Overclock.net banner
181 - 200 of 9,076 Posts
Xbox Game Bar + Windows Balanced Power plan is needed to run dual CCD X3D chips "properly" Needed to park the High frequency cores without any use of apps or registry editing..
 
  • Rep+
Reactions: LtMatt
HUB results look a bit off when you compare to the results GN are getting in SOTTR.
I saw some large variations in results between a lot of the reviews. Not sure why.
 
@LtMatt Found that on the GN review ~26m mark
 
  • Love
Reactions: LtMatt
I feel like some of these tests aren't being done right, especially when HUB start doing "virtual" 7800x3D tests like *** are you doing that for?

At the moment though I'm waiting for the 7800x3d for my purely gaming PC, for the sake of simplicity.
 
  • Rep+
Reactions: LtMatt
Discussion starter · #189 ·
Some more info on what is needed from Windows to make this work (in addition to balanced power plan and game bar)
Light Font Screenshot Operating system Technology


edit: sorry screenshot is too blurry, here’s the direct link to see those processes running

 
We may be waiting on that Frame Chasers big brain genius for that

his brain is very large

you know you want to use this in one of your videos, buddy
I2hard is my go to for tweaking. Framechaser does some questionable things, like using Borderlands 2 ftom 2012, a game known for heavy Intel favour that very few people play still, while not showing BL 3 which is far more adopted, but gains insane performance from 3D cache...
 
  • Love
Reactions: LtMatt
7800X3D with a budget A620/B650 2dimm motherboard (saves a lot of hazzle tuning ram vs 4dimm, and generally OCs better) is very tempting!
 
  • Rep+
Reactions: LtMatt
Yeah 7800X3D for me. Too much of a hassle with the 7950X3D, and I can put more power into a single CCD and get better clocks without the dual CCD temperature, power and latency hit.
 
  • Rep+
Reactions: Taraquin
Discussion starter · #193 ·
I'm planning on delidding and going direct die, so this snippet from TPU was especially interesting to me:

With my Noctua NH-U14S I've hit the 89°C thermal limit very quickly, which is basically the end of the OC journey, you can only try lowering the CO values even further. Since the 3DV-Cache die is more sensitive to thermals, AMD has decided to lower the temperature limit from 95°C (non 3DV Zen 4 CPUs) to 89°C (3DV Zen 4 CPUs). Together with the thick heatspreader, it is very difficult to keep the CPU temperature down, even though your heatsink itself is barely warm.
I managed to shave around 1°C off the temperature, by lowering the SOC voltage setting, which saves roughly 10 W heat output from the SOC (memory controller etc). These 10 W get consumed by the CPU cores automatically, the processor will boost a little bit higher, until it reaches the temperature limit, you get a bit of extra performance.
Using our Arctic Liquid Freezer II AIO made no significant difference in temperatures, not even in the Ryzen Offset mounting configuration. The problem is not getting rid of the heat at the bottom of the heatsink, but getting the heat from the tiny surface area compute dies to move through the IHS.
 
Discussion starter · #194 ·
These slides from AMD on what the reviewers needed to do and what everything means/does behind the scenes is kind of insane:


AMD's Cache Optimizer driver gives you a degree of control via toggles in the motherboard's UEFI setup program (and possibly in the future through Ryzen Master). By default, you can leave the driver alone to do its thing, or you can take control, and tell it to "prefer cache," where all workload is prioritized to the CCD with the 3D Vertical Cache. The third toggle, "prefer frequency," prioritizes workloads to the second CCD without 3D Vertical Cache, which can sustain higher boost frequencies. Or you can just disable all optimization—this is only useful if you're poking around with the tech, or troubleshooting things.
Font Screenshot Number

 
I'm planning on delidding and going direct die, so this snippet from TPU was especially interesting to me:

With my Noctua NH-U14S I've hit the 89°C thermal limit very quickly, which is basically the end of the OC journey, you can only try lowering the CO values even further. Since the 3DV-Cache die is more sensitive to thermals, AMD has decided to lower the temperature limit from 95°C (non 3DV Zen 4 CPUs) to 89°C (3DV Zen 4 CPUs). Together with the thick heatspreader, it is very difficult to keep the CPU temperature down, even though your heatsink itself is barely warm.
I managed to shave around 1°C off the temperature, by lowering the SOC voltage setting, which saves roughly 10 W heat output from the SOC (memory controller etc). These 10 W get consumed by the CPU cores automatically, the processor will boost a little bit higher, until it reaches the temperature limit, you get a bit of extra performance.
Using our Arctic Liquid Freezer II AIO made no significant difference in temperatures, not even in the Ryzen Offset mounting configuration. The problem is not getting rid of the heat at the bottom of the heatsink, but getting the heat from the tiny surface area compute dies to move through the IHS.
I remember delidding 10 years ago. It's amazing that AMD and Intel still can't get this right.
 
It's too bad TPU didn't test with a 13900KS coupled with fast memory. Anyhow, based on 4K res I think the 13900K is the clear winner given the current street pricing ($529 at Microcenter, $509 if you buy the mobo with it).

 
Discussion starter · #199 ·
It's too bad TPU didn't test with a 13900KS coupled with fast memory. Anyhow, based on 4K res I think the 13900K is the clear winner given the current street pricing ($529 at Microcenter, $509 if you buy the mobo with it).

TPU tested both with ddr5 6000 cl 36. Clearly that leaves a lot to be desired on both sides of the table, maybe a bit of a larger disadvantage for intel. but at 4k the difference between 6000 and 8000 on intel is miniscule as well. i think it might come down to individual games, and your bet on how this will play out with games coming down the pipe. the fact that spiderman with heavy RT outperforms intel shows there's something to be said about peformance in gains with increasingly complex RT implementations. certainly you're right about the current cost factor, but that also leaves out the fact that comparing apples to apples (buying both completely new systems) with intel you're investing in a dead socket, and with amd you have several years and cpu gens ahead of you.
 
TPU tested both with ddr5 6000 cl 36. Clearly that leaves a lot to be desired on both sides of the table, maybe a bit of a larger disadvantage for intel. but at 4k the difference between 6000 and 8000 on intel is miniscule as well. i think it might come down to individual games, and your bet on how this will play out with games coming down the pipe. the fact that spiderman with heavy RT outperforms intel shows there's something to be said about peformance in gains with increasingly complex RT implementations. certainly you're right about the current cost factor, but that also leaves out the fact that comparing apples to apples (buying both completely new systems) with intel you're investing in a dead socket, and with amd you have several years and cpu gens ahead of you.
Well if these x3D CPUs are anything like the 5800x3D, then there is very little to gain from faster ram. My 5800x3D gained nearly nothing going from 3600 XMP to tuned 3800, at least in the SotTR and CP2077 benches (1080p low).

OTOH Intel gains a lot with faster tuned RAM.

The upside is the x3D chips are the easy button if they don't even need fast RAM to perform well.
 
181 - 200 of 9,076 Posts