90c+ CPUs

[Upgrayedd]

So now that Intel i9 is 100c and now even amd 7600X is 90c.
What's the overall sentiment about running r5 class cpus at 90+c or any of these new cpus so hot?
For the longest or was always run it cooler or you will degrade your chip.
All of a sudden we're accepting 90C+? No one is worried about the 7900x dying in 3 years right out warranty?

 

[R-T-B]

You run the part in the range that it's rated for. Go outside that range there'll be problems. The engineers know what they are doing.

 

[R0H1T]

Underclock/undervolt & try to optimize for efficiency? No way I'm boiling my water on these not so tasty chips

 

[R-T-B]

Underclock/undervolt & try to optimize for efficiency? No way I'm boiling my water on these not so tasty chips

But my body, the room temp... it burns sir.

Thankfully winter is coming...

 

[oxrufiioxo]

I'm ok with it as long as the CPU makers allow me to tweak the cpu to run at lower temp/wattage if I choose to.

The 7950X is a good example of a cpu that performs good at 65w/105w and at stock and I'm sure with further tweaking enthusiasts will figure out a sweet spot just like 12900k owners have and I'm sure 13900k owners will as well.

 

[tabascosauz]

So now that Intel i9 is 100c and now even amd 7600X is 90c.
What's the overall sentiment about running r5 class cpus at 90+c or any of these new cpus so hot?
For the longest or was always run it cooler or you will degrade your chip.

To be fair, it's been true for a while now even on Vermeer/Alder Lake that the CPU temp has no necessary relationship with the heat you actually feel being dissipated - it's just a product of processes being so dense and difficult to cool nowadays.

I'm much more suspicious of AMD allegedly saying 115C is "in spec" for manual overclocking (especially since ensuring QC and consistent silicon quality is not AMD's strong suit), but 90-95C seems fine to me - Zen 3 has the same limits depending on core count, it's just that the stock boost algo doesn't push itself to make full use of that envelope. Raphael is just pushing harder to max out that space it's been given. Depending on board, AGESA and silicon quality, you can already manipulate PBO on a 5900X/5950X to behave similarly at 200W+. It's only the aggressively conservative throttling (aka power virus recognition) of the stock Zen 3 algorithm that keeps it cool (compare less aggressive throttling on Zen2 to produce 90C+ in stock Prime95, vs. 60-70ish on Zen3).

And mobile CPUs have been running at 90-100C round the clock for close to a decade. In the end you still use the same tricks - limit PPT/EDC, set a static Vcore negative offset, etc. - to influence how your CPU boosts.

My problem is that AMD insanely cranking up the clocks/power to stay competitive has meant that at stock limits there is no sane CPU in the current Raphael lineup. The 12900K is a bit of a challenge for air and the 12900KS is an inferno, but still everything 12600K and lower is easy to cool (probably same for 13600K). Definitely not true for the 7600X.

Gaming temps look fine, however. Similar to Zen3

Maybe I'm a little partial to Intel. If they run their CPUs hot it's because they give generous Vcore to ensure all of their CPUs behave as expected regardless of silicon quality, which gives them hefty undervolt headroom. AMD likes to live on the edge - the result is CPUs that clock far worse than other samples of the same SKU, Cache Hierarchy WHEA at stock, etc. But from how the CPUs seem to clock in reviews, AMD might be adopting a similar approach to be safe, which is a good thing for QA. e.g. see the results for Eco mode 7950X.

No one is worried about the 7900x dying in 3 years right out warranty?

I mean, it's a 3 year warranty.......they're not promising 10 years. For all the bitching about Intel 4C/8T, we got generations of hella long-lasting CPUs......but none of that was promised.

 

[oxrufiioxo]

I'm much more suspicious of AMD allegedly saying 115C is "in spec" for manual overclocking (especially since ensuring QC and consistent silicon quality is not AMD's strong suit), but 90-95C seems fine to me - Zen 3 has the same limits depending on core count, it's just that the stock boost algo doesn't push itself to make full use of that envelope. Raphael is just pushing harder to max out that space it's been given. Depending on board, AGESA and silicon quality, you can already manipulate PBO on a 5900X/5950X to produce 200W+ and similar heat. It's only the aggressively conservative throttling (aka power virus recognition) of the stock Zen 3 algorithm that keeps it cool (compare less aggressive throttling on Zen2 to produce 90C+ in stock Prime95, vs. 60-70ish on Zen3).

Slightly off topic but you being an SFF builder what do you think of how well the 7950X performs at 65w it seems to beat or match a stock 12900k. I want to see more in depth testing of it personally but I still find that impressive and would think people going SFF would be excited about it. A 4090/7950X tweaked to consume less than 450W with open loop cooling could be very impressive in a tiny case like your Cerberus. Obviously throwing Price to performance out the window.

