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

[Mr. Fox]

7 hours ago, chew said:

 

 

well there's 2 templates for AMD. well actually 3 baselines.

 

Better off running A and M operation is dimm or board support dependent with agesa and board. Early in beta testing M would not even post.....

 

A and M.

 

 

 

 

 

 

 

7 hours ago, chew said:

your typical 2 dpc boards are going to look more like this

 

Thank you. My current Ryzen 8000 memory profile is largely based on your work and I really appreciate it. 

 

On another note, the 14900KS that I purchased from @tps3443 is now bare die on a Thermal Grizzly Mycro Pro block. Just finished that a few minutes ago. And, the X870E Taichi just arrived right as I was finishing up that project.

 

[electrosoft]

Sold off some more memory and storage I had along with my 4090 so the "fall sales event" is definitely in its last gasps. 🙂

 

Asus X670E Crosshair Hero was delivered a day early so I'll dig into that later. All I know is that box is CRAZY heavy.

 

If the 9800X3D works out like I want I will end up shifting the SP109 14900KS to the wife's system either as is intact with the Z790i Lightning and TG 8200 sticks and sell off her Z690 D4 + 4000 CL15 Bdie sticks + 12900k or I will just install the 14900KS into her system and sell off the Lightning and her existing 12900K chip and keep the TG sticks to dabble with on the x670e.

 

If I find my Intel setup is still better overall for WoW (definitely not Fallout 76 as even the 7800X3D smoked my 13900KS badly), I'll just return the x670e and sell off the 9800X3D which won't be a problem.

 

After that? Shut down my eBay store from January to October again. 🙂

 

Either way, I'll need to pick up an Xmas holdover card till the 5090. Maybe see if I can find a used 4070 Super or something.

 

 

 

[tps3443]

3 hours ago, Mr. Fox said:

 

Thank you. My current Ryzen 8000 memory profile is largely based on your work and I really appreciate it. 

 

On another note, the 14900KS that I purchased from @tps3443 is not bare die on a Thermal Grizzly Mycro Pro block. Just finished that a few minutes ago. And, the X870E Taichi just arrived right as I was finishing up that project.

 

That 14900KS is bound to rip on direct die. I’ve more or less completely stepped away from delidding CPU’s and water cooling graphics cards. It hurts the value and makes it really hard to sale. 
 

Although, I do think I may need to delid

this golden SP106 (10900K) I might make an exception for that one (Have not ran it yet, but I know how LGA1200 chips love that. 🤣 They rip HARD on direct die. 

[Mr. Fox]

11 minutes ago, tps3443 said:

It hurts the value and makes it really hard to sale. 

I have not seen that. Maybe if you're wanting to target the wrong buyers. I've never expected to pay less or get less for a delidded CPU, but I don't care about what gamerboys want. Some of them also wig out about liquid metal making the etched letters on the IHS unreadable and similarly stupid noob notions.

[electrosoft]

35 minutes ago, tps3443 said:

That 14900KS is bound to rip on direct die. I’ve more or less completely stepped away from delidding CPU’s and water cooling graphics cards. It hurts the value and makes it really hard to sale. 
 

Although, I do think I may need to delid

this golden SP106 (10900K) I might make an exception for that one (Have not ran it yet, but I know how LGA1200 chips love that. 🤣 They rip HARD on direct die. 

 

What doesn't rip harder on DD especially with a little chiller action?  🤣

 

But I agree with you about delidding CPUs unless needed and it is rarely needed for myself since laptop DTRs have pretty much went bye bye.

 

21 minutes ago, Mr. Fox said:

I have not seen that. Maybe if you're wanting to target the wrong buyers. I've never expected to pay less or get less for a delidded CPU, but I don't care about what gamerboys want. Some of them also wig out about liquid metal making the etched letters on the IHS unreadable and similarly stupid noob notions.

 

I always found selling properly delidded and properly sealed relidded CPUs easier to sell and at a slightly higher value than stock off the shelf ones.

