with 2 fans, everything seems fine really). I can't really explain why but - in any case - the end result was a rock-solid stable testing session (no crashes, no lag, no freezes, everything was smooth).
===================================================
Ok so... the Benchmarks now...
@ 1440p / With 3D vCache Cores
@ 1080p / With 3D vCache Cores
Conclusion + Important Points to Consider...
There wasn't - ultimately - any significant gains by actually running the Benchmark on the 3D vCache CCD (even though, on paper, there IS some improvement compared to my first Benchmark, but you get the gist, it's not that good).
BUT...
Something else I noticed, regarding the Benchmark itself and how 'representative' it was (or rather, wasn't) of the said performance gains in VAM.
So here's the thing I noticed...
The "Performance Gains" from the Benchmark itself - after more tests in various other Scenes I have - actually shows barely HALF of the actual 'real' gains I DID get in OTHER Scenes I have. Because, yes, I've seen some Scenes where I know exactly the performance I 'used to' have, where it suddenly increased by a solid
+20 and in some cases
+30 FPS (and that's
WITH using ReShade, or Post-Magic effects that also taxes the GPU).
So, yes, it is true that the performance gains In the Benchmark itself aren't good at all, and you'd think "Well, going X3D sucks for VAM then".
NO... it doesn't, I assure people here, there
ARE substantial gains. I see them now very well. But they are
NOT well-represented (for some reason) in the Benchmark, really it isn't. And I'm not "shitting" on the Benchmark, I'm simply stating what I'm seeing by comparison when I actually load up the Scenes I actually play with when I use VAM (nobody 'plays' with the Benchmark, it's not a universal Scene everyone needs to go in first to then load something Custom, everyone just directly go to their favourite independent Scene and that's it).
So...
With all this being said, one thing I cannot deny is the following...
As you can see in the 2nd Benchmark, the best unconstested gains (at least as far as the Benchmark goes) were - of course - seen only when finally I decreased the resolution from 1440p down to 1080p. Now yes, obviously, that would have been the case by default (even without the 3D vCache stuff). The GPU works less, the CPU works more at lower resolutions; so we're seeing more then what the extra 3D vCache can do.
So there you have it guys.
Quite a rollercoaster of a post but I had to give specific details.
Now, at another date I will probably come back with more results from a 'vanilla', fresh, clean VAM installation. But for now I think the results (under my specific conditions anyway) speak for themselves: The Benchmark isn't exactly representing the actual gains I do see in _other_ Scenes AND my overall performance seemed low originally because - indeed - it wasn't running on the 3D cache CCD.
Alright, my fingers are tired now, phew.
