Precision 7680 & Precision 7780 Owner's Thread

[iieeann]

1 hour ago, Quaro5 said:

Did Dell stop producing these with an IPS screen option? It doesn’t show as available to choose anymore. It was possible a few Precisions ago

What do you mean no IPS? it is always available. Talk to the sales perosn if you can't get what you see on website, Dell is hiding some options from website customization list.

[Aaron44126]

4 hours ago, Quaro5 said:

Did Dell stop producing these with an IPS screen option? It doesn’t show as available to choose anymore. It was possible a few Precisions ago

 

I think all of the displays are IPS at this point, other than the Precision 7670/7680 OLED display, so I guess they don't bother specifically stating it.

[vooze]

4 hours ago, Quaro5 said:

Did Dell stop producing these with an IPS screen option? It doesn’t show as available to choose anymore. It was possible a few Precisions ago

 

7680 + 5680 uses the same 1920x1200 IPS panel, but the 4K is OLED I belive yes.

[Kaj E]

Hi I am new here and intended to post here, but by mistake posted it in the general Dell precision forum, so here it is:

• 
• Member
• •  1

Posted 16 minutes ago

I am new here, have read with interest posts about Dell Precision 7780 which I plan on buying. But I would require some more info before I decide between the RTX 4000 and RTX 5000.

 I have a Precision 7730 with a Quadro P3200 now, but it has become too slow for Photoshop and Adobe Camera Raw with the new AI based features. Adobe tells that the new features will be expanded and that they use tensor cores. One of the new features is AI based denoise. The photographic community on the internet has been posting on how long it takes with different GPUs to do the denoise. Unfortunately, the posts only cover desktop GPUs: https://www.fredmiranda.com/forum/topic/1804640/14 (the table in the middle of the page).

If anyone of you have access to the current versions of Photoshop with ACR or Lightroom Classic, I would greatly appreciate, if you could run a denoise test on this 60 MP reference file: https://drive.google.com/file/d/15UAKmx9IA1k2Zt877ShHeUewCcHquyhu/view , and post the time it takes.

Thank you very much for your consideration.

Kaj

 

[eirop]

On 6/21/2023 at 12:39 PM, Kaj E said:

If anyone of you have access to the current versions of Photoshop with ACR or Lightroom Classic, I would greatly appreciate, if you could run a denoise test on this 60 MP reference file: https://drive.google.com/file/d/15UAKmx9IA1k2Zt877ShHeUewCcHquyhu/view , and post the time it takes.

Thank you very much for your consideration.

Kaj

 

 

I don't have a 7780, but I have a 7770 with 12850HX and RTX A5500. It took 20-21 seconds to process for me. The machine was in "best performance" mode (in Windows settings) but I had some music playing, browser running and another program open. Maybe it could have been marginally faster otherwise. Had around 97% GPU utilization while it was running, so looked like it was maxing out the GPU. This should let you make an educated guess about what an RTX 5000 Ada would do.

 

Edit: this was in Adobe Camera Raw, not LRC.

[presence]

Fresh owner of a 7780, i9 13950 and RTX3500. Opted for 17" due to hopefully a larger/better cooling system.

 

CPU idles at around 55-60C (ambient 25C), but the CPU temperatures jump around like crazy, +20..30C in 0,5 seconds when given small workloads with Solidworks (rotating, manipulating assemblies). +100C is guaranteed every day.

 

I have already replaced the factory TIM with conductonaut, but I don't see much improvement (on my last Precision 7740, it immediately silenced the fans and dropped temperature about 10C).

 

Regarding the twitchy temperatures, perhaps the CPU is not making very good contact with the cooler, since conductonaut layer is very thin? Perhaps I should try PTM7950/7958? Or is the 20-30C temperature peaking normal?

 

I hope Throttlestop is available very soon...

[win32asmguy]

On 7/21/2023 at 3:05 AM, presence said:

Fresh owner of a 7780, i9 13950 and RTX3500. Opted for 17" due to hopefully a larger/better cooling system.

 

CPU idles at around 55-60C (ambient 25C), but the CPU temperatures jump around like crazy, +20..30C in 0,5 seconds when given small workloads with Solidworks (rotating, manipulating assemblies). +100C is guaranteed every day.

 

I have already replaced the factory TIM with conductonaut, but I don't see much improvement (on my last Precision 7740, it immediately silenced the fans and dropped temperature about 10C).

 

Regarding the twitchy temperatures, perhaps the CPU is not making very good contact with the cooler, since conductonaut layer is very thin? Perhaps I should try PTM7950/7958? Or is the 20-30C temperature peaking normal?