 

 

[tps3443]

@Mr. Fox

@electrosoft

 

So when I meant harder to sell, I mean imagine I’m running direct die on a 13900KS, and then I’m contemplating selling it. There’s a whole process to do most buyers want a sealed chip, unless you get lucky but there still work involved either way. I’ve gotta pull the chip clean the die, redo the Liquid Metal, clean the solder off the original IHS and then reseal the CPU with black silicon making sure it is exactly lined up and even, re-latch back in to a socket so it can cure with pressure being applied, let’s not forget I need to run it again to make sure temps are adequate (No hot cores) all of this takes a lot of time and work all just to sell a CPU lol. Usually half a day. And then you want to let it sit in the socket for 24 hours for good cure so IHS doesn’t lift on you causing a gap between die and inner IHS. Plus LM splashes bad in shipping it’s gonna be a mess under the IHS of a delidded cpu that gets shipped if the buyer lifted the IHS again. 
 

Now with GPU’s it’s the exact same thing. Selling waterblocked GPU’s is tough. You’re limiting your market buyers to 10% of gamers who buy GPU’s who have custom loops maybe mess. Then I have to remove the block and re-mount the original cooler may need to replace pads/paste, then reinstall to check temps+function then repackage. 
 

With a stock CPU, I pull the chip wipe the paste and I’m repackaging. With GPU, I just pull GPU and package. 

[Mr. Fox]

7 hours ago, Mr. Fox said:

On another note, the 14900KS that I purchased from @tps3443 is now bare die on a Thermal Grizzly Mycro Pro block. Just finished that a few minutes ago. And, the X870E Taichi just arrived right as I was finishing up that project.

[electrosoft]

2 hours ago, tps3443 said:

@Mr. Fox

@electrosoft

 

So when I meant harder to sell, I mean imagine I’m running direct die on a 13900KS, and then I’m contemplating selling it. There’s a whole process to do most buyers want a sealed chip, unless you get lucky but there still work involved either way. I’ve gotta pull the chip clean the die, redo the Liquid Metal, clean the solder off the original IHS and then reseal the CPU with black silicon making sure it is exactly lined up and even, re-latch back in to a socket so it can cure with pressure being applied, let’s not forget I need to run it again to make sure temps are adequate (No hot cores) all of this takes a lot of time and work all just to sell a CPU lol. Usually half a day. And then you want to let it sit in the socket for 24 hours for good cure so IHS doesn’t lift on you causing a gap between die and inner IHS. Plus LM splashes bad in shipping it’s gonna be a mess under the IHS of a delidded cpu that gets shipped if the buyer lifted the IHS again. 
 

Now with GPU’s it’s the exact same thing. Selling waterblocked GPU’s is tough. You’re limiting your market buyers to 10% of gamers who buy GPU’s who have custom loops maybe mess. Then I have to remove the block and re-mount the original cooler may need to replace pads/paste, then reinstall to check temps+function then repackage. 
 

With a stock CPU, I pull the chip wipe the paste and I’m repackaging. With GPU, I just pull GPU and package. 

 

From your use case, I can see how that would involve extra work. For my use case, I am meticulous as I am resealing for personal use so I take everything to the next level and the CPU is naturally tested over a protracted period of time so when I resell, it is just a lift out, wipe down, package and on its way. I've pried apart some truly nasty liquid metal jobs on CPUs, GPUs and laptops and it's pretty horrendous.

 

I am sure blocking GPUs then having to revert back to their original state can be time consuming, but the absence of liquid metal in the equation has to help.

 

Work is just part of being a competent reseller of your equipment because you wouldn't want that karma/ bad joojoo hit for not properly taking care of your gear and/or making sure to present it in the best light to the next buyer. I break down everything in exacting order and clean everything as much to factory spec as possible and box it up to look as much as new as possible so the buyer will be happy and I'm happy because I couldn't let it go out the door looking like PigPen from Charlie Brown was the last to use it. 🙂

 

 

 

 

[electrosoft]

Crosshair setup for flashing and 7600x testing / recert before slapping in the 9800X3D.....

 

 

7600x = SP115. Not a bad Vid at all.... this is the one that could hit 2200. I'll be selling it once I'm secure in the 9800X3D (or keep it has a backup for any/all future baseline testing/flashing).

 

[win32asmguy]

Got a few parts in today for the Sapphire Rapids build.

 

 

Still waiting on the W7-2495X. It may arrive this week if the ebay seller decided to use USPS instead of UPS. Otherwise it was a holiday weekend so if its delayed that is fine.

 

For now I will try to fit the build into my Antec P7 Neo. It isn't exactly a great case but was one of the few that did not have a window and also had a built in 5.25 drive bay. The same company here in Colorado that sells Clevo laptops with open firmware also has a line of custom desktop cases they manufacture in Denver. They have a model that can fit EATX called Nebula49 which should fit the board other than maybe needing to remove a fan header and not use the custom CPU heatsink duct if it interferes with the Noctua U14S.

