Precision 7530 & Precision 7730 owner's thread

Perhaps not. However, I've done it for businesses, under clear instructions from the IT department managers. I can only quote from experience.

You really are overstating the risks of undervolting. There is an inherent, low, risk, yes. But there is a risk of a CPU not being stable at any given voltage, because of the inherent uncertainties in CPU manufacturing.

In fact, how do you know CPUs at their stock voltage are stable at all? Because they were 'tested at the factory'? I repeat my point once again that these machines are all mass-produced. The very fact that you've managed to receive a lemon Precision 7530, is undeniable proof that their quality control is hit-or-miss for such mass-produced items. These machines are likely manufactured at the same factories as Dell G7s, Alienwares, XPSes and so on. The only thing we pay more for, is the assurance of reliable post-purchase support. In view of overwhelming evidence to the contrary, you are quoting that 'no, undervolts are extremely risky'. If you don't want to use them, please, carry on. But to make it sound like doom and gloom, is what's misleading. I proved my point with the stats to back it up: an undervolt improves both computational and thermal performance, while also improving electrical efficiency. If you don't want to do it for fear of a small increase in the risk of instability, fine by me. And furthermore, ECC memory exists precisely to correct for the sort of errors you're talking about. I don't want to introduce a logical fallacy, but I can't help but notice that your machine's CPU does not support ECC. Given your winding point about uncaught errors, I'm surprised that you didn't choose either of the Xeons, and the ECC memory to go with it.

At this point, it really looks like you simply don't want to believe the facts... Intel specifies their 8750H on their website as:

.Note the emboldened word.

Further down, the tooltip for 'Processor Base Frequency' says:

Then, we have the tooltip for 'Max Turbo Frequency':

Intel uses the word 'capable', not 'guaranteed'.

Next, the tooltip for Intel® Turbo Boost Technology:

Long story short, Intel specified this CPU to run at 2.2 GHz. Full stop. Given any additional thermal and power headroom, the CPU will increase its frequency as requested, up to a maximum of 4.1 GHz on a single core, 4.1 GHz on 2 cores, 4 GHz on 4 cores, and 3.9 GHz on 6 cores. The 4.1 GHz value is in no way guaranteed. If you want guaranteed 4.1 GHz boost speeds, you should dispense with the Dell Precision and get a MSI desktop-replacement workstation, or dispense with notebooks altogether.

I was speaking arbitrarily, and phrased it as such ('a lot' can be taken to mean anything)—if you take my words literally, that's on you, I'm sorry.

Why do I need to protect Dell? You refuse to understand that an undervolt is a safe, simple, easy, and easily reversible software modification that provides additional performance and efficiency for a negligible risk increase (which you have consistently overstated and over-emphasised). And yes, certainly—-if your wiring is unable to provide a mere 180 W of power, that is a safety and fire risk, and you should seriously consider changing your electrical cabling, as soon as possible.

Dell has locked down the Quadro discrete GPU in this notebook. They've used a 6-bit panel when they could've used any of the nicer 8-bit ones. They could've included a couple more USB-A ports, instead of just two. There are many things wrong with this machine, but thermal headroom is certainly not one of them.

I've shown that the CPU in the Precision can perform beyond 2.7 GHz (base frequency on my Xeon). That is all the components are built to do. The maximum Turbo Boost frequency is, once again, not a guaranteed value. OEMs can choose to prioritise performance, portability, or power efficiency, and they can specify their components as such. As long as the CPU achieves its base frequency without approaching throttle conditions, the PC is built to spec. Any additional headroom is entirely left to the OEMs' discretion.

 

Wait... so you take Intel's words literally, then turn around and chastise me for taking your words literally? You get one or the other. And no, a lot doesn't mean anything. It only means anything when you get called out on it, proven wrong, and then have to figure out a way to twist words and make it still fit your narrative..

Intel says these things because, guess what? It's up to the laptop manufacturer to implement a legitimate cooling solution to allow it to occur. That's not on Intel. I don't get why this is so hard for you to understand.

You also don't seem to understand how throttling works. When the processor gets hot, it cuts the core. It doesn't down clock it, it doesn't back off - it cuts it to the lowest possible. I'm not sure in what world that could be considered as a decent cooling solution. That also does not fit into thermal headroom. PROCHOT is the CPU hitting Tjunc temp, and overheating. That's also taking the processor BELOW that oh-so-touted-by-you base frequency. By the very definition of that, it's not decent cooling with thermal headroom.

You also refuse to understand how Intel won't warranty anything run out of spec.You say it's safe - but it's not. It's running the transistors at a voltage they were not specced by Intel to run at. PERIOD. It doesn't matter if they can, if they should, if they will. It's about SPECs. In engineering or things that lives are dependent on, you are taking a huge risk. Period. Just because you feel it's safe or okay, doesn't mean it is. Just because some software you used to test the computer for one or two or however many hours, doesn't mean it's safe. I'll take the word of the manufacturer, who puts untold numbers of hours on testing to determine their specs and what they'll warranty and what they say is the best to run their chips at.

