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

[Mr. Fox]

44 minutes ago, Papusan said:

This is almost as a defective product. Or just say what MSI is able to throw out. If this is what MSI say is a binned product and want premium on top. No thanks. But your Newegg SKU is more the normal and what to expect for binned products. Slightly better vram than what I had. But all in all... Disapontment products from Msi. And in line with my experience with MSI graphics cards. Thanks for the testing brother. This just confirm my feelings about MSI🙂

I seriously doubt that they are actually binning the core silicon. I doubt that any of them are, even if they say they are. None of them care enough about the products they sell or the people that buy them to do that, but they certainly don't mind pretending they do and lying about it. I could be wrong and I am open to being shown evidence that I am wrong, but otherwise I believe they are just lying if they say that. A binned product should exclude everything other than golden samples and let the average stuff go to the gamerkids that don't know, don't care, want cheap.

 

If they are binning them, what they are doing is identifying below average  trash samples and putting the garbage samples on their least expensive gamerpunk bottomfeeder GPUs and only using average or better on the Suprim X. That's binning, but not the right way. Even EVGA did not do it the right way. The right way would be to identify the top tier silicon samples, set them aside and reserve them exclusively for the halo product. When those chips run out, you don't make or sell any more of the halo product.

 

EVGA binned the 3090 Kingpin. They actually had felt tip marker numbers written on the silver rim around the GPU core. Mine sucks at core overclocking... super average. Not horrible, but nothing worth being excited about. But, it cost the same as the better samples that lucky winners received. The only benefit I got out of it was avoiding something below average. Because the core overclocking isn't well above average the product was a waste of my money and regret buying it for that reason only. Anyone that buys a halo product should be a lucky winner. That is what they paid extra for, and that's what they deserve. If I am honest, I have to admit that the memory overclocking is superior, and notably better than average. But, that by itself that isn't good enough. I expected both core and memory overclocking to be superior.

 

Brother @electrosofthaving two remarkably different Suprim X in his possession is evidence that they are not binning, or that their QC failed. Not sure which is true, or if maybe both are true; meaning that they are lying and have lousy QC.

[Papusan]

21 minutes ago, Mr. Fox said:

If they are binning them, what they are doing is identifying below average  trash samples and putting the garbage samples on their least expensive gamerpunk bottomfeeder GPUs and only using average or better on the Suprim X. That's binning, but not the right way.

I remember when EVGA did proper binning. Old days they binned the correct way and used 3 tiers. And you paid accordingly what you wanted to pay. Now they put the “worst best” chips together with the “better bin” and put all of the so called binned chips into ther best and most expensive halo products. I don’t call this proper binning. The range in quality is too wide. You have a big risk getting trash vs other brands that put in whatever chips in their custom SKUs. What you really get for premium is slightly better cooler and amusing design that is meant to match what they charge. And Nvidia don’t need the best coolers for their FE cards. They just use better binned chips for free. AIC partners have to make better coolers to hide what they really use and what they get from Nvidia (they don’t want pay the premium or Nvidia don’t have enough of good silicon). Same for vram. 

[Raiderman]

1 hour ago, Papusan said:

I remember when EVGA did proper binning. Old days they binned the correct way and used 3 tiers. And you paid accordingly what you wanted to pay. Now they put the “worst best” chips together with the “better bin” and put all of the so called binned chips into ther best and most expensive halo products. I don’t call this proper binning. The range in quality is too wide. You have a big risk getting trash vs other brands that put in whatever chips in their custom SKUs. What you really get for premium is slightly better cooler and amusing design that is meant to match what they charge. And Nvidia don’t need the best coolers for their FE cards. They just use better binned chips for free. AIC partners have to make better coolers to hide what they really use and what they get from Nvidia (they don’t want pay the premium or Nvidia don’t have enough of good silicon). Same for vram. 

Would be nice if EVGA would pick up AMD, and Intel, and make graphics cards again. Nvidia burnt that bridge. One of the best gpu's I ever owned was the 1080Ti FTW. Solid build quality, and OC'ed really good. I love EVGA products, but they are a niche company, that needs to expand their product line. They are not very diverse when it comes to PC tech. Last I checked they have not released any Ryzen 7000 motherboards either.

