Intel's CPU can work up to 100C before throttling, that's a LOT of heat.
No it's not. It's quite a standard amount of heat that CPU's have experienced since the days of Pentium 3 and the first generation of Athlons. Additionally most of this temperature does not couple into the motherboard where even if your die is at 100C you will read barely more than a 20C rise on the board itself. On the otherhand VRMs do couple heat directly into the motherboard and they are rated to 150C though most of them will sensibly throttle at around 130C mainly to prevent cooking other components on the mainboard.
I've also seen people over-tighten them, cracking the chip or having liquid leak all over.
Wow. Someone has been buying coolers from aliexpress. There isn't a waterblock currently on the market where the instructions don't say to just do up the bolts as hard as you can. They all bottom out and use a spring based retention system.
Crack the chip? I call bullshit. You will do damage to every other component on the board long before you crack a CPU with an IHS on it. Now theres plenty of people who take the IHS off, but your assertion is that we're talking about people who don't know what they are doing. And yes if you're building a custom loop there's risk of leaks, but by your own admission you're worried about incompetence, so why target the enthusiast? An AIO doesn't leak.
they will dissipate the heat into the tiny watercooler sink which then becomes a radiator to the neighboring areas, melting out capacitors and various other things.
I'm sorry but physics doesn't work like that. You can radiate from the entire surface of the CPU without cooling and your neighboring components will barely be warm to the touch at the pathetically cool 100C where the CPU will throttle. See the top of my post. 100C is a trivial temperature to work with and we've worked with it for 20 years now.
The thermal cycles of a modern CPU are no different to those from 10 years ago, we ran them up to 100C back then, we run them up to 100C now. The difference is that a modern CPU getting to 100C on die will cause it to protect itself whereas the old ones happily let the smoke come out.
TDP limit is a figure given to designers of cooling systems and my ancient Pentium 4 had thermal cycles almost identical to my Threadripper (which has 4x the TDP). Oh and my Pentium 4 still works 15 years later.
And yes, people will install the chip the wrong way, overtighten the cooling, not applying or too much cooling paste (resulting in uneven heat dissipation) or somehow destroy the chip and simply return it to the store and ask for their money back.
And? That applies to the cooler on a Celeron as well and has nothing to do with high TDP.
The GPU is a lot larger
No it's not. The GPU is firstly direct die and is smaller than a heatspreadder. On top of that the cooler is required to connect to multiple parts of the card. If you're worried about enthusiasts destroying any component while installing a heatsink, then the GPU is orders of magnitude more likely to be ruined.
They also have the heatsink already assembled in most cases and removing it voids the warranty.
You do realise that installing the heatsink on e.g. a modern 250W TDP TR4 chip or an LGA 3647 chip is no different than installing the fan on your old Pentium 3 right?
Oh wait it is different. The modern chips have heat spreaders making it far less likely to damage the CPU while installing the cooler, and modern coolers have pretensioned retaining systems meaning you just throw on 4 bolts and call it a day. Then there's the advent of AIOs which makes the component sitting on your CPU significantly smaller and even easier to install than a heatsink. Seriously have you not built a computer in the last 10 years? The cooler on my Pentium 4 was significantly more difficult to install than the cooler on my current 180W TDP (rated, it actually draws closer to 230W overclocked) cooler.
Errr no. The Roman Empire fell due to idiots running the place, poor military decisions, and ultimately was overthrown rather than collapsing in on itself. The Roman empire was a lot of things but it was far from bureaucratic by any modern standards.
Now let me quote something I read on the internet to you:
That may be surprisingly ineffective at actual cooling. The radiators in a house are designed to create natural draft and in order to do that you need a very high temperature differential between the water loop and the ambient air. That is what you get when you have a vertical fins running lengthwise through a radiator. Additionally most depend on convection and radiate poorly thanks to the default colour choice of off-white and the thick layer of lacquer covering the metal.
