Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

What is recommended operating temp of HBM?

WhiteSkyMage
 Share

Hey guys, my HBM2 memory on Vega would sometimes go like 85C. Is this normal? It's on watercooling on a 280mm rad. I have bought a new rad, but it hasn't come yet. I also ordered a few sensors.

 

Question is, what temperature is dangerous for High Bandwidth Memory to be running at and what is optimal to aim for?

Intel Core i9-9900K | Asrock Phantom Gaming miniITX Z390 | 32GB GSkill Trident Z DDR4@3600MHz C17 | EVGA RTX 3090 FTW3 Watercooled | Samsung 970 EVO 1TB M.2 SSD | Crucial MX500 2TB SSD | Seasonic Focus Plus Gold 1000W | anidees AI Crystal Cube White V2 | Corsair M95 | Corsair K50 | Beyerdynamic DT770 Pros 250Ohm

Link to comment
Share on other sites

Link to post
Share on other sites

i think anything above 80*C makes hbm memory go unstable , it could just cause instability , damage would occur at 120*C+ probably with a minimum of 110 +-

 

optimal for hbm is basicly lower than 80*C because above it reduces ur oc / stability capacity

(◑‿◐)

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, Valkyrie Lenneth said:

i think anything above 80*C makes hbm memory go unstable , it could just cause instability , damage would occur at 120*C+ probably with a minimum of 110 +-

ok, well I had a "hottest spot" on my GPU at 101C once. However HBM2 only reached 90C and it was still stable.

You think it's normal that GPU die is less hot than the actual HBM?

Intel Core i9-9900K | Asrock Phantom Gaming miniITX Z390 | 32GB GSkill Trident Z DDR4@3600MHz C17 | EVGA RTX 3090 FTW3 Watercooled | Samsung 970 EVO 1TB M.2 SSD | Crucial MX500 2TB SSD | Seasonic Focus Plus Gold 1000W | anidees AI Crystal Cube White V2 | Corsair M95 | Corsair K50 | Beyerdynamic DT770 Pros 250Ohm

Link to comment
Share on other sites

Link to post
Share on other sites

2 minutes ago, WhiteSkyMage said:

ok, well I had a "hottest spot" on my GPU at 101C once. However HBM2 only reached 90C and it was still stable.

You think it's normal that GPU die is less hot than the actual HBM?

not sure, that depends on the chip itself, however its going to degrade ur chip running it at 95+ for longer periods x)

(◑‿◐)

Link to comment
Share on other sites

Link to post
Share on other sites

That's an EK full cover, right?  Not sure how many people here have a Vega on water, so I'd get in touch with EK to ask if it's normal. 

Link to comment
Share on other sites

Link to post
Share on other sites

3 minutes ago, Valkyrie Lenneth said:

i think anything above 80*C makes hbm memory go unstable , it could just cause instability , damage would occur at 120*C+ probably with a minimum of 110 +-

Do you have a source for that? I couldn't find much information on HBM safe/stable temperatures. GDDR can stand some pretty high temperatures, but HBM might be different.

 

@WhiteSkyMage If you're on water, perhaps you should make sure that the waterblock's thermal pads are making proper contact with the vram chips and gpu, check that the pump is running properly and the fans are working. Safe or not, it probably shouldn't get that hot with a custom loop. I would be more worried with the gpu running at 100C than with HBM running at 85, the gpu is generally more vulnerable to heat.

Don't ask to ask, just ask... please 🤨

sudo chmod -R 000 /*

What is scaling and how does it work? Asus PB287Q unboxing! Console alternatives :D Watch Netflix with Kodi on Arch Linux Sharing folders over the internet using SSH Beginner's Guide To LTT (by iamdarkyoshi)

Sauron'stm Product Scores:

Spoiler

Just a list of my personal scores for some products, in no particular order, with brief comments. I just got the idea to do them so they aren't many for now :)

Don't take these as complete reviews or final truths - they are just my personal impressions on products I may or may not have used, summed up in a couple of sentences and a rough score. All scores take into account the unit's price and time of release, heavily so, therefore don't expect absolute performance to be reflected here.