[Etern4l]

48 minutes ago, Raiderman said:

Would be nice if EVGA would pick up AMD, and Intel, and make graphics cards again. Nvidia burnt that bridge. One of the best gpu's I ever owned was the 1080Ti FTW. Solid build quality, and OC'ed really good. I love EVGA products, but they are a niche company, that needs to expand their product line. They are not very diverse when it comes to PC tech. Last I checked they have not released any Ryzen 7000 motherboards either.

 

Unfortunately it's not a great time for them to be taking any risks, as the PC market is currently suffering a downturn (as per Intel's latest gloomy results).

[electrosoft]

2 hours ago, Papusan said:

This is almost as a defective product. Or just say what MSI is able to throw out (Bad QC or just greed). If this is what MSI say is a binned product and want premium on top. No thanks. But your Newegg SKU is more the normal and what to expect for binned products. Slightly better vram than what I had. But all in all... Disapontment products from Msi. And in line with my experience with MSI graphics cards. Thanks for the testing brother. This just confirm my feelings about MSI🙂

 

After testing my Newegg one, I was like, "this is not bad at all and it's quiet." I was expecting similar or better results with the Best Buy one and it is acting like a Trio. I actually did a search on poor 4090 memory OC to see if such low numbers were possible and apparently they are for memory OC on some duds to hit + 700-900 max on memory and 150ish on GPUs.

 

 

2 hours ago, Mr. Fox said:

I seriously doubt that they are actually binning the core silicon. I doubt that any of them are, even if they say they are. None of them care enough about the products they sell or the people that buy them to do that, but they certainly don't mind pretending they do and lying about it. I could be wrong and I am open to being shown evidence that I am wrong, but otherwise I believe they are just lying if they say that. A binned product should exclude everything other than golden samples and let the average stuff go to the gamerkids that don't know, don't care, want cheap.

 

If they are binning them, what they are doing is identifying below average  trash samples and putting the garbage samples on their least expensive gamerpunk bottomfeeder GPUs and only using average or better on the Suprim X. That's binning, but not the right way. Even EVGA did not do it the right way. The right way would be to identify the top tier silicon samples, set them aside and reserve them exclusively for the halo product. When those chips run out, you don't make or sell any more of the halo product.

 

EVGA binned the 3090 Kingpin. They actually had felt tip marker numbers written on the silver rim around the GPU core. Mine sucks at core overclocking... super average. Not horrible, but nothing worth being excited about. But, it cost the same as the better samples that lucky winners received. The only benefit I got out of it was avoiding something below average. Because the core overclocking isn't well above average the product was a waste of my money and regret buying it for that reason only. Anyone that buys a halo product should be a lucky winner. That is what they paid extra for, and that's what they deserve. If I am honest, I have to admit that the memory overclocking is superior, and notably better than average. But, that by itself that isn't good enough. I expected both core and memory overclocking to be superior.

 

Brother @electrosofthaving two remarkably different Suprim X in his possession is evidence that they are not binning, or that their QC failed. Not sure which is true, or if maybe both are true; meaning that they are lying and have lousy QC.

 

It does make one wonder what silicon quality is in the HOF 4090 for $2500+?

 

Looking at some reviews, even some Founders cards had problems hitting +1200 memory OC, MSI +1000 and even the mighty Strix 4090 with a monster memory OC of +2000 (max) was stable at +1750 and GPU +180 with it's 120% slider:

 

https://www.tomshardware.com/reviews/asus-rtx-4090-rog-strix-oc-review/2

 

Then you have tech power up and their nifty little chart of sample OCs:

 

 

 

I definitely remember my KPE 3090 being average both in memory and GPU overclocking. It did hit the promised 2025 out of the box. It was right around 2040 I think. But my KPE 3090ti was a much better sample both GPU and memory OC on average. It boosted to ~2070-2085 and memory was definitely >=+1000.

[Raiderman]

Hows this for a Time Spy score 🤣

 

[Mr. Fox]

2 hours ago, electrosoft said:

 

After testing my Newegg one, I was like, "this is not bad at all and it's quiet." I was expecting similar or better results with the Best Buy one and it is acting like a Trio. I actually did a search on poor 4090 memory OC to see if such low numbers were possible and apparently they are for memory OC on some duds to hit + 700-900 max on memory and 150ish on GPUs.

