Switched from AS5 to Thermaltake TG-7. Huge difference!

[Baumber12]

About 4 years ago I bought my Corsair H60 and a tube of AS5. Going from the original hsf to a water-cooled system was a huge drop in temps. I was able to overclock my i5 3570k to 4.4ghz 24/7. Everything was going great. Games like ARMA 3 almost doubled my FPS with the overclock. Then I got my 980ti and upgraded to 1440p. So happy!

2 years later, I was experiencing slowly increasing temperatures and less than perfect stability. With age, my computer required higher voltages to keep my 4.4ghz OC, but the temperatures were getting to be too much. So, I dropped down to 4.2ghz at 1.2v. Games still ran good but it was noticeable in hugely cpu intensive games. But temps were good again; not ever going above 65c.

Now, 2 more years pass and my temps are up to 75-80c. Absolutely no hardware changes. Me, thinking its just my cpu degrading, didn't do anything about it... until now! Just today I opened my beautiful Bitfenix Colossus case, cleaned the already dust free fans, cleaned my radiator, and even pulled out my cpu to clean it. Unfortunately, I had misplaced my AS5 tube and was forced to make a trip to the local bestbuy to pick up a tube of Thermaltake TG-7. At first it was only to be a place holder until I had the time to purchase and re-paste some Thermal Grizzly Kryonaut. But what happened next came as a huge shock.

The results:
After I got everything set back up and plugged my computer in, I was shocked! Running Prime95, my temps were at 80c using AS5. But after repasting my cpu with Thermaltake TG-7, my max temps dropped to 64c. A 16c drop in temperature! I was expecting maybe 1-5c, but no, 16c!

I know what you're thinking, my previous application must have been awful or something along those lines. But no, it was an excellent application. Both times I used the X method and both times brought great returns. I've used both the Pea and X methods and firmly believe the X brings lower temps. With my 3570k, I used the same application method for both pastes. So it's hard to describe what happened here. Either it was the age of the AS5 that caused it lose some of its effectiveness, or the new Thermaltake TG-7 is really an amazing thermal paste. Either way, I know exactly what I'm using from now on.

 

[EarthDog]

Difference in pastes is actually around 1-2c with emperical testing. The difference between the best pastes (not tg-7) is up to ~3c.

Clearly something else was going on... but, awesome drops!

 

[Ruru]

I used AS5 in 2005 or something, I just don't think it's that good anymore..

I prefer MX-4, GC-Extreme or Liquid Ultra these days.

 

[Vya Domus]

TIM was dry as hell. That's most likely the explanation.

 

[EarthDog]

Newest quality roundup...

http://www.tomshardware.com/reviews/thermal-paste-comparison,5108-8.html

 

[eidairaman1]

AS5 takes time to burn in.

 

[EarthDog]

AS5 takes time to burn in.

for that 1-2c woooo!

 

[Nuckles56]

AS5 takes time to burn in.

4 years might be a bit to long for burn in

 

[Baumber12]

So far I have it running stable at 4.4ghz 1.275v. My chip definitely isn't the best but it gets the job done. On a side note, i am thinking about changing just my cpu to either a 3770k for $240 or a 2600k for $140. After overclocking both of them, would they give me better performance in games?

 

[jboydgolfer]

On a side note, i am thinking about changing just my cpu to either a 3770k for $240 or a 2600k for $140. After overclocking both of them, would they give me better performance in games?

if your current CPU is a 3570k, your not going to see a in game improvement to (IMO)justify a $240 purchase, neither the 3770, or the 2600 will do much. my guess would be a max improvement of around 5FPS (resolution & title dependent OFC)

I know the draw for eight threads is strong, but trust me on this one, you're not going to see an improvement that will be noticeable in any environment outside of benchmarks or specific (thread hungry) programs.

Edit
If you have hundreds to sidegrade a cpu, id recommend instead that you save up & upgrade out of that older chipset to a newer 1150 or equivalent current tech build

 

[Ruru]

Delidding would be a huge improvement too.

 

[eidairaman1]

4 years might be a bit to long for burn in

On my AXP it took a few months lol

 

[natr0n]

Diamond powder will be better than most pastes. Last time I used that type is would etch up metal when you would reapply it.

I might try this out though. Youtube videos show it's thinner than the old stuff I tried(was like clay).

Gonna head to bestbuy and get a tube tomorrow. =D

 

[EarthDog]

Diamond powder will be better than most pastes. Last time I used that type is would etch up metal when you would reapply it.

I might try this out though. Youtube videos show it's thinner than the old stuff I tried(was like clay).

consider reading my link earlier in the thread and see where the diamond paste landed.

Its in the top third, so ill go with 'most', but it isnt the beesknees..

 

[Mr.Scott]

Diamond powder will be better than most pastes. Last time I used that type is would etch up metal when you would reapply it.

