Negative or positive pressure? Does it matter?

[maxfly]

I dont even use doors/panels on my cases.

I feel like this is more of a peer pressure scenario.

Not if you have tiny fingers or furry ones beboppin around. Its absolutely a safety hazard in my house. All i need is one of our cats to get onto one of my rigs. My wife would hurt me in ways id rather not think about!

 

[eidairaman1]

This video with smoke is comprehensive.

Neutral or positive air flow please.

If you had micron filters you would need stronger fans which can get noisy.

Just keep you case clean a month or 2, keep it off floor/carpet.

 

[Mussels]

Slightly positive always works. as air creeps out every gap
That does help with dust but primarily it helps cool all the random components in the case - your VRM's, ram modules, random heatsinks and whatever... the greater the air*flow* the better it is for all your parts.

I have 4 in and 3 out in my system, with as many cracks and gaps sealed up as possible - feeling the heat passively venting out from my monster 3090 *where i want it to be* is really nice in winter (i'll cry in summer)

 

[robot zombie]

Err Long post , ended well, theorising isn't going to get you there.

That sentence says a lot.

If you really want to know what's best, try it and note the outcome. I like forums for info sometimes, and honestly this has spawned a great discussion. But what this is making me acutely aware of is this sense that one cannot factor in every possible variable and outcome. Coming from HVAC, what makes sense on paper (or in conversation/planning) doesn't always play out how you think, even when you think that it should. Because you can't think of everything. But you can learn from experience, whereas advice and commentary sometimes leave more questions. You can use those questions to inform your own experience.

Fluid systems are very strange. The physics behind them are difficult. All I can say. There are all of these general rules that tend to fly. But ultimately you must configure them for the exact circumstances they are placed in. Occasionally changes need to be made later. There are hidden details. Hell, in my own work there are some times when I don't even know why something works better, and think it shouldn't. And I'm supposed to know what I'm doing!

My experience tells me that a lot of these conversations are just that, conversations. No offense to anyone here. To me, there are just practical limits to internet thought exercises and sometimes you just have to see what happens in the real world or spend your days wondering about things that may or may not matter.

 

[Vayra86]

That just means that your power limit is set too high. Something like that shouldn't happen in prime95, as it is close to many BOINC workloads or encoding stuff. It's not unrealistic load and your hardware has to handle that for long periods of time. For thermal testing, you could just disable turbo at all and test at base speed and fixed fan rpms.



I don't find Furmark malicious. I used all my cards with Furmark at some point and nothing spectacular happened. Neither in thermals or power usage Furmark wasn't much different from Unigine benchmarks or games or BOINC workloads.

As for Cinebench, my watt meter and CPU temperature sensor says that it's just as bad as prime95, maybe a little bit worse. A light CPU stress test is CPU-Z test, but only stable version, which doesn't utilize AVX. But it's a poor test for thermals and it doesn't check stability well, so I would avoid it.

Furmark is malicious for GPUs in the sense of it presenting an unrealistic load; GPUs will clock back but still use almost maximum board power until the bios limits them (specifically for Furmark!); the result is not testing anything useful - low efficiency power usage at a low clock at the very best is a ridiculous way to test your cooling. You are not testing OC stability, not testing temp at a realistic load scenario and you are not testing behaviour under boost clocks either. Simply because you wont get them.

Time Spy or other actual benchmark runs do present all of those things and still serve as max load/temp scenarios for GPU.

 

[AusWolf]

Then it's not fine. It's set too high, lower it and it will be fine in prime and you shouldn't see any performance effect everywhere else.

It is fine for everything, except for Prime95. Why would I want to undersize my power limits just to suit a power virus type program that has no practical use?

Not sure what hell is, but it's not much different from using your GPU for Collatz@Home or MilkyWay@Home or playing a video game that utilizes it to 100%.

Just as @Vayra86 said it. It is very different.

Well, I currently have set PL1 to 75 watts and PL2 to 80 watts, Tau is Intel spec so probably 28 seconds. Anyway, my cooling handles this chip with power limits lifted it's just that it tops out at high 70s, maybe low 80s. I prefer not to exceed 70C as is is recommended temperature limit for maximum turbo boost, else cooler with stock fan curve will be at 100% speed. And it's nice for VRMs too as I likely have one of those boards, that can't handle i9k.