[electrosoft]

Trying to figure out what's going on here but I'm leaning towards this 9800X3D being problematic:

 

A-die = G.skill 2x32GB kit F5-6000J3040G32GX2-TZ5RK - I used these for over a year in my MSI x670e Carbon with the 7600x, 7800x3d and 7950x3d with numerous BIOS revisions and no problems along with testing in several motherboards.

 

M-die = Hynix Green generic Gen 1 2x16GB greenies good for up to ~6600

 

Scenarios:

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

Asus B650 Prime w/ 7600x and M-die no problem. Ran for months tuned 6000 and tested up to 6400

Updated BIOS to support 9800X3D

9800X3D only ran up to 5400 same M-die setup. Anything higher and it sometimes wouldn't post and when it did post it was sporadic in nature. Anything past ~5200 failed TM5 even with lose timings at 1.35v->1.45v

Swapped back in 7600x no problems running like before after some adjustments with m-die 6000

Swapped back in 9800x used new adjustments same problems

7600x on older bios could do 2200 w/ GPU installed

9800X3D couldn't go past 2000

A-die flat out wouldn't post but the B650 has no error codes.

Swap back in 7600x one more time and test with M and A, no problem.

---

Ordered this x670e Hero

Setup with 7600x for flashing with A-die at just jedec for 9800X3D support

7600x SP=115

Bios flashed fine and 7600x and A-die worked fine at jedec

Swap in 9800X3D = Code 15. Each boot is red light or yellow light or sometimes both

Swap in M-die sticks now it boots

Swapped back in A-die no post

Reflash BIOS now get right SP for 9800X3D = SP113...meh

Same problems

Swap back in 7600x and it works with both M-die and A-die

Set old timings on the M-die to test and it works fine. Did some testing at 2133/6400 for fun

Swap the 9800X3D back in and fclk again stuck at 2000 no matter what I do and same problems at 5400+ on M-die and yet again A-die won't post with error code 15.

---

 

So that's two Asus boards with the newest BIOS and this 9800X3D that are proving to be problematic with the A-die while the 7600x just shrugs its shoulders and keeps it moving with either set.

 

Think I'm going to return the 9800X3D as a defective chip and get a swap from Newegg or they will refund me because they will have no stock. I can't think of anything else since the 7600x works fine, both sets of memory work fine with it across what I realize now is 5-6 boards.

 

A-die worked fine with 7600x, 7800X3D, 7950X3D tested across a couple of Asrocks 1DPC, X670e carbon x 2 and two Asus boards.

 

The common "fail" point in the end is this 9800X3D chip. It is a no go at all with the A-die and is severely limited even with the M-die.

 

I just generated a return label for Newegg to send it back on Monday. I sense they're going to be out of stock of 9800X3D chips unfortunately. Hopefully they're not and can send me another one.

 

 

 

 

 

 

 

[Mr. Fox]

11 hours ago, electrosoft said:

Trying to figure out what's going on here but I'm leaning towards this 9800X3D being problematic:

 

A-die = G.skill 2x32GB kit F5-6000J3040G32GX2-TZ5RK - I used these for over a year in my MSI x670e Carbon with the 7600x, 7800x3d and 7950x3d with numerous BIOS revisions and no problems along with testing in several motherboards.

 

M-die = Hynix Green generic Gen 1 2x16GB greenies good for up to ~6600

 

Scenarios:

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

Asus B650 Prime w/ 7600x and M-die no problem. Ran for months tuned 6000 and tested up to 6400

Updated BIOS to support 9800X3D

9800X3D only ran up to 5400 same M-die setup. Anything higher and it sometimes wouldn't post and when it did post it was sporadic in nature. Anything past ~5200 failed TM5 even with lose timings at 1.35v->1.45v

Swapped back in 7600x no problems running like before after some adjustments with m-die 6000

Swapped back in 9800x used new adjustments same problems

7600x on older bios could do 2200 w/ GPU installed

9800X3D couldn't go past 2000

A-die flat out wouldn't post but the B650 has no error codes.

Swap back in 7600x one more time and test with M and A, no problem.

