Dell Precision 7670 & Dell Precision 7770 owner's thread

[win32asmguy]

4 hours ago, Aaron44126 said:

...I don't see what I posted in the quote above as a negative for the cooling system performance; if anything, it's working well enough under light load for the fans to shut off entirely from time to time.  (I'm just complaining about the noise that the fans make when they power on...)

 

I also see what you're saying about the "bending marks near the curve", but the guy who posted the Chinese review has the same thing in his system; it may well be just the way it was designed (for better or for worse).

 

I do understand that there's a potential cooling/thermal snafu here and I'm definitely willing to take steps to improve the situation, but I'm not going to do a "stab in the dark"...  I'll still wait and see what other users' systems come back looking like.  For now, even if performance is not necessarily "optimal", it's still the fastest PC that I've ever personally used, so I'm fine to wait for more information to be available before deciding on a course of action.

No worries. If anything it is not as bad an issue as drive slots randomly disappearing or a stability issue.

 

I think the heatsink in your photo and the other review look not quite as uniformly shaped as the one on the service manual site:

 

https://www.dell.com/support/manuals/en-us/precision-17-7770-laptop/precision_7770_sm/removing-the-heat-sink-for-discrete-graphics?guid=guid-de17304b-782a-4ed7-bddd-f2e6f92aaa57&lang=en-us

 

Also, this may sound strange, but I have noticed in the past that there is a marked difference to the initial fan "startup grind/whoosh" depending on the brand of heatsink. Generally I have found that the Sunon models tend to be smoother sounding than the CCI models. Sadly it is somewhat of a lottery in which one you get, and this early in the production phase you cannot even order them as a spare part.

[1610ftw]

38 minutes ago, PHVM_BR said:

It is not expected to reach the performance of X17, GE67/77, etc...these are gaming machines focused on high performance.

This 7770 is expected to deliver consistent good performance, a more reliable machine. Precision has always prized stability, reliability with good performance.

I'm convinced that Aaron's unit has a problem because at multicore full load his CPU can't even reach the value of PL1. The expected would be to reach something between 115 and 157W (PL2) - depending on the cooling capacity - for a few seconds before throttling and staying at 85W (PL1) with a temperature below 100°C.

There are no workstations that he could point to, the first one released is from Dell. This is why I pointed to another Dell unit and an MSI with the most similar hardware and size.

 

Ultimately it will be up to Dell to let him know about what performance can be expected of the 7770 and I doubt they will tell him that he should be happy with what he has got, it definitely does not qualify as good performance.

[Redrider]

First time poster long time Precision user. I was about to purchase the 7770. But will wait. Dell should up front supply testing criteria to show what the machine is capable of. The buyer should not have to uncover these issues. 

 

I would expect for the price of this machine that it would perform close to what the chip manufactures tolerance specs out. But without any up front testing from Dell no one knows. The Chinese thread, mentions there is a lab for testing but its not used for this purpose. The OP calls himself a Guinea pig. I feel the same with the OP of this thread. 

 

I expect more from a Precision not less.  I hope this gets sorted out and fixed. 

 

On another note it also seems that Dells sales department is lacking or perhaps overwhelmed. I've yet to receive a quote or return call. Ill chalk it up to this model rolling out late. Perhaps thats a sign of even Dell having issues with the 7770. 

 

 

 

 

 

 

[Ionising_Radiation]

@Aaron44126, are you using ThrottleStop to monitor core temps and throttling behaviour? This might help narrow down why your CPU is not able to reach its maximum performance, whether it be a power, current, or thermal limit...

[Reciever]

11 hours ago, Aaron44126 said:

 

Eh, I don't mind repasting but liquid metal is a bit out of my comfort zone.  I know that it can be used in laptops but I've heard some bad stories, and since I have no experience with applying that at all, not something that I really want to take a chance with.

 

I'll mention that running the GPU at 130W (without that much of a CPU load) doesn't seem like an issue for me.  I haven't gotten to thermal throttling stage with the GPU.  I'll try to run it a little harder and see how it behaves sometime today.  I don't know why it doesn't boost up to 150W, but I wasn't really expecting it to spend much time at that power level; last year's Precision 7760's seemed to behave similarly with a 115W GPU TDP and 140W boost power level which was rarely/never invoked.