75-80 W is very low for Comet/Rocket Lake. You easily reach it in everyday applications even with 6 cores. My 11700 eats around 65-70 W in games with a 50-ish % load. If your cooling is good enough, try raising your PL1 to above 150 W, and you'll see how much different Prime95 and Cinebench are.

 

[Vayra86]

It is fine for everything, except for Prime95. Why would I want to undersize my power limits just to suit a power virus type program that has no practical use?


Just as @Vayra86 said it. It is very different.


75-80 W is very low for Comet/Rocket Lake. You easily reach it in everyday applications even with 6 cores. My 11700 eats around 65-70 W in games with a 50-ish % load. If your cooling is good enough, try raising your PL1 to above 150 W, and you'll see how much different Prime95 and Cinebench are.

Yep... my 4.7 Ghz on the 6c 8700K took upwards of 130W, and at 140-150W air just cant keep it under control anymore.

75-80W is nothing. I ran that (77W package) on the 3570K with a 20 dollar Gelid Tranquillo Rev 2

That said, in my world an OC is only useful if it is rock solid. I do stress everything with Prime95, 15 minute run at least, additionally with OCCT for a 30 minute linpack in which I want zero errors. The only crash from CPU instability Ive had was a 100C Bsod in hot summer last year.

Not running a full blown stress test because you hit thermals is basically just a shitty OC that doesnt work. You do OC for peak performance, right? Otherwise its just a fancy epeen number really.

 

[AusWolf]

Yep... my 4.7 Ghz on the 6c 8700K took upwards of 130W, and at 140-150W air just cant keep it under control anymore.

75-80W is nothing. I ran that (77W package) on the 3570K with a 20 dollar Gelid Tranquillo Rev 2

That said, in my world an OC is only useful if it is rock solid. I do stress everything with Prime95, 15 minute run at least, additionally with OCCT for a 30 minute linpack in which I want zero errors. The only crash from CPU instability Ive had was a 100C Bsod in hot summer last year.

Not running a full blown stress test because you hit thermals is basically just a shitty OC that doesnt work. You do OC for peak performance, right? Otherwise its just a fancy epeen number really.

Peak performance in what? A 30-minute Cinebench run is plenty of stability testing for me.

 

[freeagent]

Cinebench isn’t a stress test though, and furmark can be dangerous

 

[Vayra86]

Peak performance in what? A 30-minute Cinebench run is plenty of stability testing for me.

Cinebench is very light and a performance bench, not a stress test.

You can shop selectively as you like but its an illusion cinebench is for stability testing.

 

[AusWolf]

Cinebench is very light and a performance bench, not a stress test.

You can shop selectively as you like but its an illusion cinebench is for stability testing.

I know it's not, but it's more representative of a CPU load that you encounter real-life with programs that you actually use.

Another topic: I couldn't wait until after my holiday, so I did the swap. Radiator in the front, acting as intake, case fans on the top, no dust filters. My first reaction even without thermal testing is that the system just got A LOT louder. I don't know why. My guess is that my case fans don't like being mounted horizontally. They also produce a lot more airflow than the AIO fans on the rad, resulting in a negative pressure situation inside the case. With no dust filters, this won't be good long-term. This is a setup that I already don't like. I'm not even sure if I want to go through thermal testing.

 

[The red spirit]

Furmark is malicious for GPUs in the sense of it presenting an unrealistic load; GPUs will clock back but still use almost maximum board power until the bios limits them (specifically for Furmark!); the result is not testing anything useful - low efficiency power usage at a low clock at the very best is a ridiculous way to test your cooling. You are not testing OC stability, not testing temp at a realistic load scenario and you are not testing behaviour under boost clocks either. Simply because you wont get them.

Time Spy or other actual benchmark runs do present all of those things and still serve as max load/temp scenarios for GPU.

There's no evidence to suggest that. I say to you GPU uses nearly the same amount of watts, gets pretty much as hot as in Heaven. If you have such argument, you gotta have proof.

 

[Vayra86]

I know it's not, but it's more representative of a CPU load that you encounter real-life with programs that you actually use

And you claim to know what instructions your CPU will and will not use and how much power it will draw? Interesting

An example: firing up Overwatch utilizes a high performance bit of AVX and loads on multiple threads. It pushes the 8700K to 80+ C here, if only briefly. It was this game that managed to peak over 100C in the earlier example and BSODs for it.

So you might want to draw your own conclusions on that Dont for a second think you know what a 'real world' load is supposed to look like. It simply is selective shopping to lower the bar for an OC thats probably better off a few hundred mhz lower. After all, if you arent fully loading the CPU, why even OC it out of its efficiency curve? The practical use is negligible...