Mind you it will still work. The amount of water in the loop needed to fill a large radiator like that can store a lot of thermal energy and thus it's unlikely a computer would move the temperature much over time. Even with poor thermal properties it may take an hour+ for the water temperature to do anything. However you may be better off pumping water through a large metal bucket.:-)
5GHz is not even close to the record extreme overclocks have stably achieved on CoffeeLake. 7.3GHz is getting close to the limit where no amount of magic can make it run stably anymore.
Could that be because you're a good developer and not a shady shit working for Facebook? It's not really a new term. There's even been a website around that publishes a continuously rolling hall of shame for the past 10 years: https://darkpatterns.org/
One can't be expected to know every term used in every field, or even in one field. For being called "hip" this website has been around for nearly a decade calling out and shaming shady practices: https://darkpatterns.org/
When has a government collapsed under the weight of bureaucracy? Normally it's the governments that eliminate bureaucracy altogether that quickly collapse.
That last "everything went batshit" reboot might have been caused by DRAM corruption.
Interestingly not only did you just move the goalposts you made them as wide as the entire frigging field. I never said DRAM doesn't get corrupted, I said "filesystem". Your memory is chock full of data so in the incredibly remote chance that you actually suffer from an error that ECC detects you have an additionally incredibly remote chance that you happen to corrupt data in flight on the way to being written on the disk. The vast majority of these already incredibly rare events will as you say cause a "batshit reboot" or a crash, something of incredibly little significance to an average desktop user.
Fortunately we can check how much value we get out of our ECC: #ipmitool sel list 1 | 01/23/2013 | 22:43:13 | Processor | FRB2/Hang in POST failure | Asserted 2 | 08/18/2015 | 10:43:12 | Fan #0x0f | Lower Non-critical going low | Asserted 3 | 08/19/2015 | 14:18:44 | Fan #0x0f | Lower Non-critical going low | Deasserted 4 | 10/08/2018 | 09:21:35 | Fan #0x11 | Lower Non-critical going low | Asserted 5 | 10/08/2018 | 09:22:44 | Fan #0x11 | Lower Non-critical going low | Deasserted
Hmmm there's a distinct lack of memory correction events in my IPMI log on this server since it's continuous operation in 2013. What a waste of money and performance that was.
Now speaking of AMD, their Zen platform is highly sensitive to RAM timings for performance. Care to point me to a 3400MHz ECC RAM module? Or am I thanking AMD for tanking the performance of my workstation?
No. I'm saying these two are by far the most effective. The alternatives are not as good at protecting against UVB and don't last as long.
Your argument also makes no sense. Humans in general don't focus on collective good of their medical fitness alone, and decisions are imperfect targeting a specific and often changing purpose. Key West's decision had nothing to do with cancer, much less causing people to get it (that is also a really dumb stretch of logic since the type of sunscreen used doesn't "cause" anything). Key West made an environmental decision, nothing more.
Now in other leaps of logic Donald Trump ran on a campaign of causing cancer and lung disease, clearly not on keeping coal jobs or appeasing interest groups. See how silly it sounds?
I hope they do, but AMD has a lot to prove on the IPC spec. They promised to catch up with Intel with every generation and while they've made leaps and bounds they still aren't there yet.
Normal humans don't want their massive filesystems to be corruption-free.:rolleyes:
Normal humans practically never experience corruption on their filesystem. If you want to get people to care about something they first need to be affected by it.
Why? 255W TDP is nothing exciting. You can handle that easily with a large air cooler or a standard 240mm AIO. I have a 360mm radiator in my system happily handling the 430W TDP of my combined CPU + GPU with ease. I've only managed an 11C temperature rise in the water under synthetic stressing of the GPU or CPU, not even individual benchmarks were able to get the temperature to rise that much.
I understand they don't want to give this one to the masses.
What "masses" spend in excess of $2000 on a CPU and don't understand or look into the cooling solution?
While we're at it, what 1990s era CPU is capable of being damaged due to poor cooling? Thermal throttling is a thing that occurs on a millisecond basis and failure for a CPU to protect itself these days would be grounds for returning it as being defective and not fit for purpose.
