*Official Benchmark Thread* - Post it here or it didn't happen :D

[Papusan]

34 minutes ago, Rage Set said:

 

The day my wife buys me a GPU is the day I cancel the order. I love her with all of my heart and being, but she doesn't know tech. I created an actual video game for her for Christmas about ten years ago and she (like me) grew up playing video games, however, since it required a mouse/keyboard she couldn't wrap her head around the controls. I built her a 10900K/3080 rig. It just sits there, unused with a keyboard that still sprews RGB vomit. With all that said, she is the only person, personally around me that has zero problems with the fans blasting on my MORA radiators.   

My wife would never ever buy computer parts without asking me what to buy 🙂

24 minutes ago, Mr. Fox said:

No, bug it is probably a bugged run. I ran one right behind it that was just a few points less than your last submission. I forgot to take a screenshot with NVCP open so I am trying to get another buggy run with the screenshot showing it enabled.

 

I think what might be happening is ECC enabled sometimes glitches and disables itself.

Run with GPU-Z in the background. Follow the amount vram. Below 24GB means it’s disabled Enabled. 

[johnksss]

1 minute ago, Papusan said:

 

Run with GPU-Z in the background. Follow the amount vram. Below 24GB means it’s disabled. 

Below means it's enabled right? 23.5GB. Above means ECC is off 24.5GB

[Papusan]

5 minutes ago, johnksss said:

Below means it's enabled right? 23.5GB. Above means ECC is off 24.5GB

Yep, ECC Enabled eats some of the vram. 
 

A typo above. Write on a phone. 

[Mr. Fox]

  

7 minutes ago, johnksss said:

Below means it's enabled right? 23.5GB. Above means ECC is off 24.5GB

2 minutes ago, Papusan said:

Yep, ECC Enabled eats some of the vram. 
 

A typo above. Write on a phone. 

Evidently a driver glitch. Maybe a TDR error tripped it off or something. See what I posted below.

This one is valid.

https://www.3dmark.com/3dm/89116421 | https://hwbot.org/submission/5201316_

[Papusan]

10 minutes ago, Mr. Fox said:

  

Evidently a driver glitch. Maybe a TDR error tripped it off or something. See what I posted below.

This one is valid.

https://www.3dmark.com/3dm/89116421 | https://hwbot.org/submission/5201316_

If I remember correctly I saw glitches like this when I disabled all remaining Nvidia services in task manager. ECC went from Enabled to disabled. Fun times with modern, LOOL

 

if people isn’t aware the bot will be flooded with “correct” scores but not valid due the bot has more restricted rules than the benchmark software companies own rules. 
 

nvidia has created a buggy monster. Fun times forwards if they can’t make proper drivers/software. Even funnier if they remove the ECC adjustment in NCP

[MyPC8MyBrain]

i'm trying to follow this ECC debacle, some basics i am having hard time wrapping my head around,
the first is why ECC is on a gpu in the first place? if for whatever justified reason ECC is needed for some gpu scenario why are you guys insisting on keeping it on if the feature is disqualified by bench tools moving forward? in other words why slow down calculation in a bench is that important? what am i missing?

 

[johnksss]

2 minutes ago, MyPC8MyBrain said:

i'm trying to follow this ECC debacle, some basics i am having hard time wrapping my head around,
the first is why ECC is on a gpu in the first place? if for whatever justified reason ECC is needed for gpu why are you guys insisting on keeping it on if the feature is disqualified by bench tools moving forward? what am i missing?

Keeping it on is valid, but turning it off for 3dmark benchmarks is not valid.

[Papusan]

3 minutes ago, MyPC8MyBrain said:

i'm trying to follow this ECC debacle, some basics i am having hard time wrapping my head around,
the first is why ECC is on a gpu in the first place? if for whatever justified reason ECC is needed for some gpu scenario why are you guys insisting on keeping it on if the feature is disqualified by bench tools moving forward? in other words why slow down calculation in a bench? what am i missing?

Because 4090 is also for creators. 

[electrosoft]

On 2/12/2023 at 3:37 PM, Papusan said:

For the future I would say go with the 4090 if you are gamer and you keep your HW until EOL date for drivers (up to 8 years). 16GB vram or below is a  no go 6 years down the road. Just look at the 3080@10GB 4K flagship. Can't be used as intended forwards.

