Quick "Benchmark" of PS, DxO PL6, and Topaz AI Sharpener on new PC

[Karl_Guttag]

About a month ago, I started to put together a new system (see;https://www.dpreview.com/forums/post/66611828 for the result and parts list). My main goal was to speed up Photoshop, DxO DeepPrime, and Topaz AI sharpening with my R5's 45MP files.

My favored "flow is to run cRAW files through DxO PL4 with DeepPrime and export 16-bit tiff to Photoshop. Many (most by count) of my shots are hand-held pictures of propeller planes taken handheld with panning at 1/60th to 1/125th of a second at 200mm to 500mm on an RF100-500. There is a lot of culling of electronic shutter bursts, but then I use Topaz AI sharpener to improve the best shots.

On my old system, it typically took over a minute to go from DxO PL6 to Photoshop and then 1 to 2 minutes for Topaz AI sharpening to do its thing (highly variable based on image content).

My old PC was built in 2015. Key components:

1. CPU i7 4790K @ 4.38GHz
2. 32GB of DDR3 @1600MHz
3. Nvidia Quadro P4000
4. Crucial SATA SSDs (CT500MX)
5. Windows 10

My New PC:

1. CPU i7 13700 @ 5GHz
2. 64GB of DDR5 @5600MHz
3. Nvidia 3080TI
4. M.2 NVME for Program and PS Cache
5. Windows 11

I ran just a single test on one "representative image." What I found:

• DxO PL6 - 45MP R5 cRAW to 16-bit Tiff with DeepPrime XD
  • Old system 68 seconds - GPU utilization ~3%
  • New system 9 seconds (~7.5x faster) - GPU utilization ~90%
• Topaz AI Sharpen (on 16-bit 45mp image in Photoshop)
  • Old system 50 seconds - GPU utilization ~60%
  • New system 15 seconds (3.3x faster) - GPU utilization ~90%

Both systems were at less than 50% memory utilization. I have been on the new system all day, whereas the old system was just fired up, and there was not much else running.

Looking at the performance monitors, it looks like both DxO PL6 and Topaz AI Sharpen primarily highly utilize the GPU, BUT only in the new system.

On the old systems, DxO does most of its work on the CPU (~60% loaded) and barely touches the GPU (only a few percent). Topaz on the old system loads the much less capable GPU by only about 30% on the old system. On the new system, DxO and Topaz are at about 90% GPU utilization while running, while the CPU is at only at 2-3%.

On the surface, it looks like DxO took almost no advantage of the old Nvidia P4000 GPU whereas the Topaz utilized it significantly. With the new 3080TI, they both took full advantage of the graphics card.

In "real used" with more complicated images with many more layers, I would expect the new system with 64GB and much faster memory and M.2 SSDs will help. I was always bumping up against filling up the 32GB on the old system.

It would be interesting to see how much just swapping the graphics card would have done in my old system, at least on DxO and Topaz. Unfortunately, I don't think the power supply of my old system is up to driving the 3080TI, and I don't want to take the time and risk swapping it over.

Programs seem to open up about 2 times faster with the new system (8 seconds on the new vs. 16 seconds for Photoshop on the old system).

 

[Robert Zanatta]

Your new system rocks. Nice setup.

 

[Karl_Guttag]

Your new system rocks. Nice setup.

It's essentially a 9-year newer system except for the graphics card, which is about four years newer (the graphics card is the only thing I upgraded in the old computer).

"General purpose" things are a bit snappier. Photoshop opens about 2x faster, but that is only the difference between 16 and 8 seconds. Opening a CR3 file in Photoshop took about 6 seconds on the old machine and 3 seconds on the new machine. While 2x as fast, we are only talking for a few seconds and not life-changing. I hear some of the new AI features in Photoshop may utilize more CPU and GPU processing.

The one to two minutes for DxO PL6 conversion and Topaz AI sharpening was getting painful when working on a lot of pictures. That was taking a total of 4 minutes or more per photo. Plus, with Topaz, I can now play with the parameters when necessary and see the difference in a few seconds in the preview window. The preview was so slow on the old system that I just went with the suggested values.