I know Intel wants to give AMD the finger but this is 1/4 kW in the processor alone.
Multiple Threadrippers have a 250W TDP, as do many Xeon chips. This isn't anything new. It's also nothing difficult to cool. There are plenty of air coolers on the market rated in excess of 250W TDP which people put on a large variety of CPUs. Anyone who has ever done any overclocking is capable of cooling a 255W TDP chip. You can easily cool a chip like this with most 240mm AIOs just fine too.
There's only so much you can offload on the GPU. If you have a 6 year old GPU and throw a Quadro at it you're going to have a pretty bad day, even in Photoshop or Premier. I'm currently running an encode offloaded on hardware. My hardware video encoder is only sitting at around 60% many thanks to one of my CPU cores being pegged.
You're quite right about the law of diminishing returns for gaming though. The difference between a 9700K and an 8700K is only a few FPS. But hey if that matters to you then more power to you.
That said the benefit I got upgrading my CPU was support for NVMe SSDs. That has improved I/O performance on my machine by an order of magnitude. But if that's important to you then you wouldn't be chosing Intel with it's crappy PCIe lanes and the audacity to force you to buy a dongle that you plug into a PCIe slot in order to enable VROC.
With a 255-watt power draw, you know it's overvolted and overclocked.
Overvolted and overclocked imply you're pushing something beyond it's vendor spec. This is the vendor spec so by definition not "overclocked". In case you don't know basically every processor on the market is binned so by your standard every processor on the market except for the slowest model in every line is "overclocked".
Slashdot Mirror
the only way to safely use C
I know. Our firmware should be coded by highschoolers using Rust. Then it'll be 100% bug free and safe.
Intel's CPU can work up to 100C before throttling, that's a LOT of heat.
No it's not. It's quite a standard amount of heat that CPU's have experienced since the days of Pentium 3 and the first generation of Athlons. Additionally most of this temperature does not couple into the motherboard where even if your die is at 100C you will read barely more than a 20C rise on the board itself. On the otherhand VRMs do couple heat directly into the motherboard and they are rated to 150C though most of them will sensibly throttle at around 130C mainly to prevent cooking other components on the mainboard.
I've also seen people over-tighten them, cracking the chip or having liquid leak all over.
Wow. Someone has been buying coolers from aliexpress. There isn't a waterblock currently on the market where the instructions don't say to just do up the bolts as hard as you can. They all bottom out and use a spring based retention system.
Crack the chip? I call bullshit. You will do damage to every other component on the board long before you crack a CPU with an IHS on it. Now theres plenty of people who take the IHS off, but your assertion is that we're talking about people who don't know what they are doing. And yes if you're building a custom loop there's risk of leaks, but by your own admission you're worried about incompetence, so why target the enthusiast? An AIO doesn't leak.
they will dissipate the heat into the tiny watercooler sink which then becomes a radiator to the neighboring areas, melting out capacitors and various other things.
I'm sorry but physics doesn't work like that. You can radiate from the entire surface of the CPU without cooling and your neighboring components will barely be warm to the touch at the pathetically cool 100C where the CPU will throttle. See the top of my post. 100C is a trivial temperature to work with and we've worked with it for 20 years now.
The thermal cycles of a modern CPU are no different to those from 10 years ago, we ran them up to 100C back then, we run them up to 100C now. The difference is that a modern CPU getting to 100C on die will cause it to protect itself whereas the old ones happily let the smoke come out.
TDP limit is a figure given to designers of cooling systems and my ancient Pentium 4 had thermal cycles almost identical to my Threadripper (which has 4x the TDP). Oh and my Pentium 4 still works 15 years later.
And yes, people will install the chip the wrong way, overtighten the cooling, not applying or too much cooling paste (resulting in uneven heat dissipation) or somehow destroy the chip and simply return it to the store and ask for their money back.
And? That applies to the cooler on a Celeron as well and has nothing to do with high TDP.
