TR Forums

I find that a rock-solid 50+ FPS is fine. I don't game competitively and I usually game with an adult beverage or a few.

Curious if it's an A or a non-A chip like how the 2070 is apparently being segmented.

techguy wrote:

Jigar wrote:

Memory cooling seems to be issue with Turing cards, i would suggest you install extra fan to cool this card.

No, it doesn't.

https://www.gamersnexus.net/guides/3394 ... ck-screens

Never ever conclude just by reading one review site analysis.

Here read the following site that hasn't signed the NDA (5 year contract) with Nvidia.

https://www.hardocp.com/article/2018/11 ... r_8_hours/

Below you will find user's also providing their input.

https://hardforum.com/threads/rtx-2080- ... h.1972149/

https://www.reddit.com/r/nvidia/comment ... _dying_in/

jihadjoe wrote:

Curious if it's an A or a non-A chip like how the 2070 is apparently being segmented.

Likewise, but I won't be taking it apart unless I decide to put a water cooler on it down the road a bit. I'm also curious to know what the actual boost clock it settles on in-game is and I'm curious how much higher using Scanner to OC it will take it (and, later on, how the water cooler changes things). So far, I haven't seen a thread anywhere else about the 11G-P4-2281-KR from someone who ordered it with fast shipping. This may be the place to read about how the first "cheap" 2080 Ti stacks up.

I recently received a package direct from EVGA:




It's going into a new build this weekend. Yeehaw!

Nice! Of course, those were back in stock on Tuesday afternoon and I felt a tinge of regret for not waiting but, come on, I'm going to play at 2560x1080, a couple hundred MHz isn't going to wreck my experience.

drfish wrote:

Nice! Of course, those were back in stock on Tuesday afternoon and I felt a tinge of regret for not waiting but, come on, I'm going to play at 2560x1080, a couple hundred MHz isn't going to wreck my experience.

As someone once said, "Fish, you idiot, you should have waited!"

Creating/being the being the butt of that niche-meme will always make whatever poor timing I have worth it.

Jigar wrote:

techguy wrote:

Jigar wrote:

Memory cooling seems to be issue with Turing cards, i would suggest you install extra fan to cool this card.

No, it doesn't.

https://www.gamersnexus.net/guides/3394 ... ck-screens

Never ever conclude just by reading one review site analysis.

Here read the following site that hasn't signed the NDA (5 year contract) with Nvidia.

https://www.hardocp.com/article/2018/11 ... r_8_hours/

Below you will find user's also providing their input.

https://hardforum.com/threads/rtx-2080- ... h.1972149/

https://www.reddit.com/r/nvidia/comment ... _dying_in/

So you take issue with my citation of a single "professional" source by doing the same? One that has clear biases, at that. Kyle Bennett is on Nvidia's blacklist because he's a drama queen. The NDA nonsense is just his latest mountain out of a molehill.

That's not a rebuttal either. Show me TardOCP's (or anyone else's) extensive, conclusive, unbiased investigation of the matter. The fact is, no one else in the tech journalism industry has done what Steve from Gamer's Nexus has done to investigate this issue. He called for users to send him dead cards and they responded in droves. The sample size he's working with is much larger than TardOCP's 3.

So you want to take Kyle "drama queen" Bennett's opinion over cold, hard facts. Be my guest. I won't waste time trying to convince you otherwise.

The "limited test escapes" issue appears to be isolated to FE cards from Nvidia.

I follow @EVGA_JacobF on Twitter so I got a heads up on upcoming availability prior to being auto notified. I was ready to pounce when the buy button appeared.

https://twitter.com/EVGA_JacobF

Yeah, he posted about the FTW3 being back in stock on Tuesday a couple hours after I ordered the Black on Monday.

I'm looking forward to seeing how the boost clocks and temps on our cards compare. Jeff let me know this about the FE card he has (in Unigine Heaven):

with power and thermal limits maxed and fans at 100% the FE seems to want to hang out around 1965-1980 MHz

We'll see...

drfish wrote:

Creating/being the being the butt of that niche-meme will always make whatever poor timing I have worth it.

Hey, I bought a 7700K like four months before Coffee Lake launched.
And then a 7700HQ laptop like a month before mobile Coffee Lake launched.

Zero regrets (except design flaws with said laptop - damn you, Lenovo!).

Kretschmer wrote:

drfish wrote:

Creating/being the being the butt of that niche-meme will always make whatever poor timing I have worth it.