My surprise is how little DxO, and to a lesser extent, Topaz used the graphics processor on the old machine, but now they seem only to use the graphics processor and fully leverage it. I would be curious how the 3080ti (which I think is probably overkill) would compare with the more common and affordable 3060 for these tasks, as it would help others in the same boat as I was in.

In this DxO discussion thread (https://feedback.dxo.com/t/which-video-card/23447/26), someone estimated that DxO with DeepPrime that 3060ti would only be about 10% slower (and about 2.5x less expensive today) in a "modern" system. If so, then the 3080ti in my system is overkill. But I didn't want to have a bottleneck anywhere for at least several years.

 

[BobKnDP]

In my ignorance, I'm surprised that the Quadro P4000 was so little used. I don't know its GTX equivalent, but it's from the Pascal (10 series) line. Seems like it should have done better.

No matter. Your new system sounds like it will be almost life changing, if you handle a largish number of images.

Congratulations.

As a side note:

My I9-13900K was running a bit warmer than expected at idle, and I found that the thermal paste didn't appear to be spreading as uniformly as I'd like. With an anti-bending frame, the idle temp dropped about 4-5°C, and the full load went from 85° (before the AIO coolant warmed) to more like 75°. My goal was to avoid thermal throttling in the I9, at any load. (I won't overclock it.)

If your I7-13700K runs a little too warm, it might be worth the $13 investment, plus the effort to install it.

 

[Robert Zanatta]

Yes, surprising that the Quadro wasn't bring used.

I purchased the an LGA 1700 frame and my temperatures dropped about 10C on a stress tests and idle, so for me it's definitely worth it.

 

[Robert Zanatta]

A few people in this forum participated in a Topaz AI benchmark a couple years ago. There was a pretty big improvement in processing speed by higher end GPUs.

 

[Karl_Guttag]

In my ignorance, I'm surprised that the Quadro P4000 was so little used. I don't know its GTX equivalent, but it's from the Pascal (10 series) line. Seems like it should have done better.

It surprised me too, which it why I thought it was worth posting. I assume it is the DxO software that didn't take advantage of it. I could see that the Topaz software was using the P4000, but still by, not as much as it did with the 3080TI.

No matter. Your new system sounds like it will be almost life changing, if you handle a largish number of images.

It is "flow changing," which is what I wanted. I find even at fairly low ISOs with the R5, Adobe Camera RAW ends up very noisy. I can play with the sliders, but it makes the image softer. DxO DeepPrime managages to keep the detail while reducing noise.

I would have to decide whether it was "worth" sending the photo through DxO, as it added a few minutes to the flow. Whenever I didn't use DxO and went straight to ACR, the pictures looked very noisy, so much so that I went back and started over with DxO DeepPrime. On the new PC, it barely takes longer to go through DxO versus ACR so I will always go via DxO.

Congratulations.

As a side note:

My I9-13900K was running a bit warmer than expected at idle, and I found that the thermal paste didn't appear to be spreading as uniformly as I'd like. With an anti-bending frame, the idle temp dropped about 4-5°C, and the full load went from 85° (before the AIO coolant warmed) to more like 75°. My goal was to avoid thermal throttling in the I9, at any load. (I won't overclock it.)

If your I7-13700K runs a little too warm, it might be worth the $13 investment, plus the effort to install it.

Thanks, I will keep that in mind. There is always some risk with opening the system and doing surgery. I watched about 10 videos on how much and in what pattern to apply the thermal past, and they all disagreed with each other regarding the "best" method. In the end, it seems that the paste pattern only affects things by a few degrees, but if you get too much on it, you create a mess that could be a problem if you had to redo something. I went with a large pea size spot.

