CompuLab Turns An 8-Core/16-Thread Xeon, 64GB RAM, NVIDIA Quadro RTX 4000 Into Fan-Less Computer

[schmidtbag]

I don't understand your position. You're clearly not willing to sacrifice everything for the sake of passive cooling, yet you seem unconcerned with just how much you are actually losing. What if you lost 1 GHz of all-core clocks? Would that get your attention?

You're right - I'm not willing to sacrifice everything, and I'm also not concerned with how much performance is lost. Based on the graphs shown, it's pretty clear the CPU isn't throttling that hard. It might not be able to sustain boost clocks, but the only reason you'd pick this PC over anything else is because it's small and silent. If you want it silent without dips in performance here and there, get something bigger. If you want something small without dips in performance, it can't be silent. You can't have everything.

Without the benchmarks, we can't make an informed tradeoff of how much performance is being sacrificed (and extra $ charged) for sake of silence.

We can't get anything exact but we can get approximations based on the data provided. Again: people who buy a PC like this should not be expecting to sustain 100% load for hours at a time.

I don't see what Mac Pros have to do with this. Do they throttle or run at below-spec clocks? AFAIK, the only bad thing about them is their ridiculous price.

Whoops I meant Macbook Pros. But yes, they're known to heavily throttle. In fact, I think they even throttle below base clocks. And people are still buying them. They offer amazing burst performance, and that's pretty much what Apple is expecting. Even if the cooling solution were adequate, people don't generally buy a laptop for things like heavy rendering. Same goes for small silent PC.s

No, we don't have the data to support that.

Yes, we definitely do. You're taking peak performance waaaay too seriously. Products like this obviously are of no interest to you.

Most people don't have one exclusive priority. They want to know how much performance they would be giving up. If you didn't care at all about performance, then why bother with an 8-core Xeon, in the first place?

Actually, most people don't give a shit and have no idea what it is they're getting. Most people don't know what exactly a core does. This is why you'll see grandmas with multi-GPU gaming PCs (this isn't an exaggeration BTW - I know of 2 personally), or people with 8-core phones when all they do with it is text people, use Snapchat filters, and browse Facebook. Hell, even a lot of enthusiasts don't really know what it is they need, which is why you'll see people who think getting a 3900X or 9900K is necessary for gaming.
What most people do care about is whether or not the system can keep up with their workload (doesn't matter by how much), if the price is right, and if it suits the ambience they seek. In the case of this PC, silence in a small package suits a niche ambience. It has hardware that decently well-informed people know is good. That's all that is necessary in order for it to sell.
Again: if performance is a top priority, you need something different.

Michael is definitely burying something. I think he favors certain vendors - especially smaller companies that give him stuff.

If that's what you want to believe, go ahead.

EDIT:
For the record, I do agree that seeing the actual performance would be nice, but, it's not that important, because the stress tests put on this system are a little unrealistic for its potential buyers.

[coder]

Based on the graphs shown, it's pretty clear the CPU isn't throttling that hard.

If you think hard throttling is the only thing to worry about, that's just naive. Motherboard vendors have more control than ever before over how fast your CPU actually runs. Here's some reading for you:

https://www.anandtech.com/show/13544...cted-tdp-turbo

It might not be able to sustain boost clocks,

It even goes beyond boost clocks:

https://en.wikipedia.org/wiki/Therma...nfigurable_TDP

This is certainly involved in the performance issues you cited with Mac Book Pros.

but the only reason you'd pick this PC over anything else is because it's small and silent.

Your position is anti-consumer. A pro-consumer attitude would be to provide people with as much information as possible, so they can make an informed tradeoff. If you're happy to place your faith in the vendor, then so be it. I prefer not to base my purchasing decisions on faith.

We can't get anything exact but we can get approximations based on the data provided.

All I'm asking for is the same data he usually provides. It should be compared with either the same CPU in a normal enclosure or a i9-9900K in similar.

Again, it's damn suspicious that he didn't provide this information.