 

 

And I hope they burn in with loads of overpriced MB. And I hope this will affect Nvidia and AMD's gamer cards.

 

Motherboard Shipments Plummet by Ten Million Units in 2022: Report

The combined hit for Asus, Gigabyte, MSI, and ASRock reportedly around 10 million.
https://www.tomshardware.com/news/motherboard-shipments-drop-ten-million-units-in-2022-report

 

3080 w/ 10GB was always going to be an issue for a card that can do 4k. I think i even mentioned it too on the NBR forums when I was quoting VRAM allocations with my original 3090 Founders.

 

As for motherboard sales slumping.....I turn the mike over to @Mr. Fox

 

On 2/12/2023 at 3:47 PM, Mr. Fox said:

We can only hope...

 

The article misses the mark on what is to blame...

The problem, at least in great part, is due to their idiotic pricing. I do not believe it is due to lack of demand or economic downturn. People that don't have money and shouldn't spend what they do not have are being more stupid than ever about using credit cards. (That is going to bite them--and maybe the rest of us--in the butt before in the long run, but it is what it is.) I think the real issue is there are enough people that are not stupid, so they are not giving into stupid prices.

 

Even the mediocre gamer mobos are priced like the high-end enthusiast/overclocker boards now. The high-end options are beyond ridiculous pricing, but they are still selling because people like us place importance on having what we want over being intelligent about how we spend money.  The mid-range Strix mobos are a good example. I bought mine on eBay as refurbished/open-box option for what should have been new retail pricing, but new retail price is/was more than what I paid for a Dark mobo on sale direct from EVGA. It's OK, but the Strix is definitely not an enthusiast board. It is a gamer mobo prices like an overclocker mobo.

 

Well said.

 

7 hours ago, Papusan said:

Me..... Still not opened the package with my new screen (see pict) due me still waiting on Asus 4090 Strix OC..... Me, will never be the 4090 Strix OC (impossible to get, due ordering competition is so hard).... Me checked out other places every day over weeks..... Me, one day I found an other option(see above)..... Me ordered KFA2 4090 HOF (The co-branding of GALAX in EU)..... Me, now have to order the second BE QUIET! 12VHPWR PCI-E ADAPTER CABLE (dual chance for Nvidia connector melting and fire, LOOL). 

 

In short.... Me looking on my own self-awareness... Me seeing I'm Klin Kokos, LOOL

 

We have words for this here home bro Fox... "Klin kokos" https://no.wikipedia.org/wiki/Klin_kokos

 

Translated into English... KOKO

 

Oh' well, Asus lost their chance to get my hard earned money (They reduce inventory to keep up high disgusting prices)and I was forced to spend more than needed ($200). What a game. But in the end, with the Norwegian 25% Tax on top, the HOF card won't be more expensive here home. I don't pay more than what you have to pay in other countries. This is the only good I can say about this purchase.

 

The weird part.... 4090 HOF is almost impossible to get in the whole EU. Being able to get your hands on one here in little Norway is quite interesting. Will believe it when I see the shipping label. This means I will wait ordering the second power connector until I see the shipping label is delivered in my email. 

 

Electrosoft, no pict of my all black crocs now, but new pict will come a later day

My wife ordered this one. I knew nothing before I woke up and checked the mail 🙂

 

 

 

Awesome @Papusan and congrats on the purchase! I think you'll be the first of this group with a real HOF class card from the last 2-3 generations so we can see one in actual use. Looking forward to pics, performance and opinions. Again, congrats bro!

[johnksss]

8 minutes ago, Papusan said:

If I remember correctly I saw glitches like this when I disabled all remaining Nvidia services in task manager. ECC went from Enabled to disabled. Fun times with modern, LOOL

 

if people isn’t aware the bot will be flooded with “correct” scores but not valid due the bot has more restricted rules than the benchmark software companies own rules. 
 

nvidia has created a buggy monster. Fun times forwards if they can’t make proper drivers/software. 

Not sure that will fly past when it hits 3dmark website....  speculations of course.

[Etern4l]

ECC is only present on the x090 models, to support some ultra-high reliability applications. I ran it for a long time and never saw any errors so turned off as it costs some usable VRAM capacity and likely performance. 

 

I'm thinking maybe HWBOT(3dMark?) have this policy to prevent people from gaining an edge with super-overclocked and artifacting VRAM. Just failing at playing devil's advocate.