So far, the highest my CPU has gotten is about 80c, and then for only a short period. I'm using ACDsee to cull and sort photos and run background tasks that can take 20-50+% of the CPU processing. I have about 230,000 images on my computer, and I only got everything on it a couple of days ago. Early on, it was taking 60% of the CPU. It stopped using the CPU after two days, but then it started using about 30-50% of the CPU tonight.

Is there a chart someplace that tells what the temp range should be at various CPU loads for an i7 13700?

 

[BobKnDP]

"Is there a chart someplace that tells what the temp range should be at various CPU loads for an i7 13700?"

Not that I know of.

Throttling occurs at 100° C.

I check using the "Stress CPU" feature in CPU-Z.

 

[Robert Zanatta]

To apply the thermal paste, put a small amount on the IHS and then use something like a tiny spatula to evenly spread it on the IHS. This week guarantee complete coverage, which other method won't.

The maximum processor package temperature depends on what type of processing you are doing, for how long you did it, and the type of cooling you are using. Something like the stress test Prime95 using AVX 2 (or 512 if your processor supports it), and small FFTs, will stress your processor much more than most other tests and the temperature and power draw will be noticeably higher. There probably isn't a consumer application that will stress your processor to this extreme, but if your setup can handle it then anything else will likely be perfectly stable under it maximum load (even for long durations).

 

[Karl_Guttag]

"Is there a chart someplace that tells what the temp range should be at various CPU loads for an i7 13700?"

Not that I know of.

Throttling occurs at 100° C.

I check using the "Stress CPU" feature in CPU-Z.

Thanks, I ran the CPU-Z "stress test" for about 10 minutes, and my CPU topped out at 95c. I'm not playing games on my PC, so usually, my peak processing is only used in short bursts. When ACDsee is doing its sorting thing, the CPU is at 33%, and the temperature hovers in the mid 65c range. At idle, the CPU is at about 40c. I seem to have plenty of cooling of the motherboard with the 6 fans in the front as it stays in the low 30s, and the DDR is about 40c.

I watch some videos on the "frames" for the CPU. Most seem to favor them, with a few dissenters. I'm tempted to give the frames a try as I am curious about how well the thermal paste ended up spreading, and it would be easy to put one in while checking.

 

[BobKnDP]

"Is there a chart someplace that tells what the temp range should be at various CPU loads for an i7 13700?"

Not that I know of.

Throttling occurs at 100° C.

I check using the "Stress CPU" feature in CPU-Z.

Thanks, I ran the CPU-Z "stress test" for about 10 minutes, and my CPU topped out at 95c. I'm not playing games on my PC, so usually, my peak processing is only used in short bursts. When ACDsee is doing its sorting thing, the CPU is at 33%, and the temperature hovers in the mid 65c range. At idle, the CPU is at about 40c. I seem to have plenty of cooling of the motherboard with the 6 fans in the front as it stays in the low 30s, and the DDR is about 40c.

I watch some videos on the "frames" for the CPU. Most seem to favor them, with a few dissenters. I'm tempted to give the frames a try as I am curious about how well the thermal paste ended up spreading, and it would be easy to put one in while checking.

Poking around the CPU socket is often a little intimidating, but to add the frame, you can leave the CPU in place to protect the socket.

It's an easy thing to try, and cheap. I was skeptical, but it seems to have helped significantly. I haven't popped the cooler since to see whether the thermal compound spread better, but I'd bet that it has.

The other change I made was exchanging the fans on the corsair H150i Elite Capellix cooler for 3 of these: XPG . A little expensive (about $25 each at Amazon), and no RGB, but they are rated to move about the same amount of air as the Corsair fans, and they are quieter at full blow. (Not quiet, but I could stand to sit next to the PC, unlike the Corsairs.)

The CPU package temp, as reported by iCue, is about 33° at idle. (Ambient is 20-21C.) With the CPU-Z stress test running, it leaps to about 77C. As the AIO coolant warms, it creeps up to the upper 80s. Without the frame, it went to 95C. Best $13 I've ever put into the PC.

 

[Robert Zanatta]

I upgraded to a 13900K today. At default BIOS settings it ran really hot under full load and was pulling well over 350W. Crazy.

