*Official Benchmark Thread* - Post it here or it didn't happen :D

[tps3443]

Oddly enough, I was trying @Mr. Fox 8000c34 bios settings, and CPU#1 could not run the 1.480V VDD, it required 1.500 VDD to even post in bios.  Well CPU#2 works fine with 1.480 VDD. So IMC strength is definitely there, and seems to factor in your ram ability as well. 

[electrosoft]

2 hours ago, tps3443 said:

Oddly enough, I was trying @Mr. Fox 8000c34 bios settings, and CPU#1 could not run the 1.480V VDD, it required 1.500 VDD to even post in bios.  Well CPU#2 works fine with 1.480 VDD. So IMC strength is definitely there, and seems to factor in your ram ability as well. 

 

Loosen the timings and check 8000->8400 with TM5 Extreme and/or Karhu to see which one can truly stretch its wings the best unhindered by memory timing constraints but IMC or potentially MB constraints. See if 8200 can pass and if so maybe one of them can do an 8400 run.

 

Once you see how high you can fly out the gate, then go back and start tightening up profiles.

 

 

 

 

 

 

 

 

[electrosoft]

15 hours ago, Talon said:

 

I will hopefully be grabbing one day 1 LOL.

 

I swear in my next house I need an entire basement as a man cave for all of this hardware I've been collecting over the years. I want to build a rack of test benches. Setup a nice home gym in one of the corners, a coffee bar and some nice lazy boys in the other corner with a big home theater setup. 

 

 

I am fully expecting you to snag one ASAP! 🤣

 

I did the exact opposite and we downsized from a monster sized house with a huge basement to a double wide since it is only the two of us now and the one thing I miss the most is the massive basement which was all mine and where I had multiple tables set up everywhere with room to spare.

 

I had an "E" configuration with five 8' tables and closet space galore with a Vax 11/780 in the other corner I had picked up from the University of Penn they were tossing out back in the day. True fact, when we sold the house we left it in the basement because that thing was monstrous in size!

 

It is the only reason I would entertain buying a small house with a same sized basement included is to set up my man cave again.

 

 

[Papusan]

On 9/1/2026 at 3:18 PM, Mr. Fox said:

Absolutely, 100% greed motivated.

Nvidia app update fails to block unofficial DLSS multi-frame generation mod on RTX 40 series gaming GPUs

 

Modders discover unlock for DLSS Multi Frame Generation on RTX 40, NVIDIA blocks it, modders immediately unlock it again

 

 

Even 4K is all too much for $5000 graphics cards

 

This Still Feels Like Tech Demo Territory

The performance we're seeing here is pretty much in tech demo territory. Sure, if you have a $5,000 GPU, you can get a playable experience at 4K and 1440p with DLSS 5 enabled. Outside of that, it's a bleak scenario. The RTX 5080 struggles at 4K, while the RTX 5070 is only borderline usable at 1440p.

 

Whether the visual improvement is substantial enough to justify the enormous performance cost is the question that still needs an answer, but for now we've established just how expensive DLSS 5 is to run in current-gen hardware.

 

xx60 series cards is almost useless with modern tech. Same for xx70 and xx80. 

 

 

Nice.......

Microsoft limits its cloud service: Game Pass Ultimate will now only allow you to play 15 hours per month on Xbox Cloud.

Microsoft has shot itself in the foot by announcing that it will be implementing time limits on Xbox Cloud Gaming. Specifically, starting in November of this year, cloud gaming will no longer be unlimited for subscribers

 

On 9/3/2026 at 1:11 AM, Papusan said:

Asus is holding back the 5090 for now. So same sky high prices. But 5080 Astral come in more like $2100 MSRP now. The much cheaper 5080 TUF at $1900. Nvidia offer still the old MSRP.

 

How can people in their right mind pay over $2000 for a 5080? The scary thing is that they do😐

ASUS and Galax Raise GeForce RTX 50-Series Prices by Up to $74

 

Also MSI want every dime out of their graphics cards. $14 USD for every card helps on the bottom line. 

 

MSI RTX 5060 Ti 16 GB: +100 Yuan or US$14

 

MSI, GIGABYTE And Inno3D Jack Up RTX 50 And RX 9000 Prices Again, With RTX 5070 Buyers Facing A $90 Hit

 

The RTX 5060 Ti 16 GB is the least affected GPU, but the GPU has already seen prices soar dramatically in the last month, reaching $800 or more in most regions. So, while a mere $14 price increase might not seem significant, it's going to make the situation even worse. As for GIGABYTE, the company is increasing the price of the RTX 5060 Ti 16 GB as well, while other models are reportedly untouched.

