Precision 7530 & Precision 7730 owner's thread

Hah sadly it is off topic but I'm always up to chit chat via PM or in the off-topic forum. I just picked up a Chevy Colorado ZR2. Looking up recalls the 2017 and 2018 Canyon only has one recall. The 2015 had a few but it was the first year so that's expected.

I see what you're saying about the misleading information, but a blanket statement of "decent cooling" isn't too bad to me, but I see where you're coming from. Sadly there aren't many bigger critics of Dell cooling than me, you can check my 9575, 9570, and 9550 cooling threads in the XPS section, lol.

 

Update on dock issues ---

USB keyboard/mouse stuttering
I've been told that they're working on a firmware update for the system (not the dock) to address this. The update will automatically disable C-states when a dock is connected. (Brute force solution, I suppose — though it's fine, C-states enabled matters more when the system on battery power.) Apparently there are additionally other, less apparent dock issues that are also resolved with C-states turned off.

Power delivery
They've taken in "several" docks and systems from a customer experiencing this issue for further in-house investigation. No word on the solution yet.

 

OK, I humoured you and performed a Cinebench test without the undervolt. Result:


The undervolt gives a neat ~150 point boost. That's like having almost one extra core. Furthermore, the CPU with the undervolt ran at 3.9 GHz (except at the first run, where the PL2 limit was active, allowing the CPU to hit 4.1 GHz, at a temperature of around 85°C), drawing roughly 60 W of power (because of the PL1 power limit) for a ~1270 score, at core and package temperatures of 80°C.

On the other hand, the same CPU at stock voltages ran at around 3.55 GHz throughout the run, also drawing 60 W (PL1 again), but running at 93 °C, and the fans at maximum. But I don't mind the sound, as that's the sound of a lot of air moving through a relatively narrow space, rather than the annoying high-pitched whine of a motor.

I strongly disagree with your blanket statement above—just because these are so-called 'business machines' does not exempt them from the quirks of modern Intel something-Lake hardware, i.e. undervolting by about 100 mV improves performance. Business machines today use the same off-the-shelf hardware that so-called 'consumer machines' use—the days of custom hardware for businesses is long-gone. Most gigantic datacentres use off-the-shelf x86_64 Xeons (or, much more recently, AMD Epycs), of which the 'HEDT' versions (Intel Core i9, AMD Ryzen Threadripper) use the same sockets, similar memory management, similar features (ECC, remote management, etc). Beneath any extra firmware features, the hardware executes the same instructions, and use the same power. Any power-saving, efficiency-boosting measure that works on consumer machines, work on business machines, because there's no real difference between the two.

You are also wrong that businesses don't undervolt, or tweak their software such that their machines run better. I've interned at IT departments in firms that do precisely that—registry, firmware, software tweaks to simultaneously improve performance and efficiency. Electrical power is money—why would businesses willingly dispense with a tried-and-tested method (undervolting) to do what I mentioned above?

Your statement 'The only stable undervolt is no undervolt' is much more misleading than what I've mentioned, because I have provided the data to back my statement up. Undervolts, over the past seven years of Intel CPUs (and now AMD, too), have proven to be stable, provided they are not excessive. Yes, there is testing involved, but it is a day or so of work, for an easy ~10% performance and efficiency improvement over the lifetime of the machine. If one is troubled by it, the undervolt can be removed at any time, by simply deleting a file.

Im really sorry, but your expectations are very, very unrealistic. You can call me out on setting false expectations, but yours are even worse... Let me address your points one by one.

1. As for all your temperature and throttling woes—you have a lemon. As of this post, I have opened Adobe Acrobat, several tabs in Chrome, MATLAB 2018, and other background applications, and here is a crop screenshot of a HWiNFO window proving that the CPU stays well within limits:
2. It is impossible for a notebook of the Dell Precision's dimensions to provide the cooling for the performance you expect, unless you want the notebook to dispense with any form of extendability in terms of ports, and to dispense with the GPU. Physics does not agree with your expectations, and you would be happier for it if you tempered your expectations accordingly. A modern hexacore Intel CPU that runs at 4.4 GHz on all six cores is destined to draw at least 100 W of electrical power (and since CPUs are large resistors, also output an equal amount of heat). You will not be able to dissipate that heat without a notebook 4 cm thick, with half-a-dozen heatpipes, weighing 3 kg or more.
3. Your 6700K at home has a large block of metal (with a volume likely on the order of magnitude of one litre) on top of it, with a 120/140mm axial fan providing ten times the airflow that a notebook centrifugal fan does, with additional fans around the desktop case. Physics again. Your comparison is invalid.
4. You should be blaming Intel for it. Once again, it's physics. The quad-core 7700HQ, at a 2.8 GHz base clock, had a TDP of 45 W. My hexa-core (2 more cores than Kaby Lake) Coffee Lake Xeon E-2176M, at a base clock of 2.7 GHz, is also expected to draw 45 W? Nope, physics does not agree with that, given how similar Kaby and Coffee Lake cores are.
5. Dell's power adaptor is 180 W, because the CPU is rated to draw up to 90 W, the GPU up to around 70W, and the rest for the other components in the system. If the wiring in your house is unable to provide 180 W (which was the power draw of three dim incandescent filament bulbs), then you should change that. Electric ovens (microwave or infrared) draw more than a kilowatt, a modern television draws half a kilowatt, desktops draw half a kilowatt, washing machines and air-conditioners draw multiple kilowatts.

Agreed. 95% of 15/17" notebooks cannot do that. Which is why I'm very satisfied that the 7530 can do what it can do with my Xeon...