 

[MyPC8MyBrain]

18 minutes ago, johnksss said:

Keeping it on is valid, but turning it off for 3dmark benchmarks is not valid.

i was under the impression that they want to keep it off or something?

 

18 minutes ago, Papusan said:

Because 4090 is also for creators. 

so was last gen 3090, how is the 4090 different in its workflow for creators that it needs ECC?

  

also what kind of ECC is that, traditionally there's no ON/OF switch for ECC unless its some software implementation or there's actually two separate memory controllers onboard,
if anything i can maybe see this useful for miners but that's pretty much it.
 

12 minutes ago, Etern4l said:

I ran it for a long time and never saw any errors so turned off as it costs some usable VRAM capacity. 

this has to be a software based ECC mutant which in my view still makes zero sense and sort of defeats the purpose of a performance bench,

 

[johnksss]

2 minutes ago, MyPC8MyBrain said:

i was under the impression that they want to keep it off or something?

 

so was last gen 3090, how is the 4090 different in its workflow for creators that it needs ECC?

 

also what kind of ECC is that, traditionally there's no ON/OF switch for ECC unless its some software implementation or there's actually two separate memory controllers onboard,

We want to keep it off as it gives higher scores, but with it off...People run benchmarks with crazy high memory over clocks which cause the benchmark to glitch out and start running faster than normal as it's no longer rendering all the content. ECC does error correcting so it means you can not really run crazy mem over clocks stable.

[Mr. Fox]

I never knew they had a thread for this because I don't spend much time crawling through threads at HWBOT Forum. I cross-posted in the thread designed for it.

https://community.hwbot.org/topic/203413-if-you-want-your-score-to-be-deleted-ask-here/?do=findComment&comment=636481

 

I'm really liking this GPU. I still hate how overpriced it is, but dang... nice overclocking device.

 

I found there is like a dead spot of instability between about 3020MHz and 3060. If I set an overclock in that range it bugs out, but if I push past it to like 3105 or 3120MHz it doesn't crash the benchmark. 

[MyPC8MyBrain]

12 minutes ago, johnksss said:

We want to keep it off as it gives higher scores,

of curse it will ECC is additional processing layer which defeats the whole premises of running a bench,
for other application it is critical when its reliable stability they are after for implementing ECC on a server not for benching bleeding edge performance,

 

lets call a spade a spade the developers made a booboo and the only way to ensure result is to enable ECC so they can get a readout they can reliably analyze, in simple word their bench software cant keep up, is that a fair assumption? 

[Etern4l]

18 minutes ago, MyPC8MyBrain said:

this has to be a software based ECC mutant

 

 

Looks that way.

[Mr. Fox]

46 minutes ago, MyPC8MyBrain said:

of curse it will ECC is additional processing layer which defeats the whole premises of running a bench,
for other application it is critical when its reliable stability they are after for implementing ECC on a server not for benching bleeding edge performance,

 

lets call a spade a spade the developers made a booboo and the only way to ensure result is to enable ECC so they can get a readout they can reliably analyze, in simple word their bench software cant keep up, is that a fair assumption? 

Sounds about right to me. It's just like the stupid Benchmate thing. Add a bunch of extra crap that creates system overhead and slows things down just to try and stop cheaters from cheating. It's ridiculous. Expecting people to self-moderate their benchmark scores is also kind of unrealistic. Where do you start splitting hairs on what is acceptable versus unacceptable? I mean, you can have a bugged out score that is inflated a little bit and hard for anyone to tell if it is accurate or not, and then you have those where it is totally obvious that it is not accurate.

---

49 minutes ago, Mr. Fox said:

I'm really liking this GPU. I still hate how overpriced it is, but dang... nice overclocking device.

 But, I do remember one of the reasons I have never like AIO coolers. Those stinking hoses really suck compared to a custom loop. Very limiting on where you can put the radiator. My case is too huge for it to reach the front panel, so I temporarily removed one of the 200MM fans and mounted this to the roof of my case to help exhaust the hot air. I hope to have a block on it in the next 24-48 hours.