Even better would be to provide a graph of clock speed over time - or any data on actual clock speed during the runs.

most people don't give a shit and have no idea what it is they're getting. Most people don't know what exactly a core does.

Most people don't buy workstations. If you're paying a premium price for an 8-core Xeon-based workstation, you either know exactly what you're buying, or you deserve to get screwed. I don't particularly care about people in the second camp. More information won't help them.

What most people do care about is whether or not the system can keep up with their workload

And that's exactly what benchmarks are for. So, why didn't Michael publish them? You don't have any answer for that.

[schmidtbag]

If you think hard throttling is the only thing to worry about, that's just naive. Motherboard vendors have more control than ever before over how fast your CPU actually runs. Here's some reading for you:

Not exactly sure what your point is here. You and that article didn't specify what else there is to worry about. Power draw is not really much of a concern in this case.
The control motherboard vendors have is a double-edged sword in your argument. They also have more control over voltage. Using Apple's Macbooks as an example again, they actually change the voltage and thermal throttling profiles so the CPU remains at a higher speed for a longer amount of time. It wouldn't surprise me if Compulab picked binned parts that they were able to undervolt; it doesn't appear they changed when the CPU thermal throttle mapping when you consider the peak temperature.

Your position is anti-consumer. A pro-consumer attitude would be to provide people with as much information as possible, so they can make an informed tradeoff. If you're happy to place your faith in the vendor, then so be it. I prefer not to base my purchasing decisions on faith.

No, my position is to focus on what the product was built for. It wasn't built to do what you are expecting. It's built to be small and silent with decent performance, which is precisely what it does. Yes, I get it, we don't have the exact numbers. But, you can find out how fast the CPU and GPU are with their full potential. Based on the thermals Michael tested, I think we can safely assume that both them can reliably operate at their base clocks under full load for an indefinite amount of time. So, since the 2288G's base clocks are 26% lower than the boost clocks (assuming 5GHz is single-core boost), take a single-threaded benchmark of it that interests you and multiply the results by 0.74. Figure out the all-core boost clocks and do the same for a multithreaded test. That'll show you the worst-case scenario you'll get in this PC. As for the Quadro, multiply by 0.65.

All I'm asking for is the same data he usually provides. It should be compared with either the same CPU in a normal enclosure or a i9-9900K in similar.

I don't disagree.

Even better would be to provide a graph of clock speed over time - or any data on actual clock speed during the runs.

Yes, that I would definitely like to see.

Most people don't buy workstations. If you're paying a premium price for an 8-core Xeon-based workstation, you either know exactly what you're buying, or you deserve to get screwed. I don't particularly care about people in the second camp. More information won't help them.

And yet, you don't seem to realize that people who are seeking this kind of thing know exactly what they're getting into, which implies some sacrificed performance. I agree it'd be nice to know exactly how much is sacrificed, but I'm pretty confident that most people who would buy this don't care anywhere near as much as you might think. This is a very niche product.

And that's exactly what benchmarks are for. So, why didn't Michael publish them? You don't have any answer for that.

Never said I had an answer for that, so, not sure why you're challenging me on it. I'm not even saying you're wrong to be suspicious (I don't agree, but I don't care to argue over such things), I just think you're taking this way too seriously.

[coder]

Not exactly sure what your point is here. You and that article didn't specify what else there is to worry about.

The point is that motherboard vendors have precise control over boost behavior - how high it boosts and for how long. Not in terms of clockspeed, but rather by constraining power.

The article quotes from Intel documentation that "PL1 ... recommend to set equal to TDP power. PL1 should not be set higher than the thermal solution cooling limits."

Strongly implies that PL1 can go below the CPU's specified TDP, resulting in below-expected sustained performance. If that's what they're doing, you wouldn't see it in those temperature graphs, because it doesn't look like normal throttling. It's more like dialing back the base clocks.