 

I don't want to get too involved in the ongoing discussion, but I need to point something out here.

TDP != power consumption

The 45W that Intel lists as TDP is basically Intel saying "Hey manufacturers, you don't need your heatsink to deal with more then 45W. It'll be juuuuuust fiiiiine".
Every 8th hen 'H' processor has a 45W TDP. The 8950HK at 2.9/4.8GHz, but also the 8400H, with 2 cores less and 2.5/4.2GHz clock frequencies. Judging by TDP, they'd both consume the same power? No... The i9 default has a long-term power limit of 60W. This isn't an 8th gen issue either: My latitude with 4th gen i7 had this as well: a 45W TDP processor that, in reality, easily drew well over 55W, after which it would soon thermal throttle. However, once I used XTU to set the power limit of that processor to 45W, it has no thermal issues whatsoever.

So yes, it's Intel that's telling Dell a heatsink for a 45W chip should be enough to achieve intended performance. So either Intel messed up, or the intended performance is way worse than what we're getting out of these Precisions.

The 7730 with i9 that's sitting in front of me managed to maintain it's all-clock turbo for 30 minutes in 28 degrees Celcius ambient, with only a 100mV undervolt.

 

This Intel paper is a few years old but explains the Thermal Design Power philosophy in detail.

There's also an element of hare vs tortoise in the CPU behaviour. Given that hare (Turbo) mode has to run at higher voltage (power is proportional to speed x voltage) to ensure stability, work done per unit of power used is worse than running in tortoise (non-Turbo) mode. I've seen this on my Core M powered Latitude 7370. It came with a Core m5-6Y54 max turbo 2.7GHz) but I then changed the board to one with the m7-6Y75 (max turbo 3.1GHz) because I wanted the 16GB RAM of the latter board. I discovered that the nominally faster CPU was slightly slower running longer tasks as the period running in Turbo mode was significantly reduced as the thermal design limit cut in sooner.

Once upon a time (>10 years ago) faster CPUs tended to be those which production testing revealed could be stable at higher speed for a given voltage. These days I think the chips aren't tested in the same way and the speed cap is whatever the customer is willing to pay for so it's more of a lottery.

John

 

Honestly that's just a lousy design in my opinion. It's not an excuse to put a CPU in a platform if you can't cool it enough to make it operate at full speed.

You're promising something "up to x ghz" on a very optimistic (almost impossible) basis, but in reality some of these machines struggle to keep their base turbo frequency stable with stress test, let alone combined load stress tests.



7730 i9 at what clockspeed and what load program?

Can the i9 at 4.2 or 4.3 with the cpu on extended wPrime 1024 12 thread run, prime95 small FFTs, or OCCT (CPU: OCCT mode).

 

Default clock speed, which is 4.3GHz for all six cores simultaneously, If I recall correctly. Done using the stress test feature in XTU.

I couldn't care less for prime95, tbh. I only stress tested up to half an hour because that's the time the average VHDL synthesis took on my previous laptop... It's a lot quicker on this one.

 

Somebody in this thread asked for USB support for m.2 NVMe drives to get the data if the laptop dies. Here you go:

https://www.anandtech.com/show/13372/mydigitalssd-introduces-m2x-usb-enclosure-for-m2-nvme-ssds

 

This is all backwards thinking in my opinion.

1. No, most businesses do not purchase laptops and then one-by-one undervolt the CPUs. Why? Because it could be different for each and every one, as you stated. There's no way a large business is going to do this. Hell, I'm in a small business and doing this would be a nightmare trying to remember which CPU is what, and then having to do a trade or a reformat.... please. I don't think you really understand how a business works. Sure, they'll customize what they can to make ALL PCs work the way they want. Windows installs, group policy, etc. That's all easy, because it applies uniformly. Uniformly is the key word here.
2. 100mv on a CPU is not a lot of energy. At all. My business would rather keep all PCs on all the time for backing up and remote access than they would to turn them on and off at night. Sure you could WOL, but getting that set up and running for users to use, training the users, diagnosing why some won't WOL, diagnosing why some won't WOL on WiFi, and then attempting to help users on the weekends would be more headache and money than just paying for the electricity.
3. So what happens when a CPU starts becoming unstable? Sure, you can undervolt a CPU and have stability. But how long? Intel specs what they spec for warranty, errata data, and MTBF. Your CPU is running different voltages? Sorry, can't help you. Is that BSOD because of the undervolt? Is that number that's a fraction off because of the undervolt? You don't know, because you're introducing an error into the CPU, because it's running specs it wasn't designed to run. Just because it appears to work, doesn't mean it's designed to work that way.
4. No, I'm not going to blame Intel for it. Intel specs what they spec. Dell is the one taking the chip and putting into a laptop that can't cool it. What you just said was like saying that if I stick the CPU into a small case, slap some cheapie heatsink on it, and then it overheats, I should blame Intel for that. Come on, man, are you serious?
5. A lot of air moving? Hardly. Why don't you actually get a CFM measurement from those fans. the ONE fan in my 10-year old HP moves more air than BOTH fans in the Dell.
6. So now it's my electrical wiring's fault, because it's not the adapter? I'm starting to see a trend with you here - NOTHING is Dell's fault - blame everyone else! I'm starting to wonder if you're really having a debate, or just protecting Dell at all costs...

What you've showed is that the Dell laptop at stock settings cannot perform at the level that the components are built to do. It takes customization which most people won't do in the target audience. And that customization can and will vary depending on the computer you get. Which was my point from the beginning.

 

It wasn't me, but this is awesome. I found a couple on Amazon, but they were over $50 and had mixed reviews.