[MyPC8MyBrain]

11 minutes ago, Etern4l said:

Looks that way.

if that's the case its a bogus ECC!

they cant be doing software based memory error correction that's laughable calling it ECC, its like having a 4 year old babysit a 3 year old, maybe its an acronym for something else NVidia made up they just call it ECC to be boujie and confusing? I'm sorry I'm still having a hard time wrapping my head around the rational behind this nonsense from NVidia to be honest,

[johnksss]

23 minutes ago, MyPC8MyBrain said:

of curse it will ECC is additional processing layer which defeats the whole premises of running a bench,
for other application it is critical when its reliable stability they are after for implementing ECC on a server not for benching bleeding edge performance,

 

lets call a spade a spade the developers made a booboo and the only way to ensure result is to enable ECC so they can get a readout they can reliably analyze, in simple word their bench software cant keep up, is that a fair assumption? 

If it was implemented and no one even knew it was there, no one would be the wiser on scores being able to go higher. It's only because people know that makes it a real problem. The issue is people do benchmarks a certain way and along comes a bunch of cheaters messing with the system. So changes have to be made. And this is their answer thus far. I already complained about my benchmarks being removed and that was that. Time to move on and adapt is what I do.

 

 

5 minutes ago, MyPC8MyBrain said:

if that's the case its a bogus ECC!

they cant be doing software based memory error correction that's laughable calling it ECC, its like having a 4 year old babysit a 3 year old, maybe its an acronym for something else NVidia made up they just call it ECC to be boujie and confusing? I'm sorry I'm still having a hard time wrapping my head around the rational behind this nonsense from NVidia to be honest,

Most if not all of the 30 series had ECC as well as 40 Series.... The difference is they exposed it to the OS. From my understanding. 

[Papusan]

47 minutes ago, johnksss said:

Not sure that will fly past when it hits 3dmark website....  speculations of course.

One thing for sure... A lots of not valid benchmarks will play havoc on the HW points. It will go up and down as a Yo-yo as the scores jump in. You'll never know if your poins/ranking is corr3ct or not. This one is a couple of days old...  https://hwbot.org/submission/5198709_hardwarejiri_3dmark___fire_strike_ultra_geforce_rtx_4090_33382_marks

 

 

[johnksss]

2 minutes ago, Papusan said:

One thing for sure... A lots of not valid benchmarks will play havoc on the HW points. It will go up and down as a Jo-Jo as the scores jump in. You'll never know if your poins/ranking is corr3ct or not. This one is a couple of days old...  https://hwbot.org/submission/5198709_hardwarejiri_3dmark___fire_strike_ultra_geforce_rtx_4090_33382_marks

 

 

That score will be getting kicked to the curb.... but how fast after the 15th is the million dollar question.

 

I'm ready for the score changes because I don't have a bunch of bugged runs, but that 3090 NVlink I was really mad about. That cost real money!

 

Along with the mobile gpus scores they lost of mine.

[Mr. Fox]

It's going to be interesting to find out how high this sucker will clock on a chilled water block. Getting "hot" it's still averaging 3126MHz on core, LOL. I am going to have to try more than +1600 on memory, too. Maybe it will do +1700 or more.

 

https://www.3dmark.com/3dm/89118436

 

[Papusan]

11 minutes ago, johnksss said:

That score will be getting kicked to the curb.... but how fast after the 15th is the million dollar question.

Yep, I know. But it will come a flood of them and they will stay there until one some report them or the mod team catch it. The ranking and points will never be correct as it is now with the ECC mess. And buggy drivers/software coming forwards from Nvidia won't make it easier.

 

And for the records... A valid run (ECC Enabled) doesnt atomaticly mean it is valid and will stay on the bot.

[Mr. Fox]

Just now, Papusan said:

Yep, I know. But it will come a flood of them and they will stay there until one report them or the mode team catch it. The ranking and points will never be correct as it is now with the ECC mess.

I bet they go in the order of ranking and the "golden boys" will be the last to gets theirs deleted, or they will conveniently get overlooked in the purging process.

[Papusan]

10 minutes ago, Mr. Fox said:

I bet they go in the order of ranking and the "golden boys" will be the last to gets theirs deleted, or they will conveniently get overlooked in the purging process.

My best bet... They will only be concerned about the upper end of the leaderboard. If some post bugged scores in the mid/lower stack it will be there until some report it. And I talk about what will happen in the future after 15. Feb.

 

Then we have benches from the not worthy one with valid higher upper scores that will be thrown out from the bot by the chosen one.