 

-Lenovo Thinkpad X220 - [8/10]

Spoiler

A durable and reliable machine that is relatively lightweight, has all the hardware it needs to never feel sluggish and has a great IPS matte screen. Downsides are mostly due to its age, most notably the screen resolution of 1366x768 and usb 2.0 ports.

 

-Apple Macbook (2015) - [Garbage -/10]

Spoiler

From my perspective, this product has no redeeming factors given its price and the competition. It is underpowered, overpriced, impractical due to its single port and is made redundant even by Apple's own iPad pro line.

 

-OnePlus X - [7/10]

Spoiler

A good phone for the price. It does everything I (and most people) need without being sluggish and has no particularly bad flaws. The lack of recent software updates and relatively barebones feature kit (most notably the lack of 5GHz wifi, biometric sensors and backlight for the capacitive buttons) prevent it from being exceptional.

 

-Microsoft Surface Book 2 - [Garbage - -/10]

Spoiler

Overpriced and rushed, offers nothing notable compared to the competition, doesn't come with an adequate charger despite the premium price. Worse than the Macbook for not even offering the small plus sides of having macOS. Buy a Razer Blade if you want high performance in a (relatively) light package.

 

-Intel Core i7 2600/k - [9/10]

Spoiler

Quite possibly Intel's best product launch ever. It had all the bleeding edge features of the time, it came with a very significant performance improvement over its predecessor and it had a soldered heatspreader, allowing for efficient cooling and great overclocking. Even the "locked" version could be overclocked through the multiplier within (quite reasonable) limits.

 

-Apple iPad Pro - [5/10]

Spoiler

A pretty good product, sunk by its price (plus the extra cost of the physical keyboard and the pencil). Buy it if you don't mind the Apple tax and are looking for a very light office machine with an excellent digitizer. Particularly good for rich students. Bad for cheap tinkerers like myself.

 

 

Link to comment
Share on other sites

Link to post
Share on other sites

12 minutes ago, Sauron said:

Do you have a source for that? I couldn't find much information on HBM safe/stable temperatures. GDDR can stand some pretty high temperatures, but HBM might be different.

 

@WhiteSkyMage If you're on water, perhaps you should make sure that the waterblock's thermal pads are making proper contact with the vram chips and gpu, check that the pump is running properly and the fans are working. Safe or not, it probably shouldn't get that hot with a custom loop. I would be more worried with the gpu running at 100C than with HBM running at 85, the gpu is generally more vulnerable to heat.

Junction operating temperature -40˚C to 125˚C    but it was shown that above 80*C its where hbm becomes less stable ( with overclocked chips )

 

damage wont really occur from 100*C on the hbm but 110-120 can damage/degrade it overtime :) even if its within the operating temperature

 

those are just statistics from quite a few oc,s :)

 

sauce for all is

 

https://www.esilicon.com/wp-content/uploads/esilicon-high-bandwidth-memory-14lpp-hbm-gen2-phy-brochure-web-20160406.pdf

 

but its just the standard information, nothing about stability runs ect

(◑‿◐)

Link to comment
Share on other sites

Link to post
Share on other sites

3 minutes ago, Valkyrie Lenneth said:

Junction operating temperature -40˚C to 125˚C    but it was shown that above 80*C its where hbm becomes less stable ( with overclocked chips )

Interesting. I wouldn't necessarily base myself on overclocked hardware when it comes to stability though, temperatures is just as likely to be the culprit as the overclock itself in that scenario.

 

Either way those temperatures on water are not normal.

Don't ask to ask, just ask... please 🤨

sudo chmod -R 000 /*

What is scaling and how does it work? Asus PB287Q unboxing! Console alternatives :D Watch Netflix with Kodi on Arch Linux Sharing folders over the internet using SSH Beginner's Guide To LTT (by iamdarkyoshi)

Sauron'stm Product Scores:

Spoiler

Just a list of my personal scores for some products, in no particular order, with brief comments. I just got the idea to do them so they aren't many for now :)

Don't take these as complete reviews or final truths - they are just my personal impressions on products I may or may not have used, summed up in a couple of sentences and a rough score. All scores take into account the unit's price and time of release, heavily so, therefore don't expect absolute performance to be reflected here.