 

I hope Throttlestop is available very soon...

 

Welcome to the owners lounge / NBT! I agree that the 17 inch is better for cooling than the 16 inch. Both heatpipes are the larger width and it has a set of exhaust fins directly over the GPU for DOO. With the 7740 it has more exhaust fins so its still superior, but the design team likes thin bezels more than cool temperatures so that is what we get now.

 

I have tested the latest bios and it still cannot unlock undervolting. These chips also boost very high on both single and multi core loads so repaste isn't going to reduce temperatures other than with the sustained long term, default 85W power draw. Even then you have to disable Intel IPF to get consistent performance, and also may need to disable BD Prochot as it can cause throttling to still occur even with good core temperatures.

 

With LM you may need to take it back apart after a few days and reapply if some has soaked into the heatsink, as these are not nickel plated like the AW18 and other systems with factory LM. You can also double-check if any of the thermal pads are obstructing contact pressure by not sitting squarely in their place on the heatsink.

 

I need to order more PTM7950 as it is the best non-LM TIM. Although the 145W 4090 in these can do very well even with stock paste:

 

https://www.3dmark.com/3dm/97517372?

[Kaj E]

On 7/17/2023 at 7:27 PM, eirop said:

 

I don't have a 7780, but I have a 7770 with 12850HX and RTX A5500. It took 20-21 seconds to process for me. The machine was in "best performance" mode (in Windows settings) but I had some music playing, browser running and another program open. Maybe it could have been marginally faster otherwise. Had around 97% GPU utilization while it was running, so looked like it was maxing out the GPU. This should let you make an educated guess about what an RTX 5000 Ada would do.

 

Edit: this was in Adobe Camera Raw, not LRC.

Thank you very much! Very useful info. Sounds good.

[iieeann]

@Kaj E

7780 13950hx RTX5000 is 16sec.  I did not observe the gpu utilization anyway.

 

[Kaj E]

4 hours ago, iieeann said:

@Kaj E

7780 13950hx RTX5000 is 16sec.  I did not observe the gpu utilization anyway.

 

Thank you for the reply. Excellent result, both timewise and image qualitywise. Strange that there was no noticeable GPU utilization. Maybe the fact that just tensor cores were used and that the RTX 5000 is so powerful was the cause.

Based on eirop's reply I compared the specs of A5500 to the Ada gen. RTXs and decided to go for the RTX 4000. I do not need the fastest, just good performance, anything in the 20-25 s range would be fine. Went for a 7780 i9, RTX 4000, 1 +2 TB, 64 (32+32) GB SODIMM. Expecting to receive it next week.

Thank you, Ileeann and eirop!

[iieeann]

11 hours ago, Kaj E said:

Thank you for the reply. Excellent result, both timewise and image qualitywise. Strange that there was no noticeable GPU utilization. Maybe the fact that just tensor cores were used and that the RTX 5000 is so powerful was the cause.

 

 

Sorry what i meant was i did not go to look at gpu utilization whether it is high or low, i just run the ai denoising only.

[yslalan]

I just received my Precision 7680, it shows an extremely poor performance with thermal throttling.

Configuration:

- i9-13950HX

- RTX 3500Ada

I don't know if it's in a similar case as what we saw on Precision 7x70 before. I do replace the thermal paste with Honeywell 7950 thermal pad for both CPU and GPU die.

- CPU throttled at ~110W (100 Celsius degree)

- CPU idle temp at 70 Celsius degree.

 

Anyone is facing a similar issue?

 

CPU-Z looks like: 

 

Temp and package watt looks like when running CB R23 benchmark:

CB R23 score:

[Kaj E]

On 7/27/2023 at 8:54 PM, iieeann said:

 

Sorry what i meant was i did not go to look at gpu utilization whether it is high or low, i just run the ai denoising only.

My bad, I realized what you meant as soon as I had made the post.

Now I am the happy owner of the new Dell Precision 7780. I made the denoise test, and it took 18 seconds with the RTX 4000 to denoise the test file in Adobe camera Raw. The computer is also otherwise very snappy compared to the old Precision 7730. 😁

[PHVM_BR]

7 hours ago, yslalan said:

I just received my Precision 7680, it shows an extremely poor performance with thermal throttling.

Configuration:

- i9-13950HX

- RTX 3500Ada

I don't know if it's in a similar case as what we saw on Precision 7x70 before. I do replace the thermal paste with Honeywell 7950 thermal pad for both CPU and GPU die.

- CPU throttled at ~110W (100 Celsius degree)

- CPU idle temp at 70 Celsius degree.

 

Anyone is facing a similar issue?