Hey, I bought a 7700K like four months before Coffee Lake launched.
And then a 7700HQ laptop like a month before mobile Coffee Lake launched.

Zero regrets (except design flaws with said laptop - damn you, Lenovo!).

I bought a 7700k in January 2017, still a few months before Coffee Lake but in the grand scheme of things, not too far apart. I was somewhat kicking myself after Coffee Lake launch but my 7700k is a golden sample - 5.2GHz at 1.39V 24x7 AVX stable no less. Absolute beauty of a chip, probably my best result in the "silicon lottery" over the past couple decades. I'm going to hold out until Comet Lake or whatever Intel puts out next. Maybe Ryzen 3000 if they can close the gap in IPC + clocks.

techguy wrote:

Jigar wrote:

techguy wrote:

No, it doesn't.

https://www.gamersnexus.net/guides/3394 ... ck-screens

Never ever conclude just by reading one review site analysis.

Here read the following site that hasn't signed the NDA (5 year contract) with Nvidia.

https://www.hardocp.com/article/2018/11 ... r_8_hours/

Below you will find user's also providing their input.

https://hardforum.com/threads/rtx-2080- ... h.1972149/

https://www.reddit.com/r/nvidia/comment ... _dying_in/

So you take issue with my citation of a single "professional" source by doing the same? One that has clear biases, at that. Kyle Bennett is on Nvidia's blacklist because he's a drama queen. The NDA nonsense is just his latest mountain out of a molehill.

That's not a rebuttal either. Show me TardOCP's (or anyone else's) extensive, conclusive, unbiased investigation of the matter. The fact is, no one else in the tech journalism industry has done what Steve from Gamer's Nexus has done to investigate this issue. He called for users to send him dead cards and they responded in droves. The sample size he's working with is much larger than TardOCP's 3.

So you want to take Kyle "drama queen" Bennett's opinion over cold, hard facts. Be my guest. I won't waste time trying to convince you otherwise.

I told you before, i am telling you again. Never ever take one source as end of story. I presented to you with user views as well but you conveniently ignored it.

Anyway below are the links that might help.

News - https://www.digitaltrends.com/computing ... rds-dying/

More users RTX card dead https://forums.geforce.com/default/topi ... vely-die-/


Reviewer's card dead (Ignoring HardOCP 3 dead cards since you don't trust it)

Joker production - https://www.youtube.com/watch?v=ewiFtgfK0KY


UFD Tech - https://www.youtube.com/watch?v=P18d6uLZ4qk

Hardware Unboxed - https://www.youtube.com/watch?v=z75wDmwfIJY

Jigar wrote:

I told you before, i am telling you again. Never ever take one source as end of story. I presented to you with user views as well but you conveniently ignored it.

Thanks for caring so damn much.

When my RTX card dies you will be the first to know.

techguy wrote:

Kretschmer wrote:

drfish wrote:

Creating/being the being the butt of that niche-meme will always make whatever poor timing I have worth it.

Hey, I bought a 7700K like four months before Coffee Lake launched.
And then a 7700HQ laptop like a month before mobile Coffee Lake launched.

Zero regrets (except design flaws with said laptop - damn you, Lenovo!).

I bought a 7700k in January 2017, still a few months before Coffee Lake but in the grand scheme of things, not too far apart. I was somewhat kicking myself after Coffee Lake launch but my 7700k is a golden sample - 5.2GHz at 1.39V 24x7 AVX stable no less. Absolute beauty of a chip, probably my best result in the "silicon lottery" over the past couple decades. I'm going to hold out until Comet Lake or whatever Intel puts out next. Maybe Ryzen 3000 if they can close the gap in IPC + clocks.

I was really kicking myself until I saw how close a 7700K and 8700K perform when both use fast RAM. I don't really feel constrained by my 7700K for games. It'd be nice to have a few extra cores for photo processing, but overnight import and processing jobs would make the return-on-upgrade pretty niche.

What kind of cooling and/or chip modifications did you throw at that 7700K? And does it still switch to lower power states? A 15% boost is nothing to sneeze at but isn't enough to justify replacing my existing 140mm radiator.

Kretschmer wrote:

techguy wrote:

Kretschmer wrote:

Hey, I bought a 7700K like four months before Coffee Lake launched.
And then a 7700HQ laptop like a month before mobile Coffee Lake launched.