---

Ordered this x670e Hero

Setup with 7600x for flashing with A-die at just jedec for 9800X3D support

7600x SP=115

Bios flashed fine and 7600x and A-die worked fine at jedec

Swap in 9800X3D = Code 15. Each boot is red light or yellow light or sometimes both

Swap in M-die sticks now it boots

Swapped back in A-die no post

Reflash BIOS now get right SP for 9800X3D = SP113...meh

Same problems

Swap back in 7600x and it works with both M-die and A-die

Set old timings on the M-die to test and it works fine. Did some testing at 2133/6400 for fun

Swap the 9800X3D back in and fclk again stuck at 2000 no matter what I do and same problems at 5400+ on M-die and yet again A-die won't post with error code 15.

---

 

So that's two Asus boards with the newest BIOS and this 9800X3D that are proving to be problematic with the A-die while the 7600x just shrugs its shoulders and keeps it moving with either set.

 

Think I'm going to return the 9800X3D as a defective chip and get a swap from Newegg or they will refund me because they will have no stock. I can't think of anything else since the 7600x works fine, both sets of memory work fine with it across what I realize now is 5-6 boards.

 

A-die worked fine with 7600x, 7800X3D, 7950X3D tested across a couple of Asrocks 1DPC, X670e carbon x 2 and two Asus boards.

 

The common "fail" point in the end is this 9800X3D chip. It is a no go at all with the A-die and is severely limited even with the M-die.

 

I just generated a return label for Newegg to send it back on Monday. I sense they're going to be out of stock of 9800X3D chips unfortunately. Hopefully they're not and can send me another one.

Sounds like you're having as much fun as i have in the last 24 hours. I was planning to return the Carbon and replace it with the Taichi, but so far I am very unimpressed with the lack of memory stability with the Taichi. It has a lot going for it but that, in and of itself, is a deal-breaker. Not sure why, but it is not stable and has a lot worse latency in comparison. I wasted 16 hours yesterday trying to get it stable. Memory training takes 3 or 4 times longer than the Carbon and has 10ns more latency using the same memory timings. 

 

@cylix the no-lane sharing benefit of the Taichi means nothing. Here is the gotcha: If you use the second x16 slot for anything the GPU runs x8. Not being able to use my x4 PCIe add-card means I ended up losing NVMe storage capacity in spite of  being able to use all M.2 slots without affecting the GPU.  Using the PCIe slot on the Carbon does not cut the GPU bandwidth to x8. That allows me to install up to four NVMe in a x4 PCIe add-in card. So, the Taichi EATX motherboard essentially offers nothing in terms of expansion options than what you get with an ITX motherboard.

 

Very disappointed with the poor overall functionality of this motherboard so far and will most likely return it and keep the Carbon. The opposite of what I expected.

 

Edit: yeah, have tested every available BIOS and this thing is just pure trash. Not sure if it is the one I received or the product in general. Definitely going to RMA/refund on this and go back to the Carbon. It unequivocally sucks at memory overclocking like nothing I have ever seen. Rubbish.

 

Edited December 1 by Mr. Fox
added comments

[D2ultima]

So anybody seen the GN video about NZXT?

[Mr. Fox]

57 minutes ago, D2ultima said:

So anybody seen the GN video about NZXT?

I have not. I will have to check it out.

 

Edit: Wow. Shameful. Seems like shady and scummy is the new standard for all aspects of computer technology. Sad that NZXT is part of it. The trick of the day would be finding a brand that is good enough to not be shady and scummy. I don't think there is one.

 

Edited December 1 by Mr. Fox
icky NZXT

[electrosoft]

4 hours ago, Mr. Fox said:

Sounds like you're having as much fun as i have in the last 24 hours. I was planning to return the Carbon and replace it with the Taichi, but so far I am very unimpressed with the lack of memory stability with the Taichi. It has a lot going for it but that, in and of itself, is a deal-breaker. Not sure why, but it is not stable and has a lot worse latency in comparison. I wasted 16 hours yesterday trying to get it stable. Memory training takes 3 or 4 times longer than the Carbon and has 10ns more latency using the same memory timings. 

 

@cylix the no-lane sharing benefit of the Taichi means nothing. Here is the gotcha: If you use the second x16 slot for anything the GPU runs x8. Not being able to use my x4 PCIe add-card means I ended up losing NVMe storage capacity in spite of  being able to use all M.2 slots without affecting the GPU. 

 

Very disappointed with the poor overall functionality of this motherboard so far and will most likely return it and keep the Carbon. The opposite of what I expected.