Power draw is not really much of a concern in this case.

https://en.wikipedia.org/wiki/First_...thermodynamics

Yes, I get it, we don't have the exact numbers. But, you can find out how fast the CPU and GPU are with their full potential. Based on the thermals Michael tested, I think we can safely assume that both them can reliably operate at their base clocks under full load for an indefinite amount of time.

No, we cannot conclude that. The thermals of this cooling solution cannot be directly compared with the thermals of the same CPU with a different cooling solution, unless more is known about the cooling solutions. Then, you need another mapping function to translate that back onto clock speed.

It's much simpler just to have the raw benchmarks. Clockspeed graphs would be nice, too.

So, since the 2288G's base clocks are 26% lower than the boost clocks (assuming 5GHz is single-core boost), take a single-threaded benchmark of it that interests you and multiply the results by 0.74. Figure out the all-core boost clocks and do the same for a multithreaded test. That'll show you the worst-case scenario you'll get in this PC. As for the Quadro, multiply by 0.65.

As I explained above, you cannot assume that it will maintain the specified base clocks. Those relate to its advertised TDP, but they could've dialed in PL1 below that point.

Quadro surely offers similar knobs, since Nvidia now sells the same GPUs into both mobile and desktop solutions. The only difference is how the power curves are configured.

And yet, you don't seem to realize that people who are seeking this kind of thing know exactly what they're getting into,

How can they know exactly, without the data?

This article is like running a car on a dynomometer blindfolded. You're only getting the noise, without knowing the peak RPMs and torque curve. It's just dumb. You're arguing on the side of stupidity and ignorance. I don't know how you can justify this to yourself, but it really hurts your credibility.

I'm pretty confident that most people who would buy this don't care anywhere near as much as you might think.

Do you have any market research on that? Are those people reading this site? What matters is the prospective buyers who are reading this site - not the general computing public.

This is a very niche product.

Which is exactly why its buyers are more likely to care, than average.

Never said I had an answer for that, so, not sure why you're challenging me on it. I'm not even saying you're wrong to be suspicious (I don't agree, but I don't care to argue over such things), I just think you're taking this way too seriously.

You clearly do care, because you definitely are arguing that the actual performance doesn't matter and people should just take it on faith that this performs to their expectations. That completely defeats the point of measuring stuff, and it's about the dumbest position you can take, in comments on a benchmarking article.

[schmidtbag]

Strongly implies that PL1 can go below the CPU's specified TDP, resulting in below-expected sustained performance. If that's what they're doing, you wouldn't see it in those temperature graphs, because it doesn't look like normal throttling. It's more like dialing back the base clocks.

What we're seeing in the temperature graphs is the CPU's average temperature is more than cool enough to run the default base clocks. The PC is advertised at 5GHz (which obviously isn't all cores) and PTS acknowledges that. Considering how easy it is to prove if it's actually doing that, CompuLab would be sued pretty easily for false advertising.
Your skepticism isn't accomplishing anything, and the Anandtech article doesn't change that.

https://en.wikipedia.org/wiki/First_...thermodynamics

How is this relevant? You are really not doing a good job making your point clear. You keep talking about wattage as though that matters, when we don't know the thermal capacity of the cooling solution. Despite what you want to believe, there really is nothing to suggest the CPU isn't maintaining base clocks. Yes, I get it - it's possible the manufacturer tweaked the power profiles. Doesn't change the fact that the CPU most likely is capable of boosting to some degree. Like I said before, I agree with you that seeing CPU frequency would really help clear things up.

No, we cannot conclude that. The thermals of this cooling solution cannot be directly compared with the thermals of the same CPU with a different cooling solution, unless more is known about the cooling solutions. Then, you need another mapping function to translate that back onto clock speed.

*sigh* I didn't say you were going to get anything exact, I'm saying you'll get an approximate idea of the worst-case scenario. You seem to be very pessimistic over how poorly this CPU can be cooled in this PC.

As I explained above, you cannot assume that it will maintain the specified base clocks. Those relate to its advertised TDP, but they could've dialed in PL1 below that point.