There's no evidence to suggest that. I say to you GPU uses nearly the same amount of watts, gets pretty much as hot as in Heaven. If you have such argument, you gotta have proof.

Monitor your clocks and power draw then. "Pretty much" says enough. Stop deluding yourself This info on Furmark is very old news that has not changed since.

why does'nt my gpu clock speed go fully up?

hey guys so i have a zotac gtx 1050ti mini + i5 4460 i oc'd it to 1504 mhz on d clock and 2002 mhz on d mem clock but when i stress tested it with furmark, d gpu clock reaches a max of 1341 mhz. is this fine or is something wrong?(i'm a noob) also this computer is powered by a 300w psu... am i sa...

linustechtips.com

Furmark lower FPS and lower score on higher gpu clock and power limit?

Is that due to Nvidia built in Protection, when I run furmark at 113% power limit, it hits around 1750-1800Mhz and gives 200 FPS but when I run 113% power limit, it goes 2000-2040Mhz but at 155 FPS? Palit 3070 Gaming Pro

linustechtips.com

Need more?

 

[lexluthermiester]

If you have such argument, you gotta have proof.

You might as well go looking for the pot of gold at the end of the rainbow...

 

[The red spirit]

It is fine for everything, except for Prime95. Why would I want to undersize my power limits just to suit a power virus type program that has no practical use?

Because as you say, no other software makes your CPU to consume as many watts and you know that t's bad if it does.

Just as @Vayra86 said it. It is very different.

No explanation why. BTW Furmark seems to be similar to MSI Kombustor or OCCT's GPu stress test.

75-80 W is very low for Comet/Rocket Lake. You easily reach it in everyday applications even with 6 cores. My 11700 eats around 65-70 W in games with a 50-ish % load.

That's where you are wrong. i5 10400F at full load in games and maximum all core boot consumes 50-55 watts. Maximum all core boost is 4GHz. There are plenty of loads that can be executed at maximum boost and fit within 65 watt PL1. I really have to go out of my way to each 80 watts and only prime95 was able to push it to more than 90.

If your cooling is good enough, try raising your PL1 to above 150 W, and you'll see how much different Prime95 and Cinebench are.

I literally told you that I ran rime95 with PLs lifted (set to 4095). My cooling was able to cope with that, but I don't want my hardware to be able to reach such temperatures (were in low 80s on CPU). Cinebench R23 is still less demanding, so outcome would be the same.

 

[freeagent]

You might as well go looking for the pot of gold at the end of the rainbow...

Not to worry, I am Irish..

And Scottish too, but don't worry I'm not gonna show you my Caber

 

[TheoneandonlyMrK]

And you claim to know what instructions your CPU will and will not use and how much power it will draw? Interesting

An example: firing up Overwatch utilizes a high performance bit of AVX and loads on multiple threads. It pushes the 8700K to 80+ C here, if only briefly. It was this game that managed to peak over 100C in the earlier example and BSODs for it.

So you might want to draw your own conclusions on that Dont for a second think you know what a 'real world' load is supposed to look like. It simply is selective shopping to lower the bar for an OC thats probably better off a few hundred mhz lower. After all, if you arent fully loading the CPU, why even OC it out of its efficiency curve? The practical use is negligible...



Monitor your clocks and power draw then. "Pretty much" says enough. Stop deluding yourself This info on Furmark is very old news that has not changed since.

why does'nt my gpu clock speed go fully up?

hey guys so i have a zotac gtx 1050ti mini + i5 4460 i oc'd it to 1504 mhz on d clock and 2002 mhz on d mem clock but when i stress tested it with furmark, d gpu clock reaches a max of 1341 mhz. is this fine or is something wrong?(i'm a noob) also this computer is powered by a 300w psu... am i sa...

linustechtips.com

Furmark lower FPS and lower score on higher gpu clock and power limit?

Is that due to Nvidia built in Protection, when I run furmark at 113% power limit, it hits around 1750-1800Mhz and gives 200 FPS but when I run 113% power limit, it goes 2000-2040Mhz but at 155 FPS? Palit 3070 Gaming Pro

linustechtips.com

Need more?

I think many do not realise Nvidia definitely ,and AMD also possibly, watch for power viruses especially furmark and limit they're own performance via driver, and have for a while.

Furmark doesn't run unconstrained.