There's this process that (good or bad) Intel uses called binning. One CPU can run fine under or overvolted. But the next one can't. One machine, you can test. Hundreds, probably not. You know what I can guarantee? That if you run them all like Intel states, unless there's a legitimate problem with the CPU, it will run as Intel states it will run. You can't guarantee me that with undervolting. That's just the fact of the matter, no matter how you feel about it.

If you want to take the risk, then so be it. If you feel it's acceptable, then so be it. But don't sit here and state your feelings or opinions as fact. That's misinformation and misleading. The fact of the matter is that undervolting a CPU is running it out of spec, and that can and does cause errata and irregular behavior in the CPU. If you want it to perform as Intel has stated it will, then run it at Intel's specs. It's up to the laptop manufacturer to design a system around those specs, instead of making excuses and coming up with out-of-spec ways to run the machine.

 

Don't you think all this discussion about tweaking the CPUs or not is a bit off topic. Why not take it to a thread of its own to keep this thread more to 7530 and 7730?

 

This is specifically about the 7530 and what tweaks another user did to claim “decent cooling”. I’m disputing that, and disclosing that tweak and why it’s not for business users.

EDIT: I was on my phone and couldn't elaborate. I think this belongs in this thread because it is directly referencing the 7530 and the throttling that I think users who purchase the laptop are going to see, unless they do some unsupported tweaking of voltage and/or paste replacements, which could cause warranty claim denials and erratic behavior if you don't do it right.

 

I mean, I don't know how else you'd interpret 'a lot'. It's as vague as it gets. Whereas Intel's wording was specifically written to prove that it doesn't guarantee maximum clocks, and it's up to the OEM to implement a thermal solution, which is exactly what I I said.

??? You're just repeating precisely what I said in the last paragraph.

As far as modern throttling is concerned, this is downright false. With Intel SpeedStep, P-states and now Speed Shift, CPUs don't simply cut a core altogether. Core multipliers are stepped down by one and a check is performed to verify that the throttle condition (in this case, high temperatures) is removed. If not, reduce more, check again. CPUs under thermal throttling conditions don't simply drop back down to 800 MHz and then jump back up to ~ 3 GHz, they are stepped up and down; their clocks are quantised.

Intel CPUs have dynamic voltage scaling. I find myself repeating a previous point: this isn't the 1980s anymore, where CPUs are rated to run at one fixed clock and voltage. Modern CPUs and GPUs (whether Intel, AMD, x86_64 or ARMv7) dynamically scale their voltage up and down, as required by the workload and the corresponding clock speed.

I repeat another point, that you are severely overstating the risks of undervolting. It is highly plausible that Intel has overstated the voltage required for their CPUs, to account for sample variance. There is tangible evidence throughout this forum that overvolting carries little risk (except BSODs), and gives very large return (in terms of improved performance, thermals and efficiency) for very little effort. Most users in this thread have already carried it out, and new purchasers are discussing performing it as well.

To use your car analogy: my Asian-market 2008 Civic has specified that the tyres are to be inflated to 220 kPa. I realised that inflating them to 260 kPa improves handling and fuel efficiency, at the (tiny) increased risk of a burst tyre. I continue overinflating my tyres, because the chance is sufficiently low that over the lifetime of the car, I deem that I will likely not have a burst tyre, and would sooner suffer from a puncture.

You appear to have conveniently overlooked that I have proven that even without the undervolt, the Precision runs within spec (i.e. the CPU exceeds base clock, and runs at 3.55 GHz)...

 

Okay. A lot. More than an 80mm fan. More than a 120mm fan. You know, something standard like that. A lot. Not a little. Not a bit more than. A lot.

And yes. It is up to an OEM to implement a solution. The OEM solution currently overheats and throttles.

Yes, I am repeating what you said. Because you're repeating what I said. Yet you still want to blame Intel for it? I don't get it. You seem to understand the words and what they're saying, but if it's applied to the Dell computer that you have, you want to deny that it's Dell's problem.

WRONG. This is what you aren't understanding. You are not getting the difference between PROCHOT thermal throttling and normal "I'm starting to get hot" thermal throttling. One is "HOLY... I'M ON FIRE TURN OFF TURN OFF TURN OFF" and one is "Hey, I'm gonna back this down a bit because I know that I'm adding on other cores and I need to not only share the amount of current coming in, but I also need to realize that the heat is going to go up with more cores being used." It's very important you know the difference. And I didn't say completely cut - It doesn't go to 0, but it sure does go to the lowest it possibly can so that it can cut voltage and current. That's a lot different than throttling down to a lower step on the frequency list.

So, one is thermal throttling to save the life of the CPU and anything around it, and one is determined by a multitude of things including how many cores are in use, how high the load is on those cores, and what state the laptop power mode is in. Battery? Different power profiles. AC? Same thing. Eco mode on AC? Yet another profile. Those have NOTHING to do with what happens when PROCHOT hits and is triggered.