 

CPU-Z looks like: 

 

Temp and package watt looks like when running CB R23 benchmark:

CB R23 score:

When PL1 (77W) is reached and stabilized at what temperature?

Your peak power seems too low...

[yslalan]

8 hours ago, PHVM_BR said:

When PL1 (77W) is reached and stabilized at what temperature?

Your peak power seems too low...

Yea, that's also what I am concern about, when it's 77W, it stabilized at ~90 degrees

[yslalan]

I have fresh installed Windows 10 LTSC on it. 

Peak CPU power is still not what I expected, but benchmark somehow increases: CPU-Z single core score is pretty high, is it normal? (The CPU voltage is set correctly by DELL?)

CB R23: 

CPU-Z: 

With CPU voltage offset(-0.125v):

3D Mark TimeSpy:

[yslalan]

Good news here:

48GB*2 5600MHz SODIMM works pretty well on 7680 (running @5200Mhz bcz dual channel)

[PHVM_BR]

13 hours ago, yslalan said:

Yea, that's also what I am concern about, when it's 77W, it stabilized at ~90 degrees

Are these data with PTM 7950?

Was the laptop on a stand or directly on the table?

Room temperature too high?

Was the application done well?

Did you check the temperature difference between the cores?

Check the P-cores and E-cores separately.

If everything is ok the difference between the hottest and the least hot core under full load should be 5 to 7ºC, differentiating the P-cores from the E-cores.

 

I'm questioning because sustaining 77W at ~90ºC is a disappointing result, especially if you're using PTM 7950.

 

With the phase change pad my 7540 sustains 100W at 92-95ºC depending on ambient temperature and on a stand (no fan).

I can't believe that after 4 years dell has downgraded the cooling of their Precision 7000 series...

 

 

[yslalan]

4 hours ago, PHVM_BR said:

Are these data with PTM 7950?

Was the laptop on a stand or directly on the table?

Room temperature too high?

Was the application done well?

1. Yes, indeed, it's with PTM 7950.

2. directly on the table with 2 bottle caps in between.

3. Room temperature is ~22 celsius degree

4. CPU workload is very low before I started the benchmark.

I will check the core temperature later, and update it here.

 

4 hours ago, PHVM_BR said:

I can't believe that after 4 years dell has downgraded the cooling of their Precision 7000 series...

Even without too much change in cooling design compared with Precision 7670, the thermal performance is much worse than 7670. Precision 7670 can have the peak watt at ~157W and keep for a few secs.

[PHVM_BR]

5 hours ago, PHVM_BR said:

Did you check the temperature difference between the cores?

Check the P-cores and E-cores separately.

If everything is ok the difference between the hottest and the least hot core under full load should be 5 to 7ºC, differentiating the P-cores from the E-cores.

You checked?

Perhaps the contact between the heatsink and the CPU is bad.

The gain in thermal headroom with PTM 7950 should be much higher...

 

[yslalan]

23 minutes ago, PHVM_BR said:

You checked?

Short term workload:

 

Long term workload: (reaching 77W:)

[yslalan]

32 minutes ago, PHVM_BR said:

Perhaps the contact between the heatsink and the CPU is bad.

High temp P-core: 1 3 5 7

High temp E-core: 12 13 14 15 | 20 21 22 23

4 E-cores for 1 cluster

Looks like one side of Die is not fully contacted with the heatsink.

 

So, maybe the pressure of heatsink is not balanced?

 

[PHVM_BR]

13 minutes ago, yslalan said:

High temp P-core: 1 3 5 7

High temp E-core: 12 13 14 15 | 20 21 22 23

4 E-cores for 1 cluster

Looks like one side of Die is not fully contacted with the heatsink.

 

So, maybe the pressure of heatsink is not balanced?

Exactly!

You have 4 hot P-cores (P1/P3/P5/P7) and 8 hot E-cores.

One side of your chip has little contact with the heatsink.

You must do the repaste again.

Carefully go down the heatsink with each screw in line with its respective hole, tighten each screw in correct order a half turn at a time until all are secure.

[yslalan]

3 hours ago, PHVM_BR said:

You must do the repaste again.

I have done the repaste, but sadly, nothing improved. 

 

When I opened up the heatsink, the paste is not spread evenly, and the upper part is core 1,3,5,7, which can explain the problem I am facing now.

 

Should I ask Dell support to send me a heatsink replacement?

[PHVM_BR]

20 minutes ago, yslalan said:

Should I ask Dell support to send me a heatsink replacement?

Certainly!

As soon as I saw your results with the PTM 7950 I knew something was wrong.