 

 

 

It does make one wonder what silicon quality is in the HOF 4090 for $2500+?

 

Looking at some reviews, even some Founders cards had problems hitting +1200 memory OC, MSI +1000 and even the mighty Strix 4090 with a monster memory OC of +2000 (max) was stable at +1750 and GPU +180 with it's 120% slider:

 

https://www.tomshardware.com/reviews/asus-rtx-4090-rog-strix-oc-review/2

 

Then you have tech power up and their nifty little chart of sample OCs:

 

 

 

I definitely remember my KPE 3090 being average both in memory and GPU overclocking. It did hit the promised 2025 out of the box. It was right around 2040 I think. But my KPE 3090ti was a much better sample both GPU and memory OC on average. It boosted to ~2070-2085 and memory was definitely >=+1000.

I really can't complain about the memory on my 3090 KPE. I can run +1600 in benches and +1500 for extended gaming sessions. That is definitely better than average in that product generation.

 

I find myself wondering if the goofballs at MSI think they can get away with slightly lower silicon quality on the liquid cooled model due to lower temperatures compared to the Suprim X air cooled. They're selling them for roughly the same price with the air cooled version usually just a few bucks more. What I mean by that is after they are built the water cooling goes on the worse samples of the batch. There is somewhat of a bit of logic to that thought, but it should not be applied to a halo product. That would be something that an end user would do to extract a few more megahertz boost clock, not something that an OEM should be doing to help offset or compensate for a lower bin quality.

[ryan]

quick question does anyone know how i can find my asic quality with my laptops 3060...gpuz doesnt seem to have it with the newest version..also i broke my record. 2100 barrier.

 

equates to less than this but I was getting artifacting

 

also worth mentioning I got my highest port royal score with stable clocks

 

 

i mean asic quality via gpuz eternal

 

 

[Etern4l]

17 minutes ago, ryan said:

quick question does anyone know how i can find my asic quality with my laptops 3060...gpuz doesnt seem to have it with the newest version..also i broke my record. 2100 barrier.

 

 

 

Congrats!

 

(BTW GPUs are not classed as ASICs).

[Mr. Fox]

1 hour ago, ryan said:

quick question does anyone know how i can find my asic quality with my laptops 3060...gpuz doesnt seem to have it with the newest version..also i broke my record. 2100 barrier.

 

equates to less than this but I was getting artifacting

 

also worth mentioning I got my highest port royal score with stable clocks

 

 

i mean asic quality via gpuz eternal

 

 

Most modern GPUs do not report their ASIC quality. Open GPU-Z. Go to the Advanced tab. Change the drop-down menu to ASIC Quality. If it is not there, you are using a GPU that was designed after they started trying to keep trash bin silicon quality a secret. I don't know if it was Turing or Ampere when they started trying to keep the silicon quality hidden. Neither my 3060 Ti nor my 3090 KPE show the ASIC quality rating.

 

ASIC = Application Specific Integrated Circuit

 

Where is ASIC quality now?

 

 

[ryan]

thanks. i had no idea as i havent tested it since 6970m

[Etern4l]

1 hour ago, Mr. Fox said:

Most modern GPUs do not report their ASIC quality. Open GPU-Z. Go to the Advanced tab. Change the drop-down menu to ASIC Quality. If it is not there, you are using a GPU that was designed after they started trying to keep trash bin silicon quality a secret. I don't know if it was Turing or Ampere when they started trying to keep the silicon quality hidden. Neither my 3060 Ti nor my 3090 KPE show the ASIC quality rating.

 

ASIC = Application Specific Integrated Circuit

 

Where is ASIC quality now?

 

 

 

The difference between an ASIC and standard IC parts can be a bit subtle, so I would excuse Jay there. Basically ASICs are custom chips (e.g. a Bitcoin miner, which can do nothing else), whereas ICs are standard, general-purpose parts. Despite what he said, CPUs and GPUs are not ASICs.