 

[The red spirit]

Monitor your clocks and power draw then. "Pretty much" says enough. Stop deluding yourself This info on Furmark is very old news that has not changed since.

why does'nt my gpu clock speed go fully up?

hey guys so i have a zotac gtx 1050ti mini + i5 4460 i oc'd it to 1504 mhz on d clock and 2002 mhz on d mem clock but when i stress tested it with furmark, d gpu clock reaches a max of 1341 mhz. is this fine or is something wrong?(i'm a noob) also this computer is powered by a 300w psu... am i sa...

linustechtips.com

Furmark lower FPS and lower score on higher gpu clock and power limit?

Is that due to Nvidia built in Protection, when I run furmark at 113% power limit, it hits around 1750-1800Mhz and gives 200 FPS but when I run 113% power limit, it goes 2000-2040Mhz but at 155 FPS? Palit 3070 Gaming Pro

linustechtips.com

Need more?

Yes. Look mate, I modded my own Polaris card vBIOS and no matter what software it runs, it has universal power limits. There's nothing Furmark specific that can hard your GPU. Perhaps old cards, which didn't have AMD PowerTune or nV equivalent, could indeed burn out or be damaged. But that's because they had piss poor cooling and just so so electrical components. And then you would need to go back to pre Radeon HD 6000 series for that. And then again why does Furmark get all shit? MSI Kombustor, OCCT stress test run literally the same thing and people are fine with that.

I think many do not realise Nvidia definitely ,and AMD also possibly, watch for power viruses especially furmark and limit they're own performance via driver, and have for a while.

No, it's through vBIOS. Card OEM sets allowed wattages (yes wattages, there's small power limit, maximum power limit, TDC, TDP) for card.

 

[freeagent]

I think many do not realise Nvidia definitely ,and AMD also possibly, watch for power viruses especially furmark and limit they're own performance via driver, and have for a while.

That's what I thought too..

A couple of years ago when I got my 980 Classified I thought it would be neat to see how much power a 3770K running at 4600MHz static loaded with Linpack Xtreme, and a well clocked 980 Classy can draw. That number was 750w. I didn't believe it, so I just ran the GPU and that was showing 500w at the wall. Scary stuff.

 

[The red spirit]

You might as well go looking for the pot of gold at the end of the rainbow...

You know people said that Amazon's New World also killed cards, when some poorly models kicked the bucket.

 

[AusWolf]

And you claim to know what instructions your CPU will and will not use and how much power it will draw? Interesting

An example: firing up Overwatch utilizes a high performance bit of AVX and loads on multiple threads. It pushes the 8700K to 80+ C here, if only briefly. It was this game that managed to peak over 100C in the earlier example and BSODs for it.

So you might want to draw your own conclusions on that Dont for a second think you know what a 'real world' load is supposed to look like. It simply is selective shopping to lower the bar for an OC thats probably better off a few hundred mhz lower. After all, if you arent fully loading the CPU, why even OC it out of its efficiency curve? The practical use is negligible...

Fair enough - I haven't played Overwatch. The only way I've tested an AVX load is with the CPU-Z benchmark, but that doesn't even use as much power as Cinebench for some reason (just clocks 100 MHz lower because of AVX).

No explanation why. BTW Furmark seems to be similar to MSI Kombustor or OCCT's GPu stress test.

There's no evidence to suggest that. I say to you GPU uses nearly the same amount of watts, gets pretty much as hot as in Heaven. If you have such argument, you gotta have proof.

Yes. Look mate, I modded my own Polaris card vBIOS and no matter what software it runs, it has universal power limits. There's nothing Furmark specific that can hard your GPU. Perhaps old cards, which didn't have AMD PowerTune or nV equivalent, could indeed burn out or be damaged. But that's because they had piss poor cooling and just so so electrical components. And then you would need to go back to pre Radeon HD 6000 series for that. And then again why does Furmark get all shit? MSI Kombustor, OCCT stress test run literally the same thing and people are fine with that.


No, it's through vBIOS. Card OEM sets allowed wattages (yes wattages, there's small power limit, maximum power limit, TDC, TDP) for card.

OK, own example with the 5700 XT: In games (100% load) it ran at about 75 °C (90-95 hot spot) and fluctuated between 2000 and 2050 MHz. When I fired up Furmark, the core hit 90 °C (~100 hot spot), and clocked all the way down to the 1800 MHz region. You can't explain this purely based on power limits. It is not representative of a gaming load at all.