Anyway, I just set the CPU voltage to adaptive with an offset of -0.12 and that dropped max stress wattage by 100 and now it doesn't throttle. The performance is identical in CPUZ and CB R20.

 

[Tai_Ji_51]

am on the fence to order this new custom Birds in flight / Wildlife Photo centric built - no OC

i7-13700K
Asus ProArt Z790 Creator (4 x M.2 slots + 4 SATA 6Gb ports (once device is detected at m.2_4, otherwise 8 SATA ports)
or
Asus Rog Strix Z790-E/F (5/4 x M.2 slots + 4 SATA 6Gb ports)
still have to rethink how I will migrate/manage my image libraries on a new/different config, and decide how I will populate/use these M.2 and SATA slots
64 Gb DDR5 (2x32Gb Kingston Fury Fury 4800)
Noctua NH-D15S
1000W PSU
Fractal Design Torrent (4 SSD slots and 2 x HDD slots)

to replace "old faithfull" built late 2010

i7-3930K 3.2 GHz
Asus X9P79 MoBo
48 Gb DDR3 3200
Nvidia Quadro P4000
2 x 870 EVO + 2 x Raid 0 (2*16 Tb each) - do not like NAS solutions
win10

not so different from yours, apart older cpu

and run

OM Workspace
DxO Photolab 5 Elite (evaluated PL6 but didn't jump into - yet ...)
Lightroom v.12.0.1 - ACR v15
Adobe Bridge 2022 (v.2023 is deadslow - many complaints on forums)
Photoshop v24 !
Nik Collection v5
Topaz DeNoise, Sharpen, Gigapixel, Photo AI, all latest versions
DaVinci Resolve Studio v18.1 - will presumably be happy with Intel UHD Graphics 770 encoders/decoders

Can't wait to see how the Quadro P4000 will behave with i7-13700k / DDR5, expecting the new RTX 4080 to hopefully become more affordable - as is said to perform better than 3090 with much less power drain / less heat.

 

[Karl_Guttag]

Most of my comments are inline:

I would suggest putting all your selections into PCPartPicker. If you have not already, you might want to start with my "final" list (easier to start with something close and edit): https://pcpartpicker.com/list/DzYKbK

am on the fence to order this new custom Birds in flight / Wildlife Photo centric built - no OC

i7-13700K

Probably plenty of processing. Faster processing than the 12900K and less power than then 13900K

Asus ProArt Z790 Creator (4 x M.2 slots + 4 SATA 6Gb ports (once device is detected at m.2_4, otherwise 8 SATA ports)

The big thing this board gives you is 2 Thunderbolt ports and four more SATA Plugs over the Z790F,

I think you may be misunderstanding the M.2 ports. One port is on the CPU side and the other 3 are on the "Chipset side." Of the 3 chipset M.2 slots, two are NVME only, and one is NVME or SATA. In theory, the CPU side port is faster, but in practice, you probably won't notice a difference. NVME is a bit faster than SATA.

I thought about the Creator board but didn't need the Thunderbolt I could see.

or
Asus Rog Strix Z790-E/F (5/4 x M.2 slots + 4 SATA 6Gb ports)
still have to rethink how I will migrate/manage my image libraries on a new/different config, and decide how I will populate/use these M.2 and SATA slots

The Z790E has a bit of a strange M.2 slot where if you use it you get PCI-E 5 NVME but then you lose 8 lanes to the graphics card (at least that is how I understand it). I saw this as a disadvantage of the E model compared to the F, so I went with the F.

64 Gb DDR5 (2x32Gb Kingston Fury Fury 4800)

Probably more than good enough. The g.skill 5600 is jused should be slightly faster but doubtful you will notice. Note 5600 is the fastest the 13700K will run without overclocking. The g.skill ripsaw is also shorter height which might be important if you go with a large dual fan Air Cooled.