 

Edit: yeah, have tested every available BIOS and this thing is just pure trash. Not sure if it is the one I received or the product in general. Definitely going to RMA/refund on this and go back to the Carbon. It unequivocally sucks at memory overclocking like nothing I have ever seen. Rubbish.

 

 

That's a shame considering how great their 1DPC boards have been of late across both AMD and Intel in ITX and mATX form factors. I know my z390 Taichi was a hot mess but that was so long ago and they had really seemed to turn the corner with their recent offerings. You could try another or stick with the Carbon. Like I said before, I had nothing but fantastic results with my x670e Carbon and MSI is my first choice by default now in general. I may end up returning to the fold, who knows? 🙂

 

I'm hoping MSI launches some really nice 5090s again this time as I will most likely target a Founders and MSI again but let's see the lineup first.

 

 

2 hours ago, D2ultima said:

So anybody seen the GN video about NZXT?

 

I watched it around 4am and wow.....I am a huge proponent of caveat emptor and doing your due diligence, but that is shady at its best and the large investment from a known anti-consumer bent on profits and propping up valuation to resell their stake is telling (IE Corsair).

 

 

[D2ultima]

2 hours ago, Mr. Fox said:

I have not. I will have to check it out.

 

Edit: Wow. Shameful. Seems like shady and scummy is the new standard for all aspects of computer technology. Sad that NZXT is part of it. The trick of the day would be finding a brand that is good enough to not be shady and scummy. I don't think there is one.

 

Wanna know the funny part? Most of you won't remember way back then, but when BLD came out I was calling them out for poor representation of benchmarks for the games, where stuff like 1080Ti was showing as lower fps than a 1080, etc.

 

It only took 7.5 years for a large media outlet to check their BLD benchmark discrepancy bullshit.

 

I even went and found my old tweets TO THEM from 2017 asking them to fix it
 

https://x.com/D2ultima/status/872530514590269440 and https://x.com/D2ultima/status/872532130663985153

 

[Mr. Fox]

1 hour ago, electrosoft said:

That's a shame considering how great their 1DPC boards have been of late across both AMD and Intel in ITX and mATX form factors. I know my z390 Taichi was a hot mess but that was so long ago and they had really seemed to turn the corner with their recent offerings. You could try another or stick with the Carbon. Like I said before, I had nothing but fantastic results with my x670e Carbon and MSI is my first choice by default now in general. I may end up returning to the fold, who knows? 🙂

One thing that I really like about it (not enough to forgive the issues) is the in-BIOS control of the lighting. I can make all of the RGB static white and save it in the BIOS. It should be mandatory that all motherboard firmware include that feature. The A-Tuning software also works well, and the per-CCD core ratio manual overclocking works correctly (broken in the MSI firmware).

 

I am not going to try another one. The fact there is so little meaningful engagement in that ASRock X870 thread at oc.net suggests to me that nobody has a good one, or the people buying it are still content with point-and-click DDR5-6000. Neither of those are inspiring.

[electrosoft]

50 minutes ago, Mr. Fox said:

One thing that I really like about it (not enough to forgive the issues) is the in-BIOS control of the lighting. I can make all of the RGB static white and save it in the BIOS. It should be mandatory that all motherboard firmware include that feature. The A-Tuning software also works well, and the per-CCD core ratio manual overclocking works correctly (broken in the MSI firmware).

 

I am not going to try another one. The fact there is so little meaningful engagement in that ASRock X870 thread at oc.net suggests to me that nobody has a good one, or the people buying it are still content with point-and-click DDR5-6000. Neither of those are inspiring.

 

Agreed. I think I commented on the BIOS level basic RGB lighting controls with the mATX RS B650 variants and how nice it was. You can always install their software for more advanced functionality, but for me the basic functionality is perfect and keeps the RGB bloatware off the system (I'm looking at you Asus and Razer).

 

So are you going to stick with the Carbon and hope for a firmware fix or give Asus a whirl?

[Mr. Fox]

2 hours ago, electrosoft said:

 

Agreed. I think I commented on the BIOS level basic RGB lighting controls with the mATX RS B650 variants and how nice it was. You can always install their software for more advanced functionality, but for me the basic functionality is perfect and keeps the RGB bloatware off the system (I'm looking at you Asus and Razer).

 

So are you going to stick with the Carbon and hope for a firmware fix or give Asus a whirl?