 

 

[Mr. Fox]

14 minutes ago, Papusan said:

Nvidia app update fails to block unofficial DLSS multi-frame generation mod on RTX 40 series gaming GPUs

 

Modders discover unlock for DLSS Multi Frame Generation on RTX 40, NVIDIA blocks it, modders immediately unlock it again

 

Even 4K is all too much for $5000 graphics cards

 

This Still Feels Like Tech Demo Territory

The performance we're seeing here is pretty much in tech demo territory. Sure, if you have a $5,000 GPU, you can get a playable experience at 4K and 1440p with DLSS 5 enabled. Outside of that, it's a bleak scenario. The RTX 5080 struggles at 4K, while the RTX 5070 is only borderline usable at 1440p.

 

xx60 series cards is almost useless with modern tech. Same for xx70 and xx80. 

 

 

The leaders of the computer tech industry are bad people, with bad intentions, that have no honor and deserve no respect from any of us. The only thing any of them deserve is misfortune and tragedy, and I hope that both are visited upon them corporately and personally.

[tps3443]

On 9/4/2026 at 12:46 PM, electrosoft said:

 

I am fully expecting you to snag one ASAP! 🤣

 

I did the exact opposite and we downsized from a monster sized house with a huge basement to a double wide since it is only the two of us now and the one thing I miss the most is the massive basement which was all mine and where I had multiple tables set up everywhere with room to spare.

 

I had an "E" configuration with five 8' tables and closet space galore with a Vax 11/780 in the other corner I had picked up from the University of Penn they were tossing out back in the day. True fact, when we sold the house we left it in the basement because that thing was monstrous in size!

 

It is the only reason I would entertain buying a small house with a same sized basement included is to set up my man cave again.

 

 

CPU#2 might be the keeper for me. It’s only 2167fclk, but the memory controller easily stabilizes 6400c28 (Fully tuned timings) 1:1 with 1.480v vdd. I maybe can drop to even lower VDD voltage. I ran TM5 for several hours, ram temps were 51-52c and it still held up even with 65k tREFi. (I was using a fan on the memory)
 

I’m going to test CPU#1 again, but I don’t think it’ll be able to do the same thing with 6400c28 1:1 so easily. Even though it has a 2200fclk. 
 

I will move to some 1:2 testing soon. It seems like either chip will post and run 8000c34, but I haven’t stability tested it. CPU#2 can run it easier. Which makes me lean towards it being the better choice. 
 

I haven’t tested any kind of core overlocking. I’m just not really any good at it, so I’ve stayed away from it lol. But I suppose it’s time to dig in. I’ve learned a lot about ram tuning on them though. Just not really sure what to do or start with on core clocks or CCD clocks and voltage etc, or load line etc. 
 

6400 1:1 c28 performs STRONG. Just a little bit slower than 8000c34. 

[Mr. Fox]

I just tested some of the Vallhalla PC Bifrost Alloy thermal paste (company owned by drajav at overclock.net) and it is a good thermal paste. I tested using the SFF living room build that has a 360 AIO.

 

https://www.valhallapc.com/products/bifrost-alloy-thermal-paste

 

 

 

[tps3443]

3 minutes ago, Mr. Fox said:

I just tested some of the Vallhalla PC Bifrost Alloy thermal paste (company owned by drajav at overclock.net) and it is a good thermal paste. I tested using the SFF living room build that has a 360 AIO.

 

https://www.valhallapc.com/products/bifrost-alloy-thermal-paste

 

 

 

That’s pretty amazing temps for an SFF! Can you tell me what sort of settings you run here on this SFF PC? Do you run a manual OC with this?

[Mr. Fox]

3 hours ago, tps3443 said:

CPU#2 might be the keeper for me. It’s only 2167fclk, but the memory controller easily stabilizes 6400c28 (Fully tuned timings) 1:1 with 1.480v vdd. I maybe can drop to even lower VDD voltage. I ran TM5 for several hours, ram temps were 51-52c and it still held up even with 65k tREFi. (I was using a fan on the memory)
 

I’m going to test CPU#1 again, but I don’t think it’ll be able to do the same thing with 6400c28 1:1 so easily. Even though it has a 2200fclk. 
 