Noctua NH-D15S

I would think about a dual-fan model. Either the 12A (fits better and almost as well as the 15D for cooling) or 15D if you are staying with air cooling.

1000W PSU

You probably want a bigger power supply if you go with the 4090

Fractal Design Torrent (4 SSD slots and 2 x HDD slots)

I don't know the case, but it looks a bit unusual with bottom rather than top fans. Gamer's Nexus liked it back in 2021

to replace "old faithfull" built late 2010

i7-3930K 3.2 GHz
Asus X9P79 MoBo
48 Gb DDR3 3200
Nvidia Quadro P4000
2 x 870 EVO + 2 x Raid 0 (2*16 Tb each) - do not like NAS solutions
win10

not so different from yours, apart older cpu

and run

OM Workspace
DxO Photolab 5 Elite (evaluated PL6 but didn't jump into - yet ...)
Lightroom v.12.0.1 - ACR v15
Adobe Bridge 2022 (v.2023 is deadslow - many complaints on forums)
Photoshop v24 !
Nik Collection v5
Topaz DeNoise, Sharpen, Gigapixel, Photo AI, all latest versions
DaVinci Resolve Studio v18.1 - will presumably be happy with Intel UHD Graphics 770 encoders/decoders

Can't wait to see how the Quadro P4000 will behave with i7-13700k / DDR5, expecting the new RTX 4080 to hopefully become more affordable - as is said to perform better than 3090 with much less power drain / less heat.

Based on my uses of the Quadro P4000 in my old systems, I would not expect it to help that much. On my old system DxO didn't use the P4000 for almost anything and Topaz seem to work better much better with my new graphics card (RTX 3080TI). The Graphics cards don't seem to affect based-level Photoshop that much (the main CPU, size of the RAM, and disk speed can), but the GPU can have a significant impact on DxO and Topaz (and maybe Photoshop's AI software as it develops). I don't think a 4090 will be a big benefit in "Photoshop" and related editing software over a 3060 to 3080ti as you will be well into diminishing returns.

 

[BobKnDP]

The G.Skill Ripjaws S5 RAM is spec'd with a height of 33 mm. The NH-D15 (with its rear fan), 32mm. Whether there is an actual conflict there, I'm not certain. The NH-U12A may be a better choice than the NH-D15S, as you post. Noctua doesn't specify a clearance, but measurements online say 62 mm.

 

[Karl_Guttag]

The G.Skill Ripjaws S5 RAM is spec'd with a height of 33 mm. The NH-D15 (with its rear fan), 32mm. Whether there is an actual conflict there, I'm not certain. The NH-U12A may be a better choice than the NH-D15S, as you post. Noctua doesn't specify a clearance, but measurements online say 62 mm.

The NH-D15 have some flexibility in how low the fans can go as they are held on with clips. One mm is probably not a problem, but if the RAMs are several millimeters too tall, the fan gets so high that it won't cool the radiator well and might run into the side of the case.

The NH-U12A does not overhang the DDR ram as there is a single thick radiator rather than two radiators with the D15. Another option might be the D15s (single fan), but likely the U12A is a better cooling option.

I was going with the D15, but toward the end, I switched to the Artic II 360 CPU cooler on the roof of the D5000 case (I had to mount the fans on top due to the thicker Artic radiator). The NH-U12A does not overhang nearly as much as there is a single thick radiator rather than two radiators with the D15.

 

[Tai_Ji_51]

Most of my comments are inline:

I would suggest putting all your selections into PCPartPicker. If you have not already, you might want to start with my "final" list (easier to start with something close and edit): https://pcpartpicker.com/list/DzYKbK

Thanks for your detailed answer and wise comments, it's not so often that one hears from someone who went from/to similar builts as ours, for similar use / programs.

am on the fence to order this new custom Birds in flight / Wildlife Photo centric built - no OC

i7-13700K

Probably plenty of processing. Faster processing than the 12900K and less power than then 13900K

Apparently MUCH less power/heat than gen12 in typical workflow

Asus ProArt Z790 Creator (4 x M.2 slots + 4 SATA 6Gb ports (once device is detected at m.2_4, otherwise 8 SATA ports)

The big thing this board gives you is 2 Thunderbolt ports and four more SATA Plugs over the Z790F,

This would be my first choice provided my local builder can supply it (belgium)

I think you may be misunderstanding the M.2 ports. One port is on the CPU side and the other 3 are on the "Chipset side." Of the 3 chipset M.2 slots, two are NVME only, and one is NVME or SATA. In theory, the CPU side port is faster, but in practice, you probably won't notice a difference. NVME is a bit faster than SATA.