 

[Deleted member 185158]

To be fair, it's been true for a while now even on Vermeer/Alder Lake that the CPU temp has no necessary relationship with the heat you actually feel being dissipated - it's just a product of processes being so dense and difficult to cool nowadays.

I'm much more suspicious of AMD allegedly saying 115C is "in spec" for manual overclocking (especially since ensuring QC and consistent silicon quality is not AMD's strong suit), but 90-95C seems fine to me - Zen 3 has the same limits depending on core count, it's just that the stock boost algo doesn't push itself to make full use of that envelope. Raphael is just pushing harder to max out that space it's been given. Depending on board, AGESA and silicon quality, you can already manipulate PBO on a 5900X/5950X to behave similarly at 200W+. It's only the aggressively conservative throttling (aka power virus recognition) of the stock Zen 3 algorithm that keeps it cool (compare less aggressive throttling on Zen2 to produce 90C+ in stock Prime95, vs. 60-70ish on Zen3).

And mobile CPUs have been running at 90-100C round the clock for close to a decade. In the end you still use the same tricks - limit PPT/EDC, set a static Vcore negative offset, etc. - to influence how your CPU boosts.

My problem is that AMD insanely cranking up the clocks/power to stay competitive has meant that at stock limits there is no sane CPU in the current Raphael lineup. The 12900K is a bit of a challenge for air and the 12900K is an inferno, but still everything 12600K and lower is easy to cool. From the leaked clock specs I don't expect 13600K to be wildly different. Definitely not true for the 7600X.

Gaming temps look fine, however. Similar to Zen3

Maybe I'm a little partial to Intel. If they run their CPUs hot it's because they give generous Vcore to ensure all of their CPUs behave as expected regardless of silicon quality, which gives them hefty undervolt headroom. AMD likes to live on the edge - the result is CPUs that clock far worse than other samples of the same SKU, Cache Hierarchy WHEA at stock, etc. But from how the CPUs seem to clock in reviews, AMD might be adopting a similar approach to be safe, which is a good thing for QA. e.g. see the results for Eco mode 7950X.



I mean, it's a 3 year warranty.......they're not promising 10 years. For all the bitching about Intel 4C/8T, we got generations of hella long-lasting CPUs......but none of that was promised.

You can request from Intel the MTBF of their hardware because it's not listed on their site.

As a rule of thumb, most electronics are MTBF of 100,000 hours of constant use at load. Motherboards and hardware without solid caps will have a shorter expected life span.

But, no. They don't out right provide this information though it is available.

 

[Frick]

The engineers know what they are doing.

They're all on the take from Big Tech. These chips are designed to fail within two years. You read it here first. Mark the date. Sheeple. And so on.

 

[oxrufiioxo]

They're all on the take from Big Tech. These chips are designed to fail within two years. You read it here first. Mark the date. Sheeple. And so on.

So they are meant to fail within warranty costing AMD and Intel millions in replacements?

 

[GerKNG]

we ran CPUs and GPus in the 90s and even 100°C for at least 10 years without problems.
there were these "akimbo style" 8800GT cards that ran at over 100°C non stop (and that's not the hotspot temp)
they still run in these days after over 10 years.

i don't see a problem running CPUs and GPUs in the high 90s.

 

[tabascosauz]

Slightly off topic but you being an SFF builder what do you think of how well the 7950X performs at 65w it seems to beat or match a stock 12900k. I want to see more in depth testing of it personally but I still find that impressive and would think people going SFF would be excited about it. A 4090/7950X tweaked to consume less than 450W with open loop cooling could be very impressive in a tiny case like your Cerberus. Obviously throwing Price to performance out the window.

So far consensus seems to be that it runs just fine at 88W. But in what workloads that is, no idea.

All-core workloads no surprise there, strong performance in Eco mode. The 5GHz+ performance is not impressing me - it's disgustingly inefficient and clearly way way past where the N5 process currently wants to be (as is tradition of AMD asking more clock from the process than TSMC can give).

But because N5 is so comfortable (cool, low power, low Vcore) in the 4-5GHz range, Zen4 keeps so much of its performance at even 65W - which is very impressive and good for SFF who do a lot of creator stuff. Haven't seen any details yet on Fabric power optimization for 7000, but the 4GHz range performance is promising so severely underclocked SFF Raphael might be a better proposition than severely underclocked SFF Vermeer.

In gaming loads I still wouldn't be confident in the 2CCD parts anyways. Already the GN review had at least one instance of poor lows from the 7950X. Only people severely underclocking 5900X/5950X for SFF were the people who can't live without the cores, and those who just want to, so 7000 is probably no different. 2CCD not a great choice for games, throwing money away.

Base clock is high for Raphael as well, might be viable for once to just disable CPB entirely if you can't cool. Lose 600MHz for like 20C drop. Ryzen has never been a clock monster anyway, it only really makes a difference in Cinebench.

 

[Vayra86]

I'm actually more worried about the parts surrounding the CPU more so than the CPU itself.