That's where you are wrong. i5 10400F at full load in games and maximum all core boot consumes 50-55 watts. Maximum all core boost is 4GHz. There are plenty of loads that can be executed at maximum boost and fit within 65 watt PL1. I really have to go out of my way to each 80 watts and only prime95 was able to push it to more than 90.

I literally told you that I ran rime95 with PLs lifted (set to 4095). My cooling was able to cope with that, but I don't want my hardware to be able to reach such temperatures (were in low 80s on CPU). Cinebench R23 is still less demanding, so outcome would be the same.

That is strange. Comet Lake isn't very different from Rocket Lake in terms of power/performance. Maybe your motherboard has some better than average voltage distribution, or I don't know.

Edit: Another update: I turned off my case fans, and now the system is completely silent. So it's confirmed: the stock Corsair fans provided with my case don't like being mounted horizontally.

 

[The red spirit]

OK, own example with the 5700 XT: In games (100% load) it ran at about 75 °C (90-95 hot spot) and fluctuated between 2000 and 2050 MHz. When I fired up Furmark, the core hit 90 °C (~100 hot spot), and clocked all the way down to the 1800 MHz region. You can't explain this purely based on power limits. It is not representative of a gaming load at all.

Sure, it is a heavy load, not denying that, but it doesn't do any damage. I ran Furmark on ATi X800 XT Platinum Edition and everything was fine. That card didn't have any throttling and only a small heatsink.

That is strange. Comet Lake isn't very different from Rocket Lake in terms of power/performance. Maybe your motherboard has some better than average voltage distribution, or I don't know.

Nothing strange at all. Rocket Lake consumes more watts per GHz. Also my chip tops out at all core 4GHz and single core 4.3GHz, at that point you are in really bad diminishing return curve already, in fact, for Rocket Lake it is at 3.3-3.7 GHz depending on your bin or maybe even less. Yours can reach higher clock speeds and you have 25% more cores, so there's no wonder that it consumes way more power. It's more surprising, that it doesn't turn into inferno. And no, my motherboard has really basic (poor) VRMs, it's essentially the same VRM from DS3H model and that board uses VRM from Gigabyte's bottom of the barrel boards like S3, DS3 or S2H. I only got fancy semi-functional heatsink.

 

[xtreemchaos]

i like dust its my second fav dirt with fluff being my Number 1.
+ pressure all the way in one of my riggs but it dont matter in my main rigg with it being a P5.

 

[TheoneandonlyMrK]

Yes. Look mate, I modded my own Polaris card vBIOS and no matter what software it runs, it has universal power limits. There's nothing Furmark specific that can hard your GPU. Perhaps old cards, which didn't have AMD PowerTune or nV equivalent, could indeed burn out or be damaged. But that's because they had piss poor cooling and just so so electrical components. And then you would need to go back to pre Radeon HD 6000 series for that. And then again why does Furmark get all shit? MSI Kombustor, OCCT stress test run literally the same thing and people are fine with that.


No, it's through vBIOS. Card OEM sets allowed wattages (yes wattages, there's small power limit, maximum power limit, TDC, TDP) for card.

How the FF are you now trying to make out you know furmark was limited but before were standing by it as a stress test and driver or weva it sees furmark and limits output as I said.

And no the bios sets initial limits, that CAN be defeated via power play tables on AMD on Nvidia they can't be defeated but the DRIVER can limit performance, as for which limits furmark I wouldn't suggest I Know it's the driver, but I sure as shit ain't taking your opinion for anything more than a opinion.



So my actual f£#@&£ point was right, yes.

 

[maxfly]

Yawn
The primal95 and furryskidmark argument. Happy, happy, joy, joy!
There is no set in stone ocing guide everyone is obligated to follow. If your rig, does what you want it to do, the way you want it to do it, when you want it to do it. Rock on! Life is gud.
Just agree to disagree.

Another topic: I couldn't wait until after my holiday, so I did the swap. Radiator in the front, acting as intake, case fans on the top, no dust filters. My first reaction even without thermal testing is that the system just got A LOT louder. I don't know why. My guess is that my case fans don't like being mounted horizontally. They also produce a lot more airflow than the AIO fans on the rad, resulting in a negative pressure situation inside the case. With no dust filters, this won't be good long-term. This is a setup that I already don't like. I'm not even sure if I want to go through thermal testing.

Did you test your gpu thermals/noise with no front filter prior to the swap by chance?