Zero regrets (except design flaws with said laptop - damn you, Lenovo!).

I bought a 7700k in January 2017, still a few months before Coffee Lake but in the grand scheme of things, not too far apart. I was somewhat kicking myself after Coffee Lake launch but my 7700k is a golden sample - 5.2GHz at 1.39V 24x7 AVX stable no less. Absolute beauty of a chip, probably my best result in the "silicon lottery" over the past couple decades. I'm going to hold out until Comet Lake or whatever Intel puts out next. Maybe Ryzen 3000 if they can close the gap in IPC + clocks.

I was really kicking myself until I saw how close a 7700K and 8700K perform when both use fast RAM. I don't really feel constrained by my 7700K for games. It'd be nice to have a few extra cores for photo processing, but overnight import and processing jobs would make the return-on-upgrade pretty niche.

What kind of cooling and/or chip modifications did you throw at that 7700K? And does it still switch to lower power states? A 15% boost is nothing to sneeze at but isn't enough to justify replacing my existing 140mm radiator.

I de-lidded it and applied liquid metal, run a custom water loop with 2 radiators (a 360 and a 140). Running the RAM at 3600MHZ C16 (just XMP). It was a nice bump over the 4790k @ 5GHz it replaced (also de-lidded and on the same custom loop). I don't bother with offset voltage and also disable low power modes so the chip is running full-out all the time. The system is only on for a few hours at a time, and only maybe once or twice a week so its effective lifespan should still be quite long.

I started de-lidding when Ivy Bridge was released and am still doing it today. Since Intel seems to have forgotten how to solder, I would likely still do so with any future soldered chip I purchase (except AMD). For me, it's worth the effort since I do flight sim which is primarily bound by single thread performance. An extra couple hundred MHz can be just enough to smooth out performance in the worst case scenarios. Probably not worth it to most people though.

techguy wrote:

Kretschmer wrote:

techguy wrote:

I bought a 7700k in January 2017, still a few months before Coffee Lake but in the grand scheme of things, not too far apart. I was somewhat kicking myself after Coffee Lake launch but my 7700k is a golden sample - 5.2GHz at 1.39V 24x7 AVX stable no less. Absolute beauty of a chip, probably my best result in the "silicon lottery" over the past couple decades. I'm going to hold out until Comet Lake or whatever Intel puts out next. Maybe Ryzen 3000 if they can close the gap in IPC + clocks.

I was really kicking myself until I saw how close a 7700K and 8700K perform when both use fast RAM. I don't really feel constrained by my 7700K for games. It'd be nice to have a few extra cores for photo processing, but overnight import and processing jobs would make the return-on-upgrade pretty niche.

What kind of cooling and/or chip modifications did you throw at that 7700K? And does it still switch to lower power states? A 15% boost is nothing to sneeze at but isn't enough to justify replacing my existing 140mm radiator.

I de-lidded it and applied liquid metal, run a custom water loop with 2 radiators (a 360 and a 140). Running the RAM at 3600MHZ C16 (just XMP). It was a nice bump over the 4790k @ 5GHz it replaced (also de-lidded and on the same custom loop). I don't bother with offset voltage and also disable low power modes so the chip is running full-out all the time. The system is only on for a few hours at a time, and only maybe once or twice a week so its effective lifespan should still be quite long.

I started de-lidding when Ivy Bridge was released and am still doing it today. Since Intel seems to have forgotten how to solder, I would likely still do so with any future soldered chip I purchase (except AMD). For me, it's worth the effort since I do flight sim which is primarily bound by single thread performance. An extra couple hundred MHz can be just enough to smooth out performance in the worst case scenarios. Probably not worth it to most people though.

Intel didn't forget how to solder IHS though. The thermal issues with Ivy Bridge/first batch Haswell chips have never been from TIM. They were from QC issues with IHS mounting at the fab. Soldering just made IHS fitting/mounting more consistent at the fab.

For me:

Just Cause 4
Battlefield V
Far Cry 5
Shadow of the Tomb Raider
Wolfenstein II: The New Colossus
Star Wars Battlefront II
Doom

Krogoth wrote:

techguy wrote:

Kretschmer wrote:

I was really kicking myself until I saw how close a 7700K and 8700K perform when both use fast RAM. I don't really feel constrained by my 7700K for games. It'd be nice to have a few extra cores for photo processing, but overnight import and processing jobs would make the return-on-upgrade pretty niche.