 

-Lenovo Thinkpad X220 - [8/10]

Spoiler

A durable and reliable machine that is relatively lightweight, has all the hardware it needs to never feel sluggish and has a great IPS matte screen. Downsides are mostly due to its age, most notably the screen resolution of 1366x768 and usb 2.0 ports.

 

-Apple Macbook (2015) - [Garbage -/10]

Spoiler

From my perspective, this product has no redeeming factors given its price and the competition. It is underpowered, overpriced, impractical due to its single port and is made redundant even by Apple's own iPad pro line.

 

-OnePlus X - [7/10]

Spoiler

A good phone for the price. It does everything I (and most people) need without being sluggish and has no particularly bad flaws. The lack of recent software updates and relatively barebones feature kit (most notably the lack of 5GHz wifi, biometric sensors and backlight for the capacitive buttons) prevent it from being exceptional.

 

-Microsoft Surface Book 2 - [Garbage - -/10]

Spoiler

Overpriced and rushed, offers nothing notable compared to the competition, doesn't come with an adequate charger despite the premium price. Worse than the Macbook for not even offering the small plus sides of having macOS. Buy a Razer Blade if you want high performance in a (relatively) light package.

 

-Intel Core i7 2600/k - [9/10]

Spoiler

Quite possibly Intel's best product launch ever. It had all the bleeding edge features of the time, it came with a very significant performance improvement over its predecessor and it had a soldered heatspreader, allowing for efficient cooling and great overclocking. Even the "locked" version could be overclocked through the multiplier within (quite reasonable) limits.

 

-Apple iPad Pro - [5/10]

Spoiler

A pretty good product, sunk by its price (plus the extra cost of the physical keyboard and the pencil). Buy it if you don't mind the Apple tax and are looking for a very light office machine with an excellent digitizer. Particularly good for rich students. Bad for cheap tinkerers like myself.

 

 

Link to comment
Share on other sites

Link to post
Share on other sites

4 minutes ago, Sauron said:

Interesting. I wouldn't necessarily base myself on overclocked hardware when it comes to stability though, temperatures is just as likely to be the culprit as the overclock itself in that scenario.

 

Either way those temperatures on water are not normal.

depends if he had thermal pads on the memory ect :) and if hes using a 280rad for hes entire system + ocing hes cpu it could be normal because vega gets hot af

(◑‿◐)

Link to comment
Share on other sites

Link to post
Share on other sites

14 hours ago, Sauron said:

Do you have a source for that? I couldn't find much information on HBM safe/stable temperatures. GDDR can stand some pretty high temperatures, but HBM might be different.

 

@WhiteSkyMage If you're on water, perhaps you should make sure that the waterblock's thermal pads are making proper contact with the vram chips and gpu, check that the pump is running properly and the fans are working. Safe or not, it probably shouldn't get that hot with a custom loop. I would be more worried with the gpu running at 100C than with HBM running at 85, the gpu is generally more vulnerable to heat.

alright. Well  the GPU is running at max of 70C (because I am using the LC Vega BIOS, which does NOT allow higher temps than 70C). The HBM2 however goes like 85C. And no, there are no PADS for VRAM. VRAM is HBM2, and it's next to the die (i used thermal compound to cover the die and the HBM2). Go and see what the Vega package looks like, cuz maybe you think that I am with GDDR5.

 

13 hours ago, Valkyrie Lenneth said:

depends if he had thermal pads on the memory ect :) and if hes using a 280rad for hes entire system + ocing hes cpu it could be normal because vega gets hot af

Totally true. I have OCed my CPU to 4.1GHz. However I have NOT stressed both CPU and Vega, because I am afraid that this may damage the CPU and Vega die. I have to tell you that Vega 64, when running a benchmark actually HEATS my CPU (I am pretty sure that water goes up to 60C or maybe even higher than that. So when Vega is running fully, CPU heats up to 80C with ONLY 1 core actually running at 4.1GHz and others being idle. Clearly, the radiator is overwhelmed with heat. I tried touching it, and it's VERY HOT!

 

I have ordered a new 240mm radiator and it should be coming in 2 days (hopefully). My question was, will HBM2 degrade at 85C?