 

————

 

I've noticed the fans cycling on and off periodically while on a light "office work" load and that's annoying (because of the fan ramp-up sound it makes when the fans come back on).  The M6700 did the same thing and I wrote this fan management software mostly just to avoid that.  (M6700 was also very reactive with the fans if the CPU load spiked up a little; 7770 doesn't have that problem, it is pretty slow to react.)  There doesn't seem to be any option for fan control options for the Precision 7X60/7X70 at this time.

 

For now, I am going to stop running with turbo boost disabled (something I previously did when I didn't need high CPU power, to keep fan speeds low), and see if that will keep the fans from powering off so often.  With background jobs moved to the E cores, it seems like the system is happy to continue to operate at a very low fan speed in this configuration.  With turbo boost off I was seeing it hang out around 1000 RPM (and there are no speeds below 1000 RPM really, it just shuts the fans off).  With turbo boost on it is hanging out more around 1200-1300 RPM, which is still too quiet to hear.

 

(Yes, yes, I did have turbo boost enabled when running Cinebench.)

Liquid metal would help reduce the spikes but its more important to check contact pressure first, if the pressure isnt sufficient its not worth the risk. I have destroyed my y510p some time back with LM, but I took 0 precautions and did 0 research. It was a quick fix more so to meet my deadline for a paper I was writing for University and the effects of thermal mediums based on their expected performance. It was that or pay for another term lol.

 

That being said I use liquid metal for my current laptop with a 9700F @ 4.5 and can sustain it iirc around 110w? Its an absolute frankentop though, not really something to be compared to what we have here 🙂

 

Also not to bad mouth Dell but their thermal applications arent the best (as is the case with most brands) this is mostly simply due to economies of scale and introduces too much error if human hands were involved, to which I would say you personally are likely fully aware. Since they are replacing the panel I would just wait it out.

 

My experiences with Dell seem to point more to the typical issues even Grocery stores faced when news of the Pandemic hit, everyone rushed to get inventory before everyone else and Dell being the slow moving giant that they are find it difficult to access to more suppliers that can meet their needs within their price points.

[Aaron44126]

A few updates for today.

 

Thermals.

Nothing yet.  I'm trying a few different avenues to see if I can get information on whether the "bend" that we are seeing in the heatpipe is "normal" or not (without actually opening a support case just yet).

 

Display.

I've been given an August 10 date for availability of the replacement display panel.  I'll be traveling next week so I won't have it replaced then, but hopefully I can get it on the schedule for August 15.

 

Fans.

I've made necessary changes to Dell Fan Management to prevent the fans from powering off.  In "consistency mode", you can set an RPM threshold.

 

If the fan speed falls below this threshold, the program will load the CPU to nudge the fan speed up.  The CPU load thread is very low priority and shouldn't impact anything else running on the system (...though there should not be much running if it wants to run the fans at such a low speed...), and it stops as soon as the fan speed rises above the threshold.  It also won't occur if the system is running on battery power.

 

This effectively prevents the fans from shutting off, at the cost of a little bit of power.  I've found that 1010 RPM is adequate to prevent the fans from shutting off, but this value might vary by unit.  I'll share this update in the DFM thread in a few days, once I've had a chance to polish it up some.  I also want to tweak it to distribute the load between cores more randomly.

[TheQuentincc]

What a shame, even my 7550 can substain 100W load on the CPU (it has been repasted)... at least since you have a 12950HX so you can undervolt it if needed.

My advise when applying LM on a laptop : make a huge thermal paste barrier all around the CPU (on the sort of metal ring around the CPU/die), and don't forget the usual Kapton to protect the contact near the CPU die.

 

@Ionising_Radiation on his cinebench screenshot you can see it's thermal throttling on hwinfo

[PHVM_BR]

45 minutes ago, TheQuentincc said:

What a shame, even my 7550 can substain 100W load on the CPU (it has been repasted)... at least since you have a 12950HX so you can undervolt it if needed.

My advise when applying LM on a laptop : make a huge thermal paste barrier all around the CPU (on the sort of metal ring around the CPU/die), and don't forget the usual Kapton to protect the contact near the CPU die.

 

@Ionising_Radiation on his cinebench screenshot you can see it's thermal throttling on hwinfo

How many Watts sustained in factory default?

My 7540 sustains up to 90W with Thermalright TFX (undervolt off) at ~93°C and ambient at ~21°C. New and with original thermal paste it was 75W at the same temperatures...

Aaron's unit certainly has a problem. His Cinebench score is lower than a Precision 3571 with (6+8)c/20t.

[win32asmguy]

On 8/1/2022 at 6:34 PM, 1610ftw said:

There are no workstations that he could point to, the first one released is from Dell. This is why I pointed to another Dell unit and an MSI with the most similar hardware and size.