What kind of cooling and/or chip modifications did you throw at that 7700K? And does it still switch to lower power states? A 15% boost is nothing to sneeze at but isn't enough to justify replacing my existing 140mm radiator.

I de-lidded it and applied liquid metal, run a custom water loop with 2 radiators (a 360 and a 140). Running the RAM at 3600MHZ C16 (just XMP). It was a nice bump over the 4790k @ 5GHz it replaced (also de-lidded and on the same custom loop). I don't bother with offset voltage and also disable low power modes so the chip is running full-out all the time. The system is only on for a few hours at a time, and only maybe once or twice a week so its effective lifespan should still be quite long.

I started de-lidding when Ivy Bridge was released and am still doing it today. Since Intel seems to have forgotten how to solder, I would likely still do so with any future soldered chip I purchase (except AMD). For me, it's worth the effort since I do flight sim which is primarily bound by single thread performance. An extra couple hundred MHz can be just enough to smooth out performance in the worst case scenarios. Probably not worth it to most people though.

Intel didn't forget to solder though. The thermal issues with Ivy Bridge/first batch Haswell chips have never been from TIM. They were from QC issues with IHS mounting at the fab. Soldering just made IHS fitting/mounting more consistent at the fab.

I've seen you make comments along these lines several times in the news post comments, and have attempted to correct you on your misconceptions about the cause of poor thermal conductivity for Intel desktop microprocessors post-Sandy Bridge. I've personally de-lidded somewhere in the range of 10 CPUs now for myself and several friends/clients since Ivy Bridge. The cause is not singular. The manufacturing process post-Sandy Bridge, having omitted solder TIM, necessitated the use of an adhesive to affix the IHS to the PCB. This introduced a measurable gap which needed to be filled. Since paste TIM has quite poor thermal conductivity (relative to solder TIM or liquid metal), the copious amounts required to thermally connect the underside of the IHS with the CPU die acted to insulate heat. The problem, therefore, is two-fold.

The solution is to use a vastly superior TIM (liquid metal) AND reduce the IHS gap by eliminating the adhesive (epoxy). Case-in-point: Intel themselves *claimed* to have corrected this problem with the Devil's Canyon generation of desktop processors through the use of higher quality paste TIM AND a reduction in IHS z-height. I actually went out and bought a 4790k to upgrade from a de-lidded 4770k, and to test out Intel's claims. Guess what: there was still a gap, and stock performance was only marginally improved over Haswell. De-lid to the rescue!

I've also de-lidded Skylake and Kaby Lake generation CPUs which, admittedly, have lower IHS z-height values than pre-Devil's Canyon chips but once again, still suffer from poor stock thermal conductivity.

So here's the thing about de-lidding. There's not much point to it if you:
1) don't overclock (or OC very little)
2) are still thermally-constrained by inadequate cooling (like air or AIO water)
So there are lots and lots of results out there where people show thermal improvement of only 5-10C under these conditions. These results are not relevant. Just like testing "CPU performance" in games with ultra settings at 4k resolution, the test conditions do not expose the full potential, so it is wrong to draw conclusions from this data.

It always has been an IHS fitting problem. People who had simply de-lid and reseated the IHS on Ivy Bridge/first batch Haswell chips got massive returns. Changing TIM yield far smaller returns. The problem was never the TIM. The packing tools and equipment were tuned to soldered chips. Intel mostly fixed it with Devil Canyon and newer silicon. You can still returns on newer chips but you'll get far more by replacing the stock TIM. The only reason to de-lid is if you want to run the chips naked or use a better TIM/liquid metal then the stock stuff.

Krogoth wrote:

It always has been an IHS fitting problem. The packing tools and equipment were tuned to soldered chips until Intel mostly fixed it with Devil's Canyon and newer silicon. The only reason to de-lid is if you want to run the chips naked or use a better TIM/liquid metal then the stock stuff.

You've made a specific claim and reiterated it above, that Devil's Canyon fixed the thermal issues - it didn't. I had a 4770k, then I bought a 4790k - stock to stock it was a marginal improvement. But once I de-lidded with liquid metal the temps dropped drastically. Not as much as Ivy Bridge, granted, thanks to the horrendous Z-height issue on those chips, but still a massive difference (in the 20C range under load when overclocked). Your theory is invalidated by simple testing.