Unless you have evidence that any manufacturer has ever underclocked the PL1, I'm sticking with "it can sustain [the default] base clocks". It's more likely they undervolted than underclocked.

How can they know exactly, without the data?

I meant "exactly" as in the scenario they're getting themselves in, not the benchmarks. I'll admit I phrased myself poorly there. But even with the data, it's inconclusive. Ambient temperature and airflow will have a dramatic difference on a passively cooled PC under heavy load. That being said: people who buy this PC know they're not getting the best performance the CPU can offer. They know they're making sacrifices. You seem to consistently ignore that.

This article is like running a car on a dynomometer blindfolded. You're only getting the noise, without knowing the peak RPMs and torque curve. It's just dumb. You're arguing on the side of stupidity and ignorance. I don't know how you can justify this to yourself, but it really hurts your credibility.

No, I'm not. I'm arguing on the side of "you're getting anal about something that doesn't really matter that much". To work off your example, it's like running an economy car on a dyno. Seriously, who cares? Most people don't give a crap how much power an economy car makes, they want to know what kind of stuff they can fit in it, how safe it is, maintenance costs, and how much fuel they need to run it. If knowing how much power the engine makes matters to you, don't buy an economy car.
This PC is no different. You're not buying it for the power. Is it nice to know? Absolutely - I don't disagree with that and I never did.

Do you have any market research on that? Are those people reading this site? What matters is the prospective buyers who are reading this site - not the general computing public.

You're changing the goalposts there... You're the one who used "most people".
But anyway, look at the market statistics on Mac users, because many of them seem well aware (or are willfully ignorant) that they are making some hefty sacrifices in performance, whether that be due to thermal issues, software compatibility, or the fact that Mac OS is just simply slow in a lot of tasks.

Which is exactly why its buyers are more likely to care, than average.

In another light, such buyers know what they're getting into and understand that if they want better performance, they ought to look elsewhere. For $1700, you can buy a better performing PC. If you buy this, you know you're not getting the hardware's full potential. And if you don't know, you won't realize it.

You clearly do care, because you definitely are arguing that the actual performance doesn't matter and people should just take it on faith that this performs to their expectations. That completely defeats the point of measuring stuff, and it's about the dumbest position you can take, in comments on a benchmarking article.

We're discussing many different things there. What I don't care about is whether you think Michael is trying to hide something. I also never said the performance doesn't matter. I'm saying it's not the main reason you buy a PC like this. Again: I agree it's nice to know, but it also isn't as sacrilegious as you make it out to be.
Stop twisting what I say into such extremes. Seeing as you care about credibility, what you're doing isn't helping yours. I don't disagree with you anywhere near as much as you think (in fact, I agree with most of your points), you're just blowing everything (including your main complaint) way out of proportion.

[coder]

What we're seeing in the temperature graphs is the CPU's average temperature is more than cool enough to run the default base clocks.

You don't know that the temperature isn't that low because it's not running at the specified base clocks.

The PC is advertised at 5GHz (which obviously isn't all cores) and PTS acknowledges that. Considering how easy it is to prove if it's actually doing that, CompuLab would be sued pretty easily for false advertising.

lawsuits are expensive and involved, and I'm sure there's enough fine print, somewhere, that lets Intel & Compulab weasel their way out of having to deliver any specific clock speed.

But, I never said it didn't or couldn't hit 5 GHz. If you read the Anandtech article, you'll know that the Tau parameter allows them to control how long it boosts at PL2.

I'm not even saying it can't sustain base clocks. All I'm saying is that it might not, we don't have the data to indicate that it does. Maybe it can do better than base clocks. Wouldn't that be nice to know? If you're trying to choose between systems, sustained performance would be a very relevant detail!

Your skepticism isn't accomplishing anything, and the Anandtech article doesn't change that.

It's accomplishing more than your assertion that these details don't matter. The Anandtech article explains how the CPU can be configured to run below base clocks. It is entirely relevant to my point that nothing about its clock speeds can be taken for granted. If you don't understand that, then I guess that's your loss.