 

Ultimately it will be up to Dell to let him know about what performance can be expected of the 7770 and I doubt they will tell him that he should be happy with what he has got, it definitely does not qualify as good performance.

The Thinkpad P16 can reach 22,700 in Cinebench R23 with the 12900HX. The RTX A4500 is about 10%-15% slower in Timespy (~9700 Graphics Score) vs ~11,000 for the Precision 7770 3080Ti. It is also significantly less expensive as the config was 3200usd (including 2x32GB memory, 1TB SSD, 4k IPS) with a coupon. It sounds like the fan behavior may also be better, as there is no on/off cycle issues. It is easily just as quiet as the Dell 7560/7760. There is only two storage slots, but I guess those 8TB drives would work so you could atleast get 16TB in it which is not bad.

[operator]

Good specs for the P16!

I just configured the P16 on the Lenovo site and it seems like 1.5k $ more expensive even with an online coupon compared to my ordered 7670. So even if the P16 is better/faster/more quiet - I am not sure if this "significant less expensive" is really reliable. Maybe one have to contact Lenovo Sales Rep. to get a better price (as it is usual from Dell).

[PHVM_BR]

2 hours ago, win32asmguy said:

The Thinkpad P16 can reach 22,700 in Cinebench R23 with the 12900HX. The RTX A4500 is about 10%-15% slower in Timespy (~9700 Graphics Score) vs ~11,000 for the Precision 7770 3080Ti. It is also significantly less expensive as the config was 3200usd (including 2x32GB memory, 1TB SSD, 4k IPS) with a coupon. It sounds like the fan behavior may also be better, as there is no on/off cycle issues. It is easily just as quiet as the Dell 7560/7760. There is only two storage slots, but I guess those 8TB drives would work so you could atleast get 16TB in it which is not bad.

where is the information from?

[win32asmguy]

39 minutes ago, operator said:

Good specs for the P16!

I just configured the P16 on the Lenovo site and it seems like 1.5k $ more expensive even with an online coupon compared to my ordered 7670. So even if the P16 is better/faster/more quiet - I am not sure if this "significant less expensive" is really reliable. Maybe one have to contact Lenovo Sales Rep. to get a better price (as it is usual from Dell).

Yeah, maybe the coupon I used was not meant to be applied to that model. Its odd because the similar configured 7670 on the Dell US site is about 2000usd more expensive.

 

41 minutes ago, PHVM_BR said:

where is the information from?

My own testing.

[Aaron44126]

Haven't had time to do any more performance testing.  I have had a "fun" experience getting my display panel replaced...

 

Got word on Friday (8/5) that my replacement display panel is in and they want to do the replacement on Monday (8/8) — today.  But, I'm traveling this week so I'm not at home for the tech to stop by and replace the panel.  I asked to delay by one week so that I can have the panel replaced next Monday (8/15) when I am back.  But... they're only allowed to hold onto the parts for five days, so the longest that I could delay for would be until Friday (8/12) and then they have to send the part back.  Which means, I have to basically cancel the service order and put a new one in.

 

In the meantime, my display panel will be shipped around between warehouses unnecessarily.  Hopefully, it's not the only spare one that they have on hand now and there won't be any delay with my second service order.

 

The timing was perfectly bad.  If they had the display panel ready one day earlier then I could have had it replaced before I left.  If they had it ready one day later then I could have delayed the service until Monday (8/15).  Not really Dell's fault (though a slightly longer delay period where they could hold onto the part for cases like this would be welcome...).

 

Still seriously considering a replacement of the heatsink assembly but I would still like to see what some other ones look like "in the wild".  I do have my 7770 with me so I might find some time to do some testing while traveling.  I want to try an extended load on the GPU and see if it has any trouble sustaining 130W, and a CPU/GPU combo load to see if I can replicate what showed up in the Chinese review (55W cap on GPU).

[PHVM_BR]

4 hours ago, win32asmguy said:

Yeah, maybe the coupon I used was not meant to be applied to that model. Its odd because the similar configured 7670 on the Dell US site is about 2000usd more expensive.

 

My own testing.

Very good!

For this score the P16 must have ~110W sustained on the CBR23?

What is the TGP of the A4500? 115+15W?

Are the temps good? 

[win32asmguy]

17 hours ago, PHVM_BR said:

Very good!

For this score the P16 must have ~110W sustained on the CBR23?