So again, if you want to run stock with light workloads you're not going to see much of a difference, and those test conditions aren't relevant as they don't push the chips.

techguy wrote:

Krogoth wrote:

It always has been an IHS fitting problem. The packing tools and equipment were tuned to soldered chips until Intel mostly fixed it with Devil's Canyon and newer silicon. The only reason to de-lid is if you want to run the chips naked or use a better TIM/liquid metal then the stock stuff.

You've made a specific claim and reiterated it above, that Devil's Canyon fixed the thermal issues - it didn't. I had a 4770k, then I bought a 4790k - stock to stock it was a marginal improvement. But once I de-lidded with liquid metal the temps dropped drastically. Not as much as Ivy Bridge, granted, thanks to the horrendous Z-height issue on those chips, but still a massive difference (in the 20C range under load when overclocked). Your theory is invalidated by simple testing.

So again, if you want to run stock with light workloads you're not going to see much of a difference, and those test conditions aren't relevant as they don't push the chips.

Intel did fixed the IHS fitting QC problems though. Your first-batch Haswell (4770K) was most likely a lucky unit that had proper IHS fitting at the fab. I have an Ivy Bridge (3570K) that was one of the lucky ones. I have seen countless others who weren't as fortunate. They saw difference of ~10-20C difference between cores under a full load which suggested an IHS fitting problem/air-gap. De-lidding and reseating the IHS yielded massive returns on their chips.

Krogoth wrote:

techguy wrote:

Krogoth wrote:

It always has been an IHS fitting problem. The packing tools and equipment were tuned to soldered chips until Intel mostly fixed it with Devil's Canyon and newer silicon. The only reason to de-lid is if you want to run the chips naked or use a better TIM/liquid metal then the stock stuff.

You've made a specific claim and reiterated it above, that Devil's Canyon fixed the thermal issues - it didn't. I had a 4770k, then I bought a 4790k - stock to stock it was a marginal improvement. But once I de-lidded with liquid metal the temps dropped drastically. Not as much as Ivy Bridge, granted, thanks to the horrendous Z-height issue on those chips, but still a massive difference (in the 20C range under load when overclocked). Your theory is invalidated by simple testing.

So again, if you want to run stock with light workloads you're not going to see much of a difference, and those test conditions aren't relevant as they don't push the chips.

Intel did fixed the IHS fitting QC problems though. Your first-batch Haswell (4770K) was most likely a lucky unit that had proper IHS fitting at the fab. I have an Ivy Bridge (3570K) that was one of the lucky ones. I have seen countless others who weren't as fortunate. They saw difference of ~10-20C difference between cores under a full load which suggested an IHS fitting problem/air-gap. De-lidding and reseating the IHS yielded massive returns on their chips.

I've de-lidded 3 Ivy Bridge CPUs, they all saw massive improvements with liquid metal.

How was my Haswell a "lucky" unit with "proper fitting" if replacing the TIM with liquid metal dropped my temps by over 20C?

The conclusions you have reached don't fit the data.

Krogoth wrote:

Intel did fixed the IHS fitting QC problems though. Your first-batch Haswell (4770K) was most likely a lucky unit that had proper IHS fitting at the fab. I have an Ivy Bridge (3570K) that was one of the lucky ones. I have seen countless others who weren't as fortunate. They saw difference of ~10-20C difference between cores under a full load which suggested an IHS fitting problem/air-gap. De-lidding and reseating the IHS yielded massive returns on their chips.

I don't see any evidence that Intel ever fixed it. Delid+refit even on modern non-solder TIM chips still give 15+ C reduction under load. My wife's 4770K dropped by nearly 30 C by delidding and refitting. I'm tempted to delid my 6700K since I have the tools and some extra liquid metal.

I've never really OC'd any Ivy Bridge chip enough to worry about temperatures, but I absolutely had to de-lid my 4790, sand down the IHS to reduce the gap a bit, and throw some Conductonaut in there instead of any old goop.Temps dropped around 12-15C at full load, and I was happier with 55-60C than I was with 70C+.

If that was Devil's canyon, I'd hate to live with a 4770K.

I'm about to build a 9900K rig (waiting for the MB to arrive before I begin). I took a look into the delid & liquid metal option and I don't think it is worth it for me as this is just a gaming rig.

Gamers Nexus took a look at a 9900K delid & liquid metal. Their conclusion - not worth it unless you are a competitive OC'er:

https://youtu.be/0DKP77tX9VE?t=278

The greatest words in the English language.