WP_20170924_14_45_37_Pro.jpg

Intel Core i9-9900K | Asrock Phantom Gaming miniITX Z390 | 32GB GSkill Trident Z DDR4@3600MHz C17 | EVGA RTX 3090 FTW3 Watercooled | Samsung 970 EVO 1TB M.2 SSD | Crucial MX500 2TB SSD | Seasonic Focus Plus Gold 1000W | anidees AI Crystal Cube White V2 | Corsair M95 | Corsair K50 | Beyerdynamic DT770 Pros 250Ohm

Link to comment
Share on other sites

Link to post
Share on other sites

dont touch it, fingergrease can cause shorts xD , ur chip wont degrade at 85*C that will start at 110+ but it will become less stable to hotter u go after 80*C

 

also the blue pads u got on there, they can be swapped for thermal pads for ur water block :) usually they come with the waterblocks o,o ( like the silver ones ) theheck the blueones tho o,o

 

 

u dont really have to worry about the temps much but it will affect ur max oc ALOT if u go above 80*C basicly :) if u can keep the hbm below 80*C  u could clock way higher

 

 

u might wanna swap out the thermal pads just buy a sheet of thermal pad for gpu and cut them proper x)

 

also if u want a custom loop for both CPU and gpu i always recommend getting a 240mm for ur cpu and then a 120mm rad for ur gpu alone x) example > 240mm to cpu then to 120mm to gpu x) that way ur gpu would get nice cool coolant also without beeing heat up from the cpu first

 

^ is just my personal recommendation,  one single 120mm rad ( thick preferred ) for gpu and 240mm rad for cpu ( doesnt have to be extreme thick but need ssome thickness )

 

the gpu at max 70*C with a oc is pretty decent btw, just ur hbm memory is running quite hot ( probably dont have good thermal pads / wkm rate low

(◑‿◐)

Link to comment
Share on other sites

Link to post
Share on other sites

2 hours ago, Valkyrie Lenneth said:

dont touch it, fingergrease can cause shorts xD , ur chip wont degrade at 85*C that will start at 110+ but it will become less stable to hotter u go after 80*C

 

also the blue pads u got on there, they can be swapped for thermal pads for ur water block :) usually they come with the waterblocks o,o ( like the silver ones ) theheck the blueones tho o,o

 

 

u dont really have to worry about the temps much but it will affect ur max oc ALOT if u go above 80*C basicly :) if u can keep the hbm below 80*C  u could clock way higher

 

 

u might wanna swap out the thermal pads just buy a sheet of thermal pad for gpu and cut them proper x)

 

also if u want a custom loop for both CPU and gpu i always recommend getting a 240mm for ur cpu and then a 120mm rad for ur gpu alone x) example > 240mm to cpu then to 120mm to gpu x) that way ur gpu would get nice cool coolant also without beeing heat up from the cpu first

 

^ is just my personal recommendation,  one single 120mm rad ( thick preferred ) for gpu and 240mm rad for cpu ( doesnt have to be extreme thick but need ssome thickness )

 

the gpu at max 70*C with a oc is pretty decent btw, just ur hbm memory is running quite hot ( probably dont have good thermal pads / wkm rate low

I think you don't really get which cooler I have. The Swiftech H240X is an AIO cooler that came with 280mm radiator/res/pump combo and Apogee XL waterblock (for my CPU). What I did was connect the GPU to that loop (i.e. i extended it). Unfortunately that 280mm radiator is NOT able to handle the heat from both CPU and GPU. I ordered a 240mm radiator to put on the front of my case. You can see all my PC specs in the signature below. Because not all 240mm radiators can fit in the front of my case, I bought a 45mm thick radiator that is 276mm in length from Magicool. I am sure that this would fit well... Basically I will have a 280mm and a 240mm radiator (idk if I need another 120mm in the back?).

 

The thermal pads (in blue) came WITH my waterblock from EK. My thermal compound is from EKWB, so i trust it's good.

Intel Core i9-9900K | Asrock Phantom Gaming miniITX Z390 | 32GB GSkill Trident Z DDR4@3600MHz C17 | EVGA RTX 3090 FTW3 Watercooled | Samsung 970 EVO 1TB M.2 SSD | Crucial MX500 2TB SSD | Seasonic Focus Plus Gold 1000W | anidees AI Crystal Cube White V2 | Corsair M95 | Corsair K50 | Beyerdynamic DT770 Pros 250Ohm

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share


×