I will move to some 1:2 testing soon. It seems like either chip will post and run 8000c34, but I haven’t stability tested it. CPU#2 can run it easier. Which makes me lean towards it being the better choice. 
 

I haven’t tested any kind of core overlocking. I’m just not really any good at it, so I’ve stayed away from it lol. But I suppose it’s time to dig in. I’ve learned a lot about ram tuning on them though. Just not really sure what to do or start with on core clocks or CCD clocks and voltage etc, or load line etc. 
 

6400 1:1 c28 performs STRONG. Just a little bit slower than 8000c34. 

Unless the core overclocking is worse on the one that runs 2200 FCLK, I would keep that one and send back the one limited to 2167 FCLK. If memory serves me correctly, they are the same SP rating and have a similar V/F curve. The one that cannot run 2200 FCLK would end upl being your weakest link in the system and memory overclocking is the least important performance factor with Ryzen, and even less important with an X3D CPU. Being limited to 2167 tells you on face value that it has a weaker fabric interconnect and is the inferior sample of the two.

[Mr. Fox]

35 minutes ago, tps3443 said:

That’s pretty amazing temps for an SFF! Can you tell me what sort of settings you run here on this SFF PC? Do you run a manual OC with this?

Hybrid OC (or Dynamic Overclock Switcher aka as "DOS" in ASUS terminology). Basically tune your PBO to the best stable undervolt that gives the highest low core use clock speed for gaming (which is the only thing PBO is actually good for) and them set your per CCD manual overclock and VCore/VID and power switch threshold.

 

If you followed the videos I shared with you previously to tune your PBO then the tedious part is over and the core overclocking is much easier. You will want to set LLC Level 6 or 7. LLC does not apply to PBO, only the manual OC.

 

For the manual overclock switch I have all cores set at 54x for each CCD and 1.275V VCore and VID voltage. I have the switch threshold set for 120A, so it runs PBO clocks until the CPU pulls 120A, then switches to the manual OC. Using the video I shared with you for PBO per-core tuning, this is where I ended up with my per core PBO negative offset (which may not work well for your dual 3D cache CPU, but you can certainly test and see). I do not use Curve Shaper, only Curve Optimizer. Curve Shaper doesn't really do anything for me other than contribute to freezing while watching YouTube videos and web browsing and tuning Curve Optimizer  properly is already a pain in the a$$ without complicating matters with Curve Shaper.

 

 

For illustration, here are the settings I used with the 9950X in the X870E Apex for the hybrid/DOS daily driver settings. I later increased the switch threshold to 120A instead of using the ASUS recommended value (89A) in the "Prediction" window.

 

 

In case you missed the post with the PBO tuning video, it is posted here. Do the PER CORE tuning (versus per CCD) to maximize your results. Otherwise, per CCD will limit your maximum stable negative offset to your WORST core on each CCD. When you tune the Curve Optimizer offset for the maxium negative offset each core can run stable, all of your cores will boost higher, the CPU will run cooler and be more stable. It's a pain in the butt, but worth the effort if you want the best overall result.

 

Your best cores (that use less voltage) will likely handle less undervolt (negative CO offset) simply because they are already using less voltage by default. If you set all the same on one CCD your best cores will under-perform due to voltage starvation, or your system will be unstable (random lockups in light loads like web browsing and watching YouTube videos).

[Mr. Fox]

56 minutes ago, Mr. Fox said:

Unless the core overclocking is worse on the one that runs 2200 FCLK, I would keep that one and send back the one limited to 2167 FCLK. If memory serves me correctly, they are the same SP rating and have a similar V/F curve. The one that cannot run 2200 FCLK would end upl being your weakest link in the system and memory overclocking is the least important performance factor with Ryzen, and even less important with an X3D CPU. Being limited to 2167 tells you on face value that it has a weaker fabric interconnect and is the inferior sample of the two.

If you want to see the difference 2200 FCLK makes, run Y-Cruncher 2.5B and look at the speed improvement compared to 2167. 

[electrosoft]

Lunacy pricing....

 

5060ti costs more than the 9070xt in the US atm retail pricing....just how much Nvidia doesn't care as consumers swirl the "edge computing" toilet bowl atm....