Thank you Karl, that kind of specs are not always that clear

I thought about the Creator board but didn't need the Thunderbolt I could see.

or
Asus Rog Strix Z790-E/F (5/4 x M.2 slots + 4 SATA 6Gb ports)
still have to rethink how I will migrate/manage my image libraries on a new/different config, and decide how I will populate/use these M.2 and SATA slots

The Z790E has a bit of a strange M.2 slot where if you use it you get PCI-E 5 NVME but then you lose 8 lanes to the graphics card (at least that is how I understand it). I saw this as a disadvantage of the E model compared to the F, so I went with the F.

64 Gb DDR5 (2x32Gb Kingston Fury Fury 4800)

Probably more than good enough. The g.skill 5600 is jused should be slightly faster but doubtful you will notice. Note 5600 is the fastest the 13700K will run without overclocking. The g.skill ripsaw is also shorter height which might be important if you go with a large dual fan Air Cooled.

DD5 4800 idea is based on Puget offers but your idea of the shorter height g.skill ripsaw is a good one

Noctua NH-D15S

I would think about a dual-fan model. Either the 12A (fits better and almost as well as the 15D for cooling) or 15D if you are staying with air cooling.

I also think I would go for the NH-D15S with dual fans, still having opportunity to substitute Noctua 14mm fans to all Fractal case ones in case cpu goes the hot side under sustained load in which case I would study Noctua again as they have several specs / capacities and use in the same size.

1000W PSU

You probably want a bigger power supply if you go with the 4090

RTX 4080 is said to suck much less, but point of detail to keep in mind, thank you.

Fractal Design Torrent (4 SSD slots and 2 x HDD slots)

I don't know the case, but it looks a bit unusual with bottom rather than top fans. Gamer's Nexus liked it back in 2021

had watched that video already but thanks. I also found several interesting ones, with air cooling, out of which

that led me the Torrent way, not too heavy (11kg) while being relatively spacious - comfortable to work in.

to replace "old faithfull" built late 2010

i7-3930K 3.2 GHz
Asus X9P79 MoBo
48 Gb DDR3 3200
Nvidia Quadro P4000
2 x 870 EVO + 2 x Raid 0 (2*16 Tb each) - do not like NAS solutions
win10

not so different from yours, apart older cpu

and run

OM Workspace
DxO Photolab 5 Elite (evaluated PL6 but didn't jump into - yet ...)
Lightroom v.12.0.1 - ACR v15
Adobe Bridge 2022 (v.2023 is deadslow - many complaints on forums)
Photoshop v24 !
Nik Collection v5
Topaz DeNoise, Sharpen, Gigapixel, Photo AI, all latest versions
DaVinci Resolve Studio v18.1 - will presumably be happy with Intel UHD Graphics 770 encoders/decoders

Can't wait to see how the Quadro P4000 will behave with i7-13700k / DDR5, expecting the new RTX 4080 to hopefully become more affordable - as is said to perform better than 3090 with much less power drain / less heat.

Based on my uses of the Quadro P4000 in my old systems, I would not expect it to help that much. On my old system DxO didn't use the P4000 for almost anything and Topaz seem to work better much better with my new graphics card (RTX 3080TI). The Graphics cards don't seem to affect based-level Photoshop that much (the main CPU, size of the RAM, and disk speed can), but the GPU can have a significant impact on DxO and Topaz (and maybe Photoshop's AI software as it develops). I don't think a 4090 will be a big benefit in "Photoshop" and related editing software over a 3060 to 3080ti as you will be well into diminishing returns.