How is this relevant? You are really not doing a good job making your point clear.

You said "Power draw is not really much of a concern in this case." Thermal dissipation must equal power draw. By constraining power, you can reduce the amount of energy that the thermal solution must dissipate. It's nothing you shouldn't already know, and the part of the Intel docs I quoted explicitly linked PL1 to the thermal solution.

I didn't explain it, because I shouldn't have to. If you need to be spoonfed, then you're not worth my time.

we don't know the thermal capacity of the cooling solution.

Yes, that's what I said in my previous post, when explaining why the temperature data tells you (almost) nothing. Therefore, it would be much easier if we just had the benchmarks.

there really is nothing to suggest the CPU isn't maintaining base clocks.

Nor anything to support that it is. The lack of data is the problem. You want to substitute data with faith, while I say "let's have the data"!

Yes, I get it - it's possible the manufacturer tweaked the power profiles. Doesn't change the fact that the CPU most likely is capable of boosting to some degree.

I never said it couldn't boost. See above.

You seem to be very pessimistic over how poorly this CPU can be cooled in this PC.

Nope. I just want the data. I'm refusing to make any assumptions.

Unless you have evidence that any manufacturer has ever underclocked the PL1, I'm sticking with "it can sustain [the default] base clocks". It's more likely they undervolted than underclocked.

Like I said, you want to take it on faith, while I want to see the data. That's why people benchmark things - so they don't have to take this stuff on faith.

people who buy this PC know they're not getting the best performance the CPU can offer. They know they're making sacrifices. You seem to consistently ignore that.

I never ignored that. I've consistently said that people should be able to make an informed tradeoff, which you cannot do without the data.

To work off your example, it's like running an economy car on a dyno. Seriously, who cares? Most people don't give a crap how much power an economy car makes, they want to know what kind of stuff they can fit in it, how safe it is, maintenance costs, and how much fuel they need to run it. If knowing how much power the engine makes matters to you, don't buy an economy car.

Dumb example. Of course the power of an economy car matters. You might want to know if it'll be able to make it up a steep hill, when it's loaded to its max capacity. You might want to know how long it will take to reach highway speed, at a dangerous on-ramp. If the car can't meet your needs, then you might look at a different model.

Seriously, why do you think car sites & publications even publish performance data on economy cars? Of course it matters! It factors into the overall decision, helping buyers make an informed choice!

This PC is no different. You're not buying it for the power.

If you don't need the power, then there are much cheaper options. Of course its performance matters. Your first post was nothing, if not acknowledgement of this fact.

In another light, such buyers know what they're getting into and understand that if they want better performance, they ought to look elsewhere.

You're trying to paint this as a binary choice. If you care about fanless, then you must completely disregard performance. Or, if you want performance, then you must completely disregard noise. Rarely is the world so simple that someone can have a single priority and no other need or constraint. Moreover, there are other fanless and quiet PCs out there.

But I'm not telling you anything you don't know. You're just being disagreeable because I caught you with your metaphorical pants down. You simply cannot justify your position.

What I don't care about is whether you think Michael is trying to hide something.

Then why didn't he post the same data that he posts in all his other benchmarks? That should send up red flags!

[Compulab]

Then why didn't he post the same data that he posts in all his other benchmarks? That should send up red flags!

A full review by Phoronix is WIP as specified in the article.

Stay tuned for our full review of the CompuLab Airtop 3 coming up soon on Phoronix as well as using this fan-less Xeon E + RTX 4000 system for other interesting Linux/BSD benchmarks moving forward.

For reference, Phoronix reviewed the 1st generation Airtop back in 2016 and published initial hands-on and 4 weeks later the full review.
I believe we should expect a similar pattern here, but Compulab cannot interfere with the procedure, timeline and content of reviews. I guess patience is a virtue.

FWIW, I expect that for Airtop3 some throttling will take place in some of the long stress tests, as we observed in our internal tests.