The GPU is a lot larger
No it's not. The GPU is firstly direct die and is smaller than a heatspreadder. On top of that the cooler is required to connect to multiple parts of the card. If you're worried about enthusiasts destroying any component while installing a heatsink, then the GPU is orders of magnitude more likely to be ruined.
They also have the heatsink already assembled in most cases and removing it voids the warranty.
You do realise that installing the heatsink on e.g. a modern 250W TDP TR4 chip or an LGA 3647 chip is no different than installing the fan on your old Pentium 3 right?
Oh wait it is different. The modern chips have heat spreaders making it far less likely to damage the CPU while installing the cooler, and modern coolers have pretensioned retaining systems meaning you just throw on 4 bolts and call it a day. Then there's the advent of AIOs which makes the component sitting on your CPU significantly smaller and even easier to install than a heatsink. Seriously have you not built a computer in the last 10 years? The cooler on my Pentium 4 was significantly more difficult to install than the cooler on my current 180W TDP (rated, it actually draws closer to 230W overclocked) cooler.
Well played :)
Errr no. The Roman Empire fell due to idiots running the place, poor military decisions, and ultimately was overthrown rather than collapsing in on itself. The Roman empire was a lot of things but it was far from bureaucratic by any modern standards.
Now let me quote something I read on the internet to you:
Now stop pretending that you know stuff.
That may be surprisingly ineffective at actual cooling. The radiators in a house are designed to create natural draft and in order to do that you need a very high temperature differential between the water loop and the ambient air. That is what you get when you have a vertical fins running lengthwise through a radiator. Additionally most depend on convection and radiate poorly thanks to the default colour choice of off-white and the thick layer of lacquer covering the metal.
Mind you it will still work. The amount of water in the loop needed to fill a large radiator like that can store a lot of thermal energy and thus it's unlikely a computer would move the temperature much over time. Even with poor thermal properties it may take an hour+ for the water temperature to do anything. However you may be better off pumping water through a large metal bucket. :-)
5GHz is not even close to the record extreme overclocks have stably achieved on CoffeeLake. 7.3GHz is getting close to the limit where no amount of magic can make it run stably anymore.
The RTX 2080 TI is a 250 watt TDP chip
So are the Threadrippers and many Xeon chips.
Could that be because you're a good developer and not a shady shit working for Facebook? It's not really a new term. There's even been a website around that publishes a continuously rolling hall of shame for the past 10 years: https://darkpatterns.org/
One can't be expected to know every term used in every field, or even in one field. For being called "hip" this website has been around for nearly a decade calling out and shaming shady practices: https://darkpatterns.org/
Or just ... you know ... not use the screenshot function?
When has a government collapsed under the weight of bureaucracy? Normally it's the governments that eliminate bureaucracy altogether that quickly collapse.
The penalty is expected to be much larger than the $22.5 million fine the agency imposed on Google in 2012.
And this is why US Regulators are an international joke.
That last "everything went batshit" reboot might have been caused by DRAM corruption.
Interestingly not only did you just move the goalposts you made them as wide as the entire frigging field. I never said DRAM doesn't get corrupted, I said "filesystem". Your memory is chock full of data so in the incredibly remote chance that you actually suffer from an error that ECC detects you have an additionally incredibly remote chance that you happen to corrupt data in flight on the way to being written on the disk. The vast majority of these already incredibly rare events will as you say cause a "batshit reboot" or a crash, something of incredibly little significance to an average desktop user.
Fortunately we can check how much value we get out of our ECC:
#ipmitool sel list
1 | 01/23/2013 | 22:43:13 | Processor | FRB2/Hang in POST failure | Asserted
2 | 08/18/2015 | 10:43:12 | Fan #0x0f | Lower Non-critical going low | Asserted
3 | 08/19/2015 | 14:18:44 | Fan #0x0f | Lower Non-critical going low | Deasserted
4 | 10/08/2018 | 09:21:35 | Fan #0x11 | Lower Non-critical going low | Asserted
5 | 10/08/2018 | 09:22:44 | Fan #0x11 | Lower Non-critical going low | Deasserted
Hmmm there's a distinct lack of memory correction events in my IPMI log on this server since it's continuous operation in 2013. What a waste of money and performance that was.
