Overclock.net banner
16501 - 16520 of 18211 Posts

·
Registered
Joined
·
1,260 Posts
My FTW3 makes 1070.6 GB/s. View attachment 2518322
Hey, that’s awesome. I haven’t adjusted any voltages to my memory yet. But so far, I am at 1,095GB bandwidth +17% on top of the default bandwidth of 936.2. But not a huge difference either way. My memory is pretty toasty on the kingpin Hybrid cooler. The default hybrid cooler doesn’t cool the memory well. it’s essentially passively cooled, unless the fan is turned on the card.

My Kingpin Hydrocopper just arrived today though, and I am hoping I can get much more out of it with additional cooling and attempting to add voltage.

Still haven’t hit a wall as of yet on this memory though. So there’s no telling how high it can go.
 

·
Registered
Joined
·
1,260 Posts
It’s time! I am going to run the 3dmark stress test exactly how the card is first, then I am
going to run it again after installing the Hydro Copper.

I have also been running the memory at +1700 which is totally fine. Maybe additional voltages, and better cooling will get me even further!


 

·
Registered
Joined
·
1,260 Posts
might be two different issues or could be related

the good question would be, why did EVGA decide to use ICX sensors and not Nvidia sensors and why the fuses ?

The chip itself drives the voltage controllers which in turn drive VRM stages and that works fine on reference 2x8pin / FE board, no fuses needed. The boost/load is controlled by Nvidia drivers also :rolleyes:
People can at least rest assured if they’re gpu is an evga. Because they’ll certainly replace it with the 2nd hand warranty, or new warranty.
 

·
Invalid Media
Joined
·
4,787 Posts
...this reminds me a bit of the 2080 Ti 'test escapes' where a bunch of them failed early on...after many on YT falsely blamed Micron etc (which was nonsensical as the regular 2080 had the same VRAM type / memory controller), it turned out that it was indeed NVidia's chip production line which had the aforementioned test escapes - meaning chips that failed QC and should never have been passed on to vendors in the first place.

...this current Amazon World Game issue might clearly have different causes, but when I looked for a MSRP - 3090 in late January, I shied away from the EVGA FTW3 even when I found one and went for the Strix instead as there were other power-related issues in the news re. the EVGA FTW3. Worth noting that the Kingpin has a different PCB and power delivery and I would take that one any day (or night...).

..finally, there just is badly-written software apps out there which can push your hardware over the edge. Theoretically, fail-safes should kick in, especially with a card with fuses and may be that is exactly what happened. Still, this is another one of those annoying 'blame the other vendor' situations for the consumer.
 

·
Facepalm
Joined
·
10,225 Posts
...this reminds me a bit of the 2080 Ti 'test escapes' where a bunch of them failed early on...after many on YT falsely blamed Micron etc (which was nonsensical as the regular 2080 had the same VRAM type / memory controller), it turned out that it was indeed NVidia's chip production line which had the aforementioned test escapes - meaning chips that failed QC and should never have been passed on to vendors in the first place.

...this current Amazon World Game issue might clearly have different causes, but when I looked for a MSRP - 3090 in late January, I shied away from the EVGA FTW3 even when I found one and went for the Strix instead as there were other power-related issues in the news re. the EVGA FTW3. Worth noting that the Kingpin has a different PCB and power delivery and I would take that one any day (or night...).

..finally, there just is badly-written software apps out there which can push your hardware over the edge. Theoretically, fail-safes should kick in, especially with a card with fuses and may be that is exactly what happened. Still, this is another one of those annoying 'blame the other vendor' situations for the consumer.
It was bad or defective solder. More precisely, the VRAM chip next to the PCIE slot that was losing its connection (which is why when the card cooled down, e.g. to be shipped out, it suddenly worked again).
That same chip is the hottest VRAM chip on Ampere cards.
 
  • Rep+
Reactions: J7SC

·
Invalid Media
Joined
·
4,787 Posts
It was bad or defective solder. More precisely, the VRAM chip next to the PCIE slot that was losing its connection (which is why when the card cooled down, e.g. to be shipped out, it suddenly worked again).
That same chip is the hottest VRAM chip on Ampere cards.
...no problem accepting that, though 'bad solder' is still a QC issue at the end of the day. However, I'm steering more towards the 'expectations' of a consumer acting on imperfect information, and this latest Amazon World game beta issue doesn't help
 

·
Registered
Joined
·
3,363 Posts
No active backplate for hydro cooper? 🤔
 

·
Registered
Joined
·
1,027 Posts
Turns out my aorus xtreme card had a really bad soldering job on one of the vrm capacitors and it broke off lol. I soldered it back on but was having issues running benches, turns out it was my bench system with the new benchmark that was unstable and my crappy soldering was fine lol. I really hate when stuff is badly done from the factory.
 

·
Registered
Joined
·
1,260 Posts
My Kingpin 3090 has some writing and dots on the die.

What could 199K mean?



 

·
Registered
Joined
·
2,404 Posts
guess that pretty much confirms they use hand binned dies :unsure:
 

·
Registered
Joined
·
3 Posts

Did anyone try to flash the Gigabyte 3090 Turbo 24G vBIOS with the Gigabyte 3090 Gaming OC one to have a higher board power limit ? is it too risky ?
Both card have almost same PCB and same pin connectors.
Here is the Gaming OC specs Target: 370.0 W Limit: 390.0 W and here is the Turbo 24G specs Target: 350.0 W Limit: 350.0 W ?
I have a watercooling custom loop so my Turbo would handle very well the heat.
best regards :)
 

·
Registered
Joined
·
400 Posts
My Kingpin 3090 has some writing and dots on the die.

What could 199K mean?



Interesting. My KP has same marks. 2 dots on the top left corner. "1" on the top right corner and "1995" on lower right corner.

Could you test your chip with ATI Tool? Make sure to max out the voltage slider and idle temps 22-23c before opening the v/f curve. 1.10v should be 2055-2070MHz at stock. Then add 270-285 to that voltage point and click run. It should hold 2325-2340MHz at 30-31C.
2518439
 

·
Registered
Joined
·
135 Posts
Interesting. My KP has same marks. 2 dots on the top left corner. "1" on the top right corner and "1995" on lower right corner.

Could you test your chip with ATI Tool? Make sure to max out the voltage slider and idle temps 22-23c before opening the v/f curve. 1.10v should be 2055-2070MHz at stock. Then add 270-285 to that voltage point and click run. It should hold 2325-2340MHz at 30-31C.
View attachment 2518439
Yeah I reckon this confirms they’re noting factory tested overclocks.

Out of interest what’s your “effective clock” with that curve, using HWInfo? I’ve found setting a curve like that Afterburner misrepresents the real (aka effective) clock speed I’m achieving by as much as 100-150mhz.
 
16501 - 16520 of 18211 Posts
Top