Best regards,
Irad Stavi
Compulab

[schmidtbag]

You don't know that the temperature isn't that low because it's not running at the specified base clocks.

This is a waste of time. You're arguing over a hypothetical situation that you have no evidence of. I'm giving CompuLab the benefit of the doubt that they didn't tamper with the PL1 clocks. Do I have evidence of that? No, but I have no good reason to believe otherwise.

But, I never said it didn't or couldn't hit 5 GHz. If you read the Anandtech article, you'll know that the Tau parameter allows them to control how long it boosts at PL2.

Considering how much you question its ability to sustain base clocks, how do you have any faith that it can reach boost clocks?

Wouldn't that be nice to know? If you're trying to choose between systems, sustained performance would be a very relevant detail!

How many times do I have to tell you that yes, it is nice to know? You seem to consistently ignore that, which is kinda weird to me, since all you seem to care about is being right.

It's accomplishing more than your assertion that these details don't matter. The Anandtech article explains how the CPU can be configured to run below base clocks. It is entirely relevant to my point that nothing about its clock speeds can be taken for granted. If you don't understand that, then I guess that's your loss.

I get that it can be configured that way, but I'm not interested in arguing over something that doesn't seem likely.

You said "Power draw is not really much of a concern in this case." Thermal dissipation must equal power draw. By constraining power, you can reduce the amount of energy that the thermal solution must dissipate. It's nothing you shouldn't already know, and the part of the Intel docs I quoted explicitly linked PL1 to the thermal solution.

I understand that... seeing as I think it's maintaining Intel's default base clocks, that means the cooling solution is sufficient. Therefore, power draw is irrelevant. Your skepticism is not my problem. If you're right that they meddled with PL1, then go ahead and come back here to rub it in.
And yes, knowing the data would help put the assumptions to rest, but again, there's nothing that leads me to believe they are meddling with PL1 clocks.

I didn't explain it, because I shouldn't have to. If you need to be spoonfed, then you're not worth my time.

You're the only one complaining here about the tests. You're the one trying to justify your complaints to me. If you want to make your point understood, then yes, you should have to explain yourself. Besides, why stop here, after all the other obvious stuff you went out of your way to explain? If I'm not worth your time now, then I never was to begin with. Leaves me suspicious if you actually know the answer yourself.

Nor anything to support that it is. The lack of data is the problem. You want to substitute data with faith, while I say "let's have the data"!

That's because I'm not a pessimist like you. I don't think CompuLab would stoop low enough to cripple the PL1 performance. Is it possible they did? Yes. Would it be great to have proof for it? Absolutely. But I don't see the lack of data as big of a conspiracy as you do. You're kinda obsessive over this...

Nope. I just want the data. I'm refusing to make any assumptions.

Except... you're making quite a lot of assumptions yourself...

I never ignored that. I've consistently said that people should be able to make an informed tradeoff, which you cannot do without the data.

You're literally ignoring it right now lol. You care about the data, regardless of the sacrifices. That's not necessarily a bad thing but it's why we're in a disagreement. People who buy this stuff don't care about the data anywhere near as much as you think. They're buying something they know is slower than alternatives for the same money. So riddle me this: why care about the in-depth details of the performance when you're willing to spend good money on something you know is slower? How are you not seeing the point here?
Again, I don't disagree that knowing the numbers would be nice. But unless the cooling solution is dismally terrible, it isn't going to affect the sales of the product. So, it's not a big focus.

Dumb example. Of course the power of an economy car matters. You might want to know if it'll be able to make it up a steep hill, when it's loaded to its max capacity. You might want to know how long it will take to reach highway speed, at a dangerous on-ramp. If the car can't meet your needs, then you might look at a different model.