I might try this out though. Youtube videos show it's thinner than the old stuff I tried(was like clay).

The thin stuff is no better. Still abrasive as hell.

 

[Vario]

I've tried a lot of different pastes but I prefer AS5. It seems to last a long time and it doesn't pump out with thermal cycles. They all seem to have the same temperatures or within 1-2 degrees.

For example, Techpowerup's own review of the popular NT-H1 vs Arctic Silver 5
https://www.techpowerup.com/reviews/Noctua/NT-H1/4.html

Its essentially identical. I find that AS5 just seems to last longer before I have to remount, holds up better to moving the case around without having the TIM shift and it it doesn't thin out or pump out over time.

Granted, in the OP's case, 4 years might be pushing it. I think any TIM might start degrading by that time. That is likely why his overclock became less stable.

 

[Baumber12]

I just got 4.6ghz stable at 1.34v. Might try for 1.335v soon. When I first applied AS5 over 4 years ago, my temps would hit 90+ when trying to get 4.5ghz stable at 1.3v. It's amazing what this little tube has been able to accomplish.

As I said before, earlier today I tested prime95 with as5 and got 80c with a 4.2ghz 1.2v overclock. Now I'm able to pull off 4.6ghz with 1.34v at 86c. During gaming it doesn'g go above 79c.

 

[EarthDog]

Good job. Dry paste as was called out earlier is what likely happened. Only a couple c between pastes... so, yeah.

Go get em.

 

[Vayra86]

I just got 4.6ghz stable at 1.34v. Might try for 1.335v soon. When I first applied AS5 over 4 years ago, my temps would hit 90+ when trying to get 4.5ghz stable at 1.3v. It's amazing what this little tube has been able to accomplish.

As I said before, earlier today I tested prime95 with as5 and got 80c with a 4.2ghz 1.2v overclock. Now I'm able to pull off 4.6ghz with 1.34v at 86c. During gaming it doesn'g go above 79c.

Be aware gaming temps are a good 10-15 C lower than stress testing. Don't use them as a baseline. If you want a good game to stress test on, TW:Warhammer. Hits all cores with varying loads, and pretty intensive at that. But only in addition to normal stress tests. 1.335v is pretty high already for this CPU, I wouldn't push it beyond 1.35v regardless of temp.

 

[John Naylor]

Since Tomshardware was purchased by the Purch Group in 2014, it's more in line with Purch's focus on product placement then a tech review site. My go to site for ZTIM data is benchmarkreviews and tho the article linkes is several years old, not much has changed other then Grizzly products. Also many products deteriote over time and many take time to burn in... not a fan of As5 for the reasons stated on their web site

"While much safer than electrically conductive silver and copper greases, Arctic Silver 5 should be kept away from electrical traces, pins, and leads. While it is not electrically conductive, the compound is very slightly capacitive and could potentially cause problems if it bridges two close-proximity electrical paths."

"Due to the unique shape and sizes of the particles in Arctic Silver 5's conductive matrix, it will take a up to 200 hours and several thermal cycles to achieve maximum particle to particle thermal conduction and for the heatsink to CPU interface to reach maximum conductivity."

Prime95 has no place really since Sandy Bridge unless ya goal is to get listed on OC web site leader boards:

a) You can test with the new versions which include modern instruction sets but that can be hazardous to your CPU
b) You can test with the older version but all that proves is that your CPU is stable ... as long as no modern instruction sets are present. kinda like testing the 4 WD off road / snow capability of your new jeep on a Florida state highway.
c) It tests your CPU at heat loads which it will never ever see again so what is the point ? Your 4.4 GHz under p95 might be a 4.6 Ghz at the same temps when faced with realistic loadings

If ya bought ya CPU to run programs / games then the best stablity testing tool to use is on which simulates such usage in a multitasking environment. I have had 24 hours stable P95 Ocs fail within 45 minutes under RoG Real Bench which, instead of single task heavy loads on each core. places varying tasks on each core at a level that your CPU will never see again in its lifetime. While doing a better job of stability testing, it comes no where near the temps experiences under synthetic benchmarks allowing you to get a more stable and higher OC

I generally use Shin Etsu G751 for CPus and Gelid Extreme for GPUs. The 751 is a bit better and at only $3 is priced ridiculously as they make it. Most everyone is a rebrand of stuff packaged for the PC market and made by industrial concerns like Skin Ertso and Dow. Hoerver with GPus, having to apply to some 20+ surfaces GPU, all the memory and VRM ships, often on two sides, Shin Etsu doesn't remain as workable and Gelid's longer pliability when exposed to air is helpful. Grizzly Kryonaugh is prolly the top non liquid metal TIM out there but its ridiculous expensive and hard to find