You are right and I don't expect nor need Photoshop to be quicker as I am the slow link in the process but would appreciate a snappier response when culling/comparing 1:1 on my 4k 32" monitors, in DxO which allows me to have a much better idea of my future processed image although I will export "dng with DeepPrime and Optical Corrections only" and process to taste in multi layer mode in Photoshop, save PS file with all layers included to get opportunity to come back to an image ages later - provided backup disks didn't crash indeed

DxO and Topaz def do seem to go GPU road.

I just installed Topaz Video AI trial (nice Black Friday rebate) and tested with 1 or 2 sec Timelapse videos of sudden actions, Orf raws shot at 50fps - mallard "exploding" out of water, or bird of prey about to land on my glove, for instance - 4x, 8x or even 16x "super slow motion" and, woooooooooooof insanely interesting and nice results depending on subject and providing camera didn't pan but is taking ages on a rig like mine so would better forget it for now, perhaps with a 4000 series - future might tell.

Enjoy your outings

 

[Tai_Ji_51]

Thanks to both of you.

Interesting comments and good food for thought

 

[Karl_Guttag]

"Is there a chart someplace that tells what the temp range should be at various CPU loads for an i7 13700?"

Not that I know of.

Throttling occurs at 100° C.

I check using the "Stress CPU" feature in CPU-Z.

Thanks, I ran the CPU-Z "stress test" for about 10 minutes, and my CPU topped out at 95c. I'm not playing games on my PC, so usually, my peak processing is only used in short bursts. When ACDsee is doing its sorting thing, the CPU is at 33%, and the temperature hovers in the mid 65c range. At idle, the CPU is at about 40c. I seem to have plenty of cooling of the motherboard with the 6 fans in the front as it stays in the low 30s, and the DDR is about 40c.

I watch some videos on the "frames" for the CPU. Most seem to favor them, with a few dissenters. I'm tempted to give the frames a try as I am curious about how well the thermal paste ended up spreading, and it would be easy to put one in while checking.

Poking around the CPU socket is often a little intimidating, but to add the frame, you can leave the CPU in place to protect the socket.

I'm going to put the frame in next week. I'm curious to see how the heat paste spread, and the frames seem to have a good reputation for improving the thermal. I've been (perhaps overly) influenced by watching Gamers Nexis, which also recommended the Artic CPU Freezer ii and Artic fans.

I got lucky in that I missed that the Artic Radiator was thicker than most (which is good in terms of removing heat), but I was able to fit it in the D5000 roof by putting the fans on the outer top of the case.

It's an easy thing to try, and cheap. I was skeptical, but it seems to have helped significantly. I haven't popped the cooler since to see whether the thermal compound spread better, but I'd bet that it has.

The other change I made was exchanging the fans on the corsair H150i Elite Capellix cooler for 3 of these: XPG . A little expensive (about $25 each at Amazon), and no RGB, but they are rated to move about the same amount of air as the Corsair fans, and they are quieter at full blow. (Not quiet, but I could stand to sit next to the PC, unlike the Corsairs.)

I have 8 Artic P12 fans in my case right now (and the two Corsair fans that came with the case), and I have 2 more that should be here next week for a total of 10 (3 with the CPU cooler, 6 in the front, and 1 in the rear). The Artic fans were 5 for $35 plus two at $9. They only go to 1800RPM and can't move as much air as the XPG fans, but then I have 10 of them so it is the ant devouring the elephant, plus more fans at lower speeds is hopefully quieter.

I'm going to install the CPU frame when the 2 Artic fans come in.

The CPU package temp, as reported by iCue, is about 33° at idle. (Ambient is 20-21C.) With the CPU-Z stress test running, it leaps to about 77C. As the AIO coolant warms, it creeps up to the upper 80s. Without the frame, it went to 95C. Best $13 I've ever put into the PC.