Are you serious? Knowing the power of the vehicle doesn't determine any of that, because you also have to account for other things like the vehicle's curb weight, drivetrain losses, the slope of the road, transmission gearing, torque, and so on. Knowing the 0-60 / 0-100 times answers your questions a hell of a lot better than how much power the engine makes. Besides, name just 1 modern car that struggles to reach highway speeds.
The point remains: people don't buy an economy car because of the power it makes. I don't deny that knowing the power is nice, but it isn't THE reason people buy it.

Seriously, why do you think car sites & publications even publish performance data on economy cars? Of course it matters! It factors into the overall decision, helping buyers make an informed choice!

And yet, CompuLab is also specifying the power of the CPU... so, nothing here is really any different. Much like a manufacturer specifying the power of an economy car, you're getting a gist of what the product can do. That's all the potential customers care about. Anything beyond that is just simply "nice to know".

If you don't need the power, then there are much cheaper options. Of course its performance matters. Your first post was nothing, if not acknowledgement of this fact.

Here you go again, ignoring the known sacrifices. And yet you claim you aren't ignoring them. How much do you need to be spoonfed this:
People who buy this PC know there are more powerful and cheaper options.
If they know this, how are you not realizing that the exact performance numbers don't matter that much to them? Key phrasing here is "that much". As I have stated over and over again, it is nice to know.

You're trying to paint this as a binary choice. If you care about fanless, then you must completely disregard performance. Or, if you want performance, then you must completely disregard noise. Rarely is the world so simple that someone can have a single priority and no other need or constraint. Moreover, there are other fanless and quiet PCs out there.

Actually, I painted a 3-way picture earlier which, as you normally do, completely ignore. If you care about small and fanless, you have to accept performance will suffer. That doesn't mean you completely disregard performance, but, it isn't crucial to people's decision of buying it.
I'm aware there are other silent PCs out there. Some of them are physically much larger. Some of them are slower.
Like I said before, you have 3 categories to pick from: size, loudness, and performance. You can only pick 2. People who pick the PC in this article are choosing a small size and silence. They know they're not getting the best performance. Therefore, knowing the exact numbers aren't as important as you think.

But I'm not telling you anything you don't know. You're just being disagreeable because I caught you with your metaphorical pants down. You simply cannot justify your position.

How ironic. I disagree with what you said there because contrary to your beliefs, I'm not as disagreeable as you think I am.
My position has remained very consistent the entire time:
Knowing the numbers is nice, but not very important to the potential buyers.

Then why didn't he post the same data that he posts in all his other benchmarks? That should send up red flags!

Again, I don't care. I agree it's a little weird but you're getting so needlessly worked-up over this.

[coder]

This is a waste of time. You're arguing over a hypothetical situation that you have no evidence of.

I'm actually not. I'm not accusing them of anything, but you need to understand that those parameters were exposed by Intel specifically for vendors to tune CPU power dissipation to suit their solution. They almost certainly customized them - we just don't know if they set PL1 below the specified TDP.

Recall that you credited them for the specs on their system and I pointed out that it's premature, as we don't actually have the data to know how well it performs. I didn't actually say it performs poorly - we just don't know.

I'm giving CompuLab the benefit of the doubt that they didn't tamper with the PL1 clocks.

Not clocks - power. And you make it sound like it's a bad thing, when the point of it is for maximizing the CPU within the power and thermal capabilities of the system. I don't consider it foul play, unless they also advertised their system had no performance penalties relative to conventional cooling.

Considering how much you question its ability to sustain base clocks, how do you have any faith that it can reach boost clocks?

You're clearly out of your depth. The point of them having a separate set of boost parameters is because the ability of a cooling solution to handle boost is somewhat separate from its capacity for sustained heat transfer.

How many times do I have to tell you that yes, it is nice to know? You seem to consistently ignore that,

I don't ignore that - rather it's inline with the core issue.

If you want to make your point understood, then yes, you should have to explain yourself.

My point was I shouldn't have to teach you science that the typical 10th grader should know.

I don't think CompuLab would stoop low enough to cripple the PL1 performance. Is it possible they did? Yes. Would it be great to have proof for it? Absolutely. But I don't see the lack of data as big of a conspiracy as you do. You're kinda obsessive over this...