I am going to stick with the Carbon. It's solid. MSI always leaves a few things to be desired, but produce an overall decent product. I'd like to steer myself away from ASUS if that is possible. I've given them enough money alread. I've got the Taichi boxed up and ready to submit my RMA. Carbon is back in use.

[Papusan]

 

Yup. It seems no one can do it right.... Also one more reason to hate/avoid the new and shiny😁 No one should need spend two whole days with troubleshooting. No time for this. 

 

Windows 11 24H2 brings the sRGB color space for everyone – When a well-intentioned Windows update causes you two days of trouble

 

And here is more stupidity.... 

 

The (a)RGB hype train continue to the absurd. Yet another new fancy feature meant to lure bling bling lovers to open their wallets. More useless features = New ways to try increase profits. Will this be the new, must have? How much are you willing to pay extra for this type of features? 20 bucks? Or even 50 ?

 

 

The first RGB WiFi antenna is officially launched on Z890 AORUS XTREME AI TOP!

 

 

[Papusan]

On 11/12/2024 at 3:19 AM, Papusan said:

Will Pat relax more now that Intel has reportedly found a solution to the gaming performance issue? You think he will dream sweetly about baby cores? 

 

On 10/23/2024 at 7:38 PM, Papusan said:

Intel's CEO Pat. Gelsinger working hard for Intel......

 

You got the boot. Well earned. From now on Intel will have to dig the hole alone. Hope they throw the baby cores you talked so nice about down that deep dark hole and forget the past to recup momentum with something better than the castrated Arrow Lake processors. And god forbid if Intel go after AMD's CEO for more glued tiles adventures. 

 

 

On 11/6/2024 at 2:11 PM, Mr. Fox said:

Pat on the back and a knife to the gut.

 

 

It seems Pat can't have sweet dreams. Only nightmares. You got fired😁

 

Intel CEO steps down after a disastrous 2024

The announcement was made via press release this morning, calling the move a retirement. Two current executives, David Zinsner and Michelle Johnston Holthaus, will function as co-CEOs while Intel’s board searches for a more permanent replacement.

Intel Announces Retirement of CEO Pat Gelsinger

SANTA CLARA, Calif., Dec. 2, 2024 – Intel Corporation (NASDAQ: INTC) today announced that CEO Pat Gelsinger retired from the company after a distinguished 40-plus-year career and has stepped down from the board of directors, effective Dec. 1, 2024. 

 

 

 

[Clamibot]

Dual GPU is back in style my dudes!

 

I did some testing with heterogeneous multi GPU in games with an RX 6950 XT and an RTX 2080 TI, and the results are really good. Using Lossless Scaling, I used the 6950 XT as my render GPU and the 2080 TI as my framegen GPU. This is the resulting usage graph for both GPUs at ultrawide 1080p (2560x1080) at a mix of high and medium settings tuned to my liking for an optimized but still very good looking list of settings.

 

 

I'm currently at the Mission of San Juan location and my usage on the 6950 XT is 70%, while it is 41% on the 2080 TI. My raw framerate s 110 fps, interpolated to 220 fps.

 

One thing that I really like aboout this approach is the lack of microstuttering. I did not notice any microstuttering using this multi GPU setup to render a game in this manner. Unlike SLI or Crossfire, this rendering pipeline is not to split the frame and have each GPU work on a portion of it. Instead, we render the game on one GPU, then do frame interpolation, upscaling, or a combination of both using the second GPU. This bypasses any timing issues from having multiple GPUs work on the same frame, which eliminates microstuttering. We also get perfect scaling relative to the amount of processing power needed for the interpolation or upscaling as there is no previous frame dependency, only the current frame is needed for either process (plus motion vectors for interpolation, which are already provided). No more wasting processing power!

 

Basically this is SLI/Crossfire without any of the downsides, the only caveat being that you need your raw framerate to be sufficiently high (preferably 100 fps or higher) to get optimal results. Otherwise, your input latency is going to suck and will ruin the experience. I recommend this kind of setup only on ultra high refresh rate monitors where you'll still get good input latency at half the max refresh rate (mine being a 200 Hz monitor, so I have my raw framerate capped at 110 fps).