 

62% faster at 1440p. 71% faster at 4k, costs less yet you know people will keep scarfing up the Nvidia brand.

 

PC Partpicker:

Cheapest 9070xt = $739.99 (~23% over MSRP $599.99 launch price)

Cheapest 5060ti 16gb = $749.99 (~75%+ over MSRP $429.99 launch price)

 

 

-------------------------------------------------------

 

4 hours ago, tps3443 said:

CPU#2 might be the keeper for me. It’s only 2167fclk, but the memory controller easily stabilizes 6400c28 (Fully tuned timings) 1:1 with 1.480v vdd. I maybe can drop to even lower VDD voltage. I ran TM5 for several hours, ram temps were 51-52c and it still held up even with 65k tREFi. (I was using a fan on the memory)
 

I’m going to test CPU#1 again, but I don’t think it’ll be able to do the same thing with 6400c28 1:1 so easily. Even though it has a 2200fclk. 
 

I will move to some 1:2 testing soon. It seems like either chip will post and run 8000c34, but I haven’t stability tested it. CPU#2 can run it easier. Which makes me lean towards it being the better choice. 
 

I haven’t tested any kind of core overlocking. I’m just not really any good at it, so I’ve stayed away from it lol. But I suppose it’s time to dig in. I’ve learned a lot about ram tuning on them though. Just not really sure what to do or start with on core clocks or CCD clocks and voltage etc, or load line etc. 
 

6400 1:1 c28 performs STRONG. Just a little bit slower than 8000c34. 

 

Hmmmm.....

 

Well 6400 vs 8000 you're not going to see much difference on average so you're right there. I talked about that over a year or so ago but 8000 is a cake and eat it too frequency to run. I kept going back and forth testing 6400 1:1 c30 and 8000 1:2 c36 and my final conclusion (YMMV of course) was that at worse, 6400 = 8000 and at best it beats on a varying scale so if your system supports it, especially running dual CCD which I don't even run on desktop, I am hard pressed to finalize around 6400.

 

I would settle in and test at 8000 across the board and see which chip is fully stable at 8000c34 and weigh that equally, if not higher than 1:1 testing at this point, but again, that's just me. That is also one reason I've been jumping through so many hoops lately testing memory cooling the last 5-6 months because of the 270k adventures and finally adopting 8000+ on AM5. All setups going forward, after having stuck to 1:1 for over a year, will be 8000+ and those temps are getting juicy.

 

On the other hand, if you're fully locked into wanting to run 2167/6400 tuned to the gills, CPU#2 would be the direction I would go too because it just seems made for 1:1. I'm hoping you get to eat your cake and eat it too and #2 also turns out to be the best at 8000. But man that not hitting 2200 is a (mild) kick in the teeth.

 

The real suck out scenario is if CPU #2 is a beast at 1:1 and CPU #1 is a beast at 1:2. In that scenario, I'm picking #1 all day everyday.

 

As for overclocking, I was pulling teeth and went with per CCD not per core like @Mr. Fox because it was quicker and you know if your weakest core is rock solid, the rest of them are too even with ambient changes. Last time I did per core was my 5800x and that took forever and even then I would sometimes get little instability and crashes. I was content with per CCD but of course you're not extracting the most performance possible so there's a trade off.

 

34 minutes ago, Mr. Fox said:

If you want to see the difference 2200 FCLK makes, run Y-Cruncher 2.5B and look at the speed improvement compared to 2167. 

 

Yeah, I don't know if I could let the 2200 chip go......

[tps3443]

Here is mine all stock no tuning to the cores or anything like that, PBO is auto so I am not sure if thats enabled or not. These Gskills 6800's are fantastic rams. Running 6400c28 with 1.45vdd and  1.40vddq. They destroy my last Corsair rams! 





 

[ssj92]

Hiya fellas. Good to see everyone around here. Thought I'd drop this here as I had a little time to play with my desktop. 

 

BTW am I the only one who doesn't like the new Cinebench 2026? I feel like R23 was better. I like seeing all the threads populate the image. 

 

W7-3465X 3.2Ghz all core stock > 4.3Ghz all core (28-cores). Single thread should be around 5-5.2ghz somewhere (I didn't check to see if it was actually sticking). 

 

 

I have a VM open and some other stuff but doubt it'd drastically change the score. 

 

Also got a Z890+270K Plus to play with next week. 

 

Still no 5000 series gpu though lol