CPUs are tough motherf... and to be honest, the worst that usually happens is degradation over a long time. That's usually voltage related though, and not heat related.

But heat must move away from the CPU and with these CPUs the increased heat is coming in tandem with higher TDPs. You need to move that heat and that price tag is on yourself. So these CPUs are in fact a lot more pricy than they seem, if you run them at stock.

 

[Sithaer]

we ran CPUs and GPus in the 90s and even 100°C for at least 10 years without problems.
there were these "akimbo style" 8800GT cards that ran at over 100°C non stop (and that's not the hotspot temp)
they still run in these days after over 10 years.

i don't see a problem running CPUs and GPUs in the high 90s.

Funny enough out of the 2 cards that died on me since 2008 was a MSI 8800 GT after 2 and half years, my bro's 8800 GTX died the same and so did 2 of my friend's 9800 GT in the coming years.
Ofc I can't tell what killed them but they all went down one after the other in a few years.

 

[Frick]

So they are meant to fail within warranty costing AMD and Intel millions in replacements?

Yes. Illuminati man. Them globalists and lizards are sucking the life and tech out of us man.

 

[INSTG8R]

Meh my brand new 5800X3D runs to me alarmingly toasty sometimes touching very close to 90 under stress tests, what's another 5C? Ever measured a GPU Hostspot? now THAT'S alarming I wish that sensor never existed...

 

[P4-630]

GPU Hostspot? now THAT'S alarming I wish that sensor never existed...

I was saying that about GPU's sometime ago, in the past there never was a "hotspot" sensor in GPU-Z.
When there was , suddenly alot of people on the net creating threads about it, worried.

The less sensors we know about the better you can sleep haha.

 

[oxrufiioxo]

I'm actually more worried about the parts surrounding the CPU more so than the CPU itself.

CPUs are tough motherf... and to be honest, the worst that usually happens is degradation over a long time. That's usually voltage related though, and not heat related.

But heat must move away from the CPU and with these CPUs the increased heat is coming in tandem with higher TDPs. You need to move that heat and that price tag is on yourself. So these CPUs are in fact a lot more pricy than they seem, if you run them at stock.

The majority of these boards are 8 layer and somewhat overbuilt in the VRM/Heatsink department I doubt they will have any issues at least when it comes to X670E/X670/B650E..... I guess we will see how many corners are cut on the vanilla B650 boards to get to 120 usd though.

 

[Toss]

I run Ryzen 5950x at perfect cool 80C tops, undervolted underclocked to 4 ghz. It's silent, fast, I don't need extra 20%, I am not in a hurry.
Also fans doesn't spin - TDP doesn't exceed 130W TDP even in harderst tasks

 

[freeagent]

I have a 7950X in my cart, have been trying to pull the trigger for a few hours now lol.

Will this be the CPU to drive me away from air cooling?

Thousand bucks man

1023 actually.

Just keep telling myself that is what I paid for my X5690 lol..

 

[oxrufiioxo]

I have a 7950X in my cart, have been trying to pull the trigger for a few hours now lol.

Will this be the CPU to drive me away from air cooling?

Thousand bucks man

1023 actually.

Just keep telling myself that is what I paid for my X5690 lol..

You can do it!!!! although a 7700X is probably better for gaming..... Seems to be already backordered here in the states with multiple expensive motherboards selling out.

 

[freeagent]

You can do it!!!! although a 7700X is probably better for gaming.....

I know

Dammit. I wanna bench too, winter is coming

 

[Space Lynx]

So now that Intel i9 is 100c and now even amd 7600X is 90c.
What's the overall sentiment about running r5 class cpus at 90+c or any of these new cpus so hot?
For the longest or was always run it cooler or you will degrade your chip.
All of a sudden we're accepting 90C+? No one is worried about the 7900x dying in 3 years right out warranty?

for Der8aur zen 4 is only running at 70 celsius after a delid

 

[oxrufiioxo]

I know

Dammit. I wanna bench too, winter is coming

Wait 1000 Cad that's just for tbe cpu right? What is it with a board and ram like 16-1700 cad lol.

 

[tabascosauz]

I have a 7950X in my cart, have been trying to pull the trigger for a few hours now lol.

Same price as 5950X..........but the temps on the 5950X and the performance it offered in 2020 made that a better deal

Will this be the CPU to drive me away from air cooling?

For you? nah, open the window

But the way you bench, you're not going anywhere past stock 230W with air.........maybe get a good block like TechN and see how far the 7950X can clock. Maybe some LM to help it out

for Der8aur zen 4 is only running at 70 celsius after a delid

That's direct die, kinda a lot more complicated than a "delid", needs a frame

I'm not sure if Intel brought back the die-thinning tricks for Alder Lake, seems like AMD completely missed that memo after Comet Lake showcased how much thermal improvement that simple trick could accomplish