 

To get this working, install any 2 GPUs of your choice in your system. Make sure you have Windows 10 22H1 or higher installed or this process may not work. Microsoft decided to allow MsHybrid mode to work on desktops since 22H1, but you'll need to perform some registry edits to make it work:

 

1. Open up Regedit to Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4d36e968-e325-11ce-bfc1-08002be10318}.
2. Identify the four digit subfolders that contain your desired GPUs (e.g. by the key DriverDesc inside since this contains the model name, making it really easy to identify). In my case, these happened to be 0001 fo the 6950 XT and 0003 for the 2080 TI (not sure why as I only have 2 GPUs, not 4)
3. Create a new DWORD key inside both four digit folders. Name this key: EnableMsHybrid.Set the value of the key 1 to assign it as the high performance GPU, or set it to a value of 2 to assign it as the power saving GPU.
4. Once you finish step 3, open up Graphics Settings in the Windows Settings app
5. Once you navigate to this panel, you can manually configure the performance GPU (your rendering GPU) and the power saving GPU (your frame interpolation and upsclaing GPU) per program. I think that the performance GPU is always used by default, so configuration is not required, but helps with forcing the system to behave how you want. It's more of a reassurance than anything else.
6. Make sure your monitor is plugged into the power saving GPU and launch Lossless Scaling.
7. Make sure your preferred frame gen GPU is set to the desired GPU
8. Profit! You now have super duper awesome performance from your dual GPU rig with no microstuttering! How does that feel?

This approach to dual GPU rendering in games works regardless of if you have a homogeneous (like required by SLI or Crossfire) or heterogeneous multi GPU setup. Do note that this approach only scales to 2 GPUs under normal circumstances, maybe 3 if you have an SLI/Crossfire setup being used to render your game. SLI/Crossfire will not help with Lossless Scaling as far as I'm aware, but if it does, say hello to quad GPU rendering again! The downside is that you get microstuttering again however. I prefer the heterogeneous dual GPU approach as it allows me to reuse my old hardware to increase performance and has no microstuttering.

[Clamibot]

Uhh, well I'm really glad my house and everything within a few miles didn't vaporize. Apparently my new super 14900KF consumed more than a quadrillion watts of power for a fraction of a second. @Mr. Fox This is some seriously powerful hardware you sold me.

 

 

I don't know if this is better or worse than someone's laptop CPU running at 90,000,000°C. I'm surprised nuclear fusion didn't occur inside their laptop.

[electrosoft]

You hear me talk about X3D and stuttering/chunk when playing that I do not experience on Intel.

 

This is an example here in Ardenweald and this is on the 9800X3D from Jansn Benchmarks. Watch closely for it. This is a mild version out in open areas with minimal player data and higher GPU utilization. It happens more with RT on and especially at 4k.

 

Side note: on a 7900XTX with RT on, the stuttering is just brutal yet it plays so much smoother with it off.

 

 

PvP/Raids with X3D is just brutal...BRUTAL when that cache gives up the ghost but I have to say it doesn't look as bad as the 7800X3D/7950X3D but it is still there. It comes down to when does the massive amount of player data micro stutters end and the X3D chunk begins....

 

 

Remember, this is at 1440p not even 4k on the 9800X3D.

 

Still want to try it myself as I don't know about his memory settings. 9800X3D is on its way back to newegg for an exchange (hopefully).

 

 

[Clamibot]

1 hour ago, electrosoft said:

You hear me talk about X3D and stuttering/chunk when playing that I do not experience on Intel.

 

This is an example here in Ardenweald and this is on the 9800X3D from Jansn Benchmarks. Watch closely for it. This is a mild version out in open areas with minimal player data and higher GPU utilization. It happens more with RT on and especially at 4k.

 

Side note: on a 7900XTX with RT on, the stuttering is just brutal yet it plays so much smoother with it off.

 

 

PvP/Raids with X3D is just brutal...BRUTAL when that cache gives up the ghost but I have to say it doesn't look as bad as the 7800X3D/7950X3D but it is still there. It comes down to when does the massive amount of player data micro stutters end and the X3D chunk begins....

 

 

Remember, this is at 1440p not even 4k on the 9800X3D.

 

Still want to try it myself as I don't know about his memory settings. 9800X3D is on its way back to newegg for an exchange (hopefully).

 

 

 

Solid proof of why high speed, low latency RAM does in fact matter a lot in gaming. Even with the amount of 3D vcache these X3D CPUs have, it's still possible to exhaust that cache many times over in some games. This is exactly the reason Intel CPUs scale so well with fast memory; it helps mititgate the shortcomings of the cache on Intel's CPUs vs AMD's X3D CPUs (in regards to cache size that is).