The setting is there to be configured. I think it would be a legit thing to do.

Except... you're making quite a lot of assumptions yourself...

Nope. No assumptions. I just want the data.

You care about the data, regardless of the sacrifices.

This makes no sense.

They're buying something they know is slower than alternatives for the same money. So riddle me this: why care about the in-depth details of the performance when you're willing to spend good money on something you know is slower? How are you not seeing the point here?

This is really the core issue. You act as if there's only one fanless PC in the world. That people have a binary choice, do I go fanless or not? Totally ignoring that there are other fanless PCs and that someone might want performance data to help them decide which to buy. Even within the offerings of a single vendor, they might want data to help them decide if a higher-end model is worth the additional money. And, if they don't need it to be completely fanless, they might be willing to consider other low-noise options, especially if they were a cheaper and offered more performance.

The consequence of your position is that gaming PC reviews should not have acoustic noise or power consumption data, because gaming PCs are loud and burn a lot of power. And because they're a niche product and people should realize that they'll be louder and burn more power, that they shouldn't care about the specifics or have the ability to make an informed decision between various alternatives, even within the category.

Another consequence of your position is that sports car buyers shouldn't care about fuel mileage or cargo capacity, because they know those parameters will be worse and therefore must completely abandon any concerns other than performance and handling.

You live in a very strange universe, where people lack competitive alternatives and have only one single priority, to the exclusion of all else. Thankfully, this is not a world I recognize. I will leave you to inhabit this lonely world, bereft of non-overlapping categories and informed compromises.

Knowing the power of the vehicle doesn't determine any of that,

I might know that in a vehicle of similar power and weight, I have trouble maintaining speed up a particularly steep road or that I have trouble merging from a dangerous on-ramp. Numbers can be used to help compare things, even if they're harder to use entirely in the abstract.

My position has remained very consistent the entire time

Yes, in spite of not knowing quite a lot about the subject matter, it seems. Changing one's position in light of new information is a virtue, not a character deficiency.

Unfortunately, ignorance and obstinance go together like gasoline and oxygen.

Again, I don't care. I agree it's a little weird but you're getting so needlessly worked-up over this.

I understand that you don't care, because you seem to live in some kind of alternate reality.

The bigger issue would be if Michael is cherry-picking benchmarks to cast vendors' products in the best light. That said, I saw Compulab's latest reply, and will approach the full review with an open mind.

See also: https://www.phoronix.com/scan.php?pa...ir-force-mp600

[danmcgrew]

This entire discussion has devolved into a "How many angels can dance on the head of a pin?" argument. Drop it, guys. We don't care any more. We don't even READ any more--a major problem for Phoronix--when we see your names pop up.

Absolute facts:

1) Phoronix is a 'class' operation; proven by history.
2) Compulab is a ''class; operation, proven by history.
3) Everyone knows that the same electronics will run hotter in a sealed box than in an open one (caveat emptor).
4) We're not talking about some 'crowd-funded', "iffy" piece of hardware here.
5) If one simply waits, points #1 and #2 will be made manifestly obvious by forthcoming, additional information.
6) One's position does NOT gain credibility by the casting of aspersions; as a matter of absolute fact, one's credibility takes a major hit when one stoops to this tactic.
7) Michael Larabel needs to reign in this type of self-serving activity, or risk compromising point #1. Make no mistake about it: the only reason for the continuation of this circus, by two people, is ego.

A simile which keeps coming to mind about this thread now is provided by a sports writer who normally covered football games of all major football conferences--PAC 10, Big 10, SEC, ACC, etc. On a vacation to the Northeast, he was invited--with no small amount of pride--by his hosts to a game between two big-name, but lower-achieving (sports-wise only) Ivy-League schools. When asked later what he thought about the game, his reply was, "It was somewhat like watching two mules fighting over a turnip."

This thread has become "...two mules fighting over a turnip." Careful, Michael.