What is the TGP of the A4500? 115+15W?

Are the temps good? 

It can burst as high as 157W but is still thermally limited. Sustained it can only maintain 85W. There is 12C core temperature deltas so a repaste could help.

 

The A4500 should be 115W with 15W of dynamic boost.

 

The Precision does have a superior hardware configuration for my use case as it allows running the internal display and both TB4 ports driven by the Intel GPU. On the P16 every video output is directly connected to the Nvidia GPU, other than the Internal display which has a MUX option in the bios. Sadly this information can be hard to discover on new models.

 

I think the Precision can also run up to 64GB memory at 4800MT, but the Lenovo forces any memory configuration to 4000MT even with two memory slots unpopulated. Maybe some higher performance DDR5 SODIMMS can be used, at least, to reduce latency and get it closer to where we were at with high performance DDR4 SODIMMS.

[PHVM_BR]

1 hour ago, win32asmguy said:

It can burst as high as 157W but is still thermally limited. Sustained it can only maintain 85W. There is 12C core temperature deltas so a repaste could help.

 

The A4500 should be 115W with 15W of dynamic boost.

This is exactly what is to be expected for the Precision 7770. 157W starting and 85W sustained.

 

I intend to change my Precision by the end of this year and I liked the P16. My preference would be for the 7670, but this CAMM module and in the future only 2 SODIMM modules, will probably make me go to the competition.

[Redrider]

7 minutes ago, PHVM_BR said:

This is exactly what is to be expected for the Precision 7770. 157W starting and 85W sustained.

 

I intend to change my Precision by the end of this year and I liked the P16. My preference would be for the 7670, but this CAMM module and in the future only 2 SODIMM modules, will probably make me go to the competition.

Whats the competition? 

[serpro69]

53 minutes ago, Redrider said:

Whats the competition? 

 

Probably the P16. But looking at the price tag, it's not that "competitive" really :D 

[Redrider]

46 minutes ago, serpro69 said:

 

Probably the P16. But looking at the price tag, it's not that "competitive" really :D 

Perhaps this requires its own thread. I've waited 4 years and the 7770 was the one I was going for.  With the OPs current performance and the China input Ill need to wait. 

 

[Conan11]

On 8/8/2022 at 11:16 AM, win32asmguy said:

It sounds like the fan behavior may also be better, as there is no on/off cycle issues. It is easily just as quiet as the Dell 7560/7760.

Could you write more about it? How loud are the fan at the lowest (but not off) speed? I guess it would be really useful if you could start a P16 thread, like Aaron did for 7770.

[win32asmguy]

1 hour ago, Conan11 said:

Could you write more about it? How loud are the fan at the lowest (but not off) speed? I guess it would be really useful if you could start a P16 thread, like Aaron did for 7770.

I added a thread. Sorry I forgot to put in observations about fan noise.

 

 

[Aaron44126]

BIOS update 1.5.6.

https://www.dell.com/support/home/en-us/drivers/driversdetails?driverid=D0RFG

 

Fixes:
- Firmware updates to address security vulnerabilities including (Common Vulnerabilities and Exposures - CVE) such as CVE-2022-0778 and CVE-2022-32482.

 

Updated without issue.  It did have the scary "black screen with no activity" for a minute or so after the firmware update completed and before the laptop started the normal boot process.  (I'm sort of used to this with Dell firmware updates these days.)

[jeamn]

@Aaron44126 Have you used it much on battery yet?  How is battery life with the efficiency cores?  Is it much different than your 7560?

[Aaron44126]

6 hours ago, jeamn said:

@Aaron44126 Have you used it much on battery yet?  How is battery life with the efficiency cores?  Is it much different than your 7560?

 

With screen brightness at 90%, Optimus on, 10+ apps open, and no particular effort to minimize power use (other than using Process Lasso to make sure background tasks stay on E cores as described earlier), I am getting about four hours of battery life.  I'm sure that it could do much better with lower screen brightness and some tuning, but I rarely use the system away from power, so four hours is totally fine with me.

[serpro69]

12 hours ago, Aaron44126 said:

With screen brightness at 90%, Optimus on, 10+ apps open, and no particular effort to minimize power use (other than using Process Lasso to make sure background tasks stay on E cores as described earlier), I am getting about four hours of battery life.  I'm sure that it could do much better with lower screen brightness and some tuning, but I rarely use the system away from power, so four hours is totally fine with me.

 

This actually sounds like a good result for such a beast w/o any optimizations. I thought it would be less.