As for range of temps (about 4.5C) between TIMs....in this test went from GC Electronics HTC 10-8135's 42.10°C to Shin-Etsu MicroSi G751's 37.55°C
https://archive.benchmarkreviews.co...&id=150&Itemid=99999999&limit=1&limitstart=12

Application methods also vary by type of heat sink
https://archive.benchmarkreviews.co...ntent&task=view&id=170&Itemid=38&limitstart=5


As we see above, the difference between TIMs is greater than 1-2C (4.5) tho I don't see the logic as to why this is considered insignificant. You might see just a 1-2C improvement going from a $35 cooler to a $90 cooler. In fact some cheaper coolers like the $37 Scyther Mugen max and $45 Scythe Fuma actually outperform some $90 flagship models from Cryotek and Noctua

 

[xkm1948]

Should have gone thermal grizzly kryonaut

https://www.techpowerup.com/forums/...-vs-conductonaut-liquid-metal-73-w-mk.229167/

This is NOT the LM one. Having this on my D15 now, pretty good paste even comparing to noctua nt-h1

 

[Mr.Scott]

Prime95 has no place really since Sandy Bridge unless ya goal is to get listed on OC web site leader boards:

If you think you P95 to get listed on leader boards, you have never been on one.
The lead scores are never prime stable, I can assure you.

 

[Vayra86]

Since Tomshardware was purchased by the Purch Group in 2014, it's more in line with Purch's focus on product placement then a tech review site. My go to site for ZTIM data is benchmarkreviews and tho the article linkes is several years old, not much has changed other then Grizzly products. Also many products deteriote over time and many take time to burn in... not a fan of As5 for the reasons stated on their web site

"While much safer than electrically conductive silver and copper greases, Arctic Silver 5 should be kept away from electrical traces, pins, and leads. While it is not electrically conductive, the compound is very slightly capacitive and could potentially cause problems if it bridges two close-proximity electrical paths."

"Due to the unique shape and sizes of the particles in Arctic Silver 5's conductive matrix, it will take a up to 200 hours and several thermal cycles to achieve maximum particle to particle thermal conduction and for the heatsink to CPU interface to reach maximum conductivity."

Prime95 has no place really since Sandy Bridge unless ya goal is to get listed on OC web site leader boards:

a) You can test with the new versions which include modern instruction sets but that can be hazardous to your CPU
b) You can test with the older version but all that proves is that your CPU is stable ... as long as no modern instruction sets are present. kinda like testing the 4 WD off road / snow capability of your new jeep on a Florida state highway.
c) It tests your CPU at heat loads which it will never ever see again so what is the point ? Your 4.4 GHz under p95 might be a 4.6 Ghz at the same temps when faced with realistic loadings

If ya bought ya CPU to run programs / games then the best stablity testing tool to use is on which simulates such usage in a multitasking environment. I have had 24 hours stable P95 Ocs fail within 45 minutes under RoG Real Bench which, instead of single task heavy loads on each core. places varying tasks on each core at a level that your CPU will never see again in its lifetime. While doing a better job of stability testing, it comes no where near the temps experiences under synthetic benchmarks allowing you to get a more stable and higher OC

I generally use Shin Etsu G751 for CPus and Gelid Extreme for GPUs. The 751 is a bit better and at only $3 is priced ridiculously as they make it. Most everyone is a rebrand of stuff packaged for the PC market and made by industrial concerns like Skin Ertso and Dow. Hoerver with GPus, having to apply to some 20+ surfaces GPU, all the memory and VRM ships, often on two sides, Shin Etsu doesn't remain as workable and Gelid's longer pliability when exposed to air is helpful. Grizzly Kryonaugh is prolly the top non liquid metal TIM out there but its ridiculous expensive and hard to find

As for range of temps (about 4.5C) between TIMs....in this test went from GC Electronics HTC 10-8135's 42.10°C to Shin-Etsu MicroSi G751's 37.55°C
https://archive.benchmarkreviews.co...&id=150&Itemid=99999999&limit=1&limitstart=12

Application methods also vary by type of heat sink
https://archive.benchmarkreviews.co...ntent&task=view&id=170&Itemid=38&limitstart=5


As we see above, the difference between TIMs is greater than 1-2C (4.5) tho I don't see the logic as to why this is considered insignificant. You might see just a 1-2C improvement going from a $35 cooler to a $90 cooler. In fact some cheaper coolers like the $37 Scyther Mugen max and $45 Scythe Fuma actually outperform some $90 flagship models from Cryotek and Noctua

The effects of paste are quite overrated. First and foremost is a good mount, and a cooler capable of dissipating the TDP you want to burn. After that its down to details that come at high price and/or effort and highly varying effect; a delid being most effective for temps, followed by case airflow, and only then paste - 4-5 C is the range you will be talking about from the best to the baseline of what's available. Its not something that will make or break any OC. It just adds up to all the other measures but at the same time its the only thing you can save money and effort on - on all the other elements of good cooling, you can not.