Finally, and the last time I will debate you on this, is that I'm not overstating anything. First, I have not applied a number to this, but merely stated that the risk is there and it's obviously enough of a risk that Intel will also say what I say. If I said 100% of the time you undervolt you will have a problem, then I'm over-stating it. If I said that it's a minimal risk, then I'm understating it, like you. It's a risk, period. Second, you can use synthetic benchmarks all you want - do they actually show you the precision of the calculations and errata data? No? Then you can sit here and say your synthetic benchmark shows a noticeable increase in speed, and yeah, it would be right, because it would finish faster. But is the accuracy and precision still the same? Or did something that the synthetic benchmark doesn't care about or check have a problem? You can't tell me that.

Oh, and you also get that businesses don't buy things just for speed, right? They have to factor in longevity, because sure, that processor under liquid nitrogen cooling can be boosted to 10GHz, but how long do you think that processor will last? How practical do you think that system will be for the average office worker? Again, you're completely missing the point that the target audience of this machine is a business. Businesses expect these things to work, as advertised, at performance levels, right out of the box. No tweaking that's not supported under warranties required. This completely negates users here and there buying this laptop personally for personal use. If it was my laptop, I'd be all over it repasting it. Easiest way to reduce heat without taking anything out of Intel spec. It will most likely void the Dell warranty, though, so that's something to consider for a personal user if they feel like risking that. A business user doing that? You're risking your job.

You state again that sample variance is why Intel states the required voltage. I say again - correct. That is exactly what I've said. Uniformity. Key word there, again. Uniformity. Intel has set the bar for what voltage will work with ALL CPUs at that bin. Businesses do not have the time or the will or the money to take every single laptop and customize it to every single CPU. No way. They'd be out of business in no time doing that, and if not, IT would at least be out of a job for their budget spending.

Any company that wants to avoid the liability of having a problem will not run outside of Intel specs. That's why the OEMs ship the laptops at Intel specs. Even Dell doesn't want to void the warranty on the CPU and take that liability. If YOU choose to accept the risk because you THINK it's minimal, then that's your choice and your opinion. But do not state it as fact, because it simply is not.

Now, we're just going in circles and repeating the same things over and over, and not only am I tired of doing so, but I bet the other people on this thread are tired of it, too. I've made my points, time and time again, so it's time to let this rest. It's up to each and every individual user to determine whether or not they feel like making a tweak like this and accept the liability in order to use the laptop at full performance, or if they want to leave it stock but suffer PROCHOT throttling.

 

It overheats and throttles... for you. You tell me not to mislead, but unfortunately you're doing the same thing, by projecting and extrapolating your experiences with your single sample of the machine, to the entire population. You are consistently and doggedly missing my point, that the Precision's thermal solution is capable and sufficient, and that you have a lemon, which you should fix, while you moved the discussion off on a tangent by calling me out on my undervolt. You asked me to remove my undervolt and perform the same benchmark, and I did so. I still proved that without it, the CPU has slightly reduced performance, but still does not encounter throttle conditions.

Also, I was talking about PROCHOT throttling.

 

ORIGINAL: And a lot of other users in this thread that returned it because of that exact problem.

One could say you're doing the exact same thing with your single sample, as well.

And if you were talking about PROCHOT, then according to my understanding, you were wrong with how it throttles. If you have a link to how Intel does this now, I'd be happy to read it, but my understanding and my experience is that PROCHOT is end-all be-all of shutting down due to heat.

EDIT: I'm going to rephrase this, because as you pointed out, I AM mainly using my experience as the baseline for the cooling problem. So, yes, it very well could be that I have one of the laptops that does not cool well, due to either quality control, or a bad CPU, or something else. I will also say that I've read about other users having the same problem and returning their laptops based on that, even in this very thread. It is also entirely possible you received a good laptop, and because of that we are both on completely different ends of the spectrum with our experiences. My intention was not to mislead, and if I did that, then I sincerely regret it, apologize for it, and that's the point of this edit. I think right now, if I'm completely and totally objective, Dell has a quality control problem with these laptops. Some things I've seen in this thread unrelated to thermal throttling have made me shake my head; mainly the fact that screens can ship with dead pixels and dust sandwiched inside. That's just pure laziness OR overworked workers who don't have time to do good QC. I'm guessing it's the latter instead of the former, since many people seem to be happy with other Dell products.

 

That's because they, too, have had lemons, and subsequently received machines that worked fine after a warranty claim.

Perhaps you could ask @Div033, @Aaron44126, @Regular_Ragnor, or @CR3, all of whom have received exemplars.

You have consistently committed fallacies by appealing to the authority of Intel engineers, while users have produced substantial evidence that undervolting does little in the way of harm to the machine.

Intel and Dell will also likely not refuse warranty claims with a software undervolt—what you've said is akin to saying that they would not honour warranties if I installed Adobe CC, or a video game on these so-called 'business machines'. The reason is that the undervolt couldn't possibly be more reversible than it already is.

 

Man, you just don't stop, do you?

Go read Intel's errata data. Go read their warranties. Harm to the machine? Never said it. Affect the CPU and the calculations it does, sure can. And, just so you know, yes, an undervolt can cause damage. It can cause a stuck transistor, just the same way an overvolt can fry one. Same thing.

But I'm done, now. It's very clear you do not want to stop for reason, so I'll just let you go ahead and have the last word after this post, since I know a reply will be coming.