Now speaking of AMD, their Zen platform is highly sensitive to RAM timings for performance. Care to point me to a 3400MHz ECC RAM module? Or am I thanking AMD for tanking the performance of my workstation?
No. I'm saying these two are by far the most effective. The alternatives are not as good at protecting against UVB and don't last as long.
Your argument also makes no sense. Humans in general don't focus on collective good of their medical fitness alone, and decisions are imperfect targeting a specific and often changing purpose. Key West's decision had nothing to do with cancer, much less causing people to get it (that is also a really dumb stretch of logic since the type of sunscreen used doesn't "cause" anything). Key West made an environmental decision, nothing more.
Now in other leaps of logic Donald Trump ran on a campaign of causing cancer and lung disease, clearly not on keeping coal jobs or appeasing interest groups. See how silly it sounds?
What is a $2000+ processor if not a tool for a very frigging specific job?
The other reason is to enrich government
Given the state of USA infrastructure and national debt this is not a bad thing.
AMD 7nm chips should change that
I hope they do, but AMD has a lot to prove on the IPC spec. They promised to catch up with Intel with every generation and while they've made leaps and bounds they still aren't there yet.
You do hit a point where it's cheaper to build a 2-socket server system for the needed performance
What needed performance? Show us a Xeon with the single core performance of this chip.
Normal humans don't want their massive filesystems to be corruption-free. :rolleyes:
Normal humans practically never experience corruption on their filesystem. If you want to get people to care about something they first need to be affected by it.
Why? 255W TDP is nothing exciting. You can handle that easily with a large air cooler or a standard 240mm AIO. I have a 360mm radiator in my system happily handling the 430W TDP of my combined CPU + GPU with ease. I've only managed an 11C temperature rise in the water under synthetic stressing of the GPU or CPU, not even individual benchmarks were able to get the temperature to rise that much.
I understand they don't want to give this one to the masses.
What "masses" spend in excess of $2000 on a CPU and don't understand or look into the cooling solution?
While we're at it, what 1990s era CPU is capable of being damaged due to poor cooling? Thermal throttling is a thing that occurs on a millisecond basis and failure for a CPU to protect itself these days would be grounds for returning it as being defective and not fit for purpose.
I know Intel wants to give AMD the finger but this is 1/4 kW in the processor alone.
Multiple Threadrippers have a 250W TDP, as do many Xeon chips. This isn't anything new. It's also nothing difficult to cool. There are plenty of air coolers on the market rated in excess of 250W TDP which people put on a large variety of CPUs. Anyone who has ever done any overclocking is capable of cooling a 255W TDP chip. You can easily cool a chip like this with most 240mm AIOs just fine too.
There's only so much you can offload on the GPU. If you have a 6 year old GPU and throw a Quadro at it you're going to have a pretty bad day, even in Photoshop or Premier. I'm currently running an encode offloaded on hardware. My hardware video encoder is only sitting at around 60% many thanks to one of my CPU cores being pegged.
You're quite right about the law of diminishing returns for gaming though. The difference between a 9700K and an 8700K is only a few FPS. But hey if that matters to you then more power to you.
That said the benefit I got upgrading my CPU was support for NVMe SSDs. That has improved I/O performance on my machine by an order of magnitude. But if that's important to you then you wouldn't be chosing Intel with it's crappy PCIe lanes and the audacity to force you to buy a dongle that you plug into a PCIe slot in order to enable VROC.
With a 255-watt power draw, you know it's overvolted and overclocked.
Overvolted and overclocked imply you're pushing something beyond it's vendor spec. This is the vendor spec so by definition not "overclocked". In case you don't know basically every processor on the market is binned so by your standard every processor on the market except for the slowest model in every line is "overclocked".