The CPU temperature is a function of the fan profile. With the Artic Cooler, the fan and pump are common. I have the fans set at 40% at 40C and then ramping to 100% at 70C.

The funny thing is that CPUID HWMonitor says the CPU is at 40C but Armory Crate (ASUS utility) says the CPU is a 36C. When I run the stress test for a while, the CPUID temp levels out at 94c while Amory Crate says only 83C or a significant 11C difference. Now I'm not sure which to believe. [edit} I just tried iCue and it reads within a degree of CPUID at both idle and full load.

 

[BobKnDP]

"Is there a chart someplace that tells what the temp range should be at various CPU loads for an i7 13700?"

Not that I know of.

Throttling occurs at 100° C.

I check using the "Stress CPU" feature in CPU-Z.

Thanks, I ran the CPU-Z "stress test" for about 10 minutes, and my CPU topped out at 95c. I'm not playing games on my PC, so usually, my peak processing is only used in short bursts. When ACDsee is doing its sorting thing, the CPU is at 33%, and the temperature hovers in the mid 65c range. At idle, the CPU is at about 40c. I seem to have plenty of cooling of the motherboard with the 6 fans in the front as it stays in the low 30s, and the DDR is about 40c.

I watch some videos on the "frames" for the CPU. Most seem to favor them, with a few dissenters. I'm tempted to give the frames a try as I am curious about how well the thermal paste ended up spreading, and it would be easy to put one in while checking.

Poking around the CPU socket is often a little intimidating, but to add the frame, you can leave the CPU in place to protect the socket.

I'm going to put the frame in next week. I'm curious to see how the heat paste spread, and the frames seem to have a good reputation for improving the thermal. I've been (perhaps overly) influenced by watching Gamers Nexis, which also recommended the Artic CPU Freezer ii and Artic fans.

I got lucky in that I missed that the Artic Radiator was thicker than most (which is good in terms of removing heat), but I was able to fit it in the D5000 roof by putting the fans on the outer top of the case.

It's an easy thing to try, and cheap. I was skeptical, but it seems to have helped significantly. I haven't popped the cooler since to see whether the thermal compound spread better, but I'd bet that it has.

The other change I made was exchanging the fans on the corsair H150i Elite Capellix cooler for 3 of these: XPG . A little expensive (about $25 each at Amazon), and no RGB, but they are rated to move about the same amount of air as the Corsair fans, and they are quieter at full blow. (Not quiet, but I could stand to sit next to the PC, unlike the Corsairs.)

I have 8 Artic P12 fans in my case right now (and the two Corsair fans that came with the case), and I have 2 more that should be here next week for a total of 10 (3 with the CPU cooler, 6 in the front, and 1 in the rear). The Artic fans were 5 for $35 plus two at $9. They only go to 1800RPM and can't move as much air as the XPG fans, but then I have 10 of them so it is the ant devouring the elephant, plus more fans at lower speeds is hopefully quieter.

I'm going to install the CPU frame when the 2 Artic fans come in.

The CPU package temp, as reported by iCue, is about 33° at idle. (Ambient is 20-21C.) With the CPU-Z stress test running, it leaps to about 77C. As the AIO coolant warms, it creeps up to the upper 80s. Without the frame, it went to 95C. Best $13 I've ever put into the PC.

The CPU temperature is a function of the fan profile. With the Artic Cooler, the fan and pump are common. I have the fans set at 40% at 40C and then ramping to 100% at 70C.

The funny thing is that CPUID HWMonitor says the CPU is at 40C but Armory Crate (ASUS utility) says the CPU is a 36C. When I run the stress test for a while, the CPUID temp levels out at 94c while Amory Crate says only 83C or a significant 11C difference. Now I'm not sure which to believe. [edit} I just tried iCue and it reads within a degree of CPUID at both idle and full load.

I found that Prime95 stresses the CPU more than CPU-Z. With Prime95, the CPU (I9-13900K, at stock) spiked to 100C (throttling).