 

https://en.m.wikipedia.org/wiki/Application-specific_integrated_circuit

 

https://www.pcmag.com/encyclopedia/term/asic

 

In the case of GPUs it's more accurate to refer to IC quality, chip quality, silicon quality etc. 

[tps3443]

Guys! Look at these GPU temps. It’s just stupid. 
 

This is the normal 520 watt KP HC LN2 bios. 22C! 
 

My chiller is MONSTER. Water flow is absolutely key with a chiller. I had no idea. I have heard this before, but I didn’t think it would offer such a drastic difference. I thought I maybe just needed a larger water chiller. Turns out I had low head pressure all this time. Bear in mind those GPU temps are also with a 6Ghz 13900K. 
 

🤯🤯🤯

 

 

[Papusan]

23 hours ago, ryan said:

quick question does anyone know how i can find my asic quality with my laptops 3060...gpuz doesnt seem to have it with the newest version..also i broke my record. 2100 barrier.

ASIC = Gone with Pascal (10xx). Maxwell was the last nvidia arc with this feature. 

 

Here's how EVGA priced it for 980Ti  K|ngp|n. My Zotac 980Ti Amp Extreme have 81.8% Asic quality. Costed $679.99 USD at launch or +50% cheaper than the best binned variant from EVGA https://www.techpowerup.com/review/zotac-geforce-gtx-980-ti-amp-extreme/28.html

 

 

My Zotac 980Ti Amp Extr can do above 1600 and Jayz 980Ti K|ngp|n sample (his 2nd  one) still barely do 1552MHz and thats near 1000+$ cards from EVGA. And Jayz wasn't happy with what it could do.

 

And as you can see.... Buying the Halo product doesn't mean you will win in the silicon lottery... Not at all. So a great idea remove such features as Nvidia think. Not all will be happy seing they paid premium for nothing.

https://forums.evga.com/What-are-your-GTX-980-Ti39s-ASIC-qualities-m2347656.aspx

[tps3443]

3 minutes ago, Papusan said:

ASIC = Gone with Pascal (10xx). Maxwell was the last nvidia arc with this feature. 

 

Here's how EVGA priced it for 980Ti K|ngp|n. My Zotac 980Ti Amp Ext have 81.8% Asic quality.

 

I had a launch day AMD RX480 and it was 98.5% ASIC. It overclocked very well for air cooling. It was rivaling a stock GTX980 for sure. I loved that GPU. 

[Mr. Fox]

1 hour ago, Etern4l said:

 

The difference between an ASIC and standard IC parts can be a bit subtle, so I would excuse Jay there. Basically ASICs are custom chips (e.g. a Bitcoin miner, which can do nothing else), whereas ICs are standard, general-purpose parts. Despite what he said, CPUs and GPUs are not ASICs.

 

https://en.m.wikipedia.org/wiki/Application-specific_integrated_circuit

 

https://www.pcmag.com/encyclopedia/term/asic

 

In the case of GPUs it's more accurate to refer to IC quality, chip quality, silicon quality etc. 

He is not the one making things up. I do not know whose idea it was to start referring to it as ASIC quality. It is/was a real thing that some people cared about. Not remarkably different conceptually than an Asus ROG SP rating 

[Papusan]

2 hours ago, Etern4l said:

 

The difference between an ASIC and standard IC parts can be a bit subtle, so I would excuse Jay there. Basically ASICs are custom chips (e.g. a Bitcoin miner, which can do nothing else), whereas ICs are standard, general-purpose parts. Despite what he said, CPUs and GPUs are not ASICs.

 

https://en.m.wikipedia.org/wiki/Application-specific_integrated_circuit

 

https://www.pcmag.com/encyclopedia/term/asic

 

In the case of GPUs it's more accurate to refer to IC quality, chip quality, silicon quality etc. 

All is from EVGA's own page for their top tier graphics card. See the link I posted above. I put it here....

 

 

I have two older GTX 480. One with Asic at below 60% and one above 86%. The lower binned card is awful for oc'ing and run damn hot (FERMY was known for running HOT so High ASIC score was very needed). And if you didn't clean up you almost saw smoke coming from that card, LOOL

32 minutes ago, Mr. Fox said:

He is not the one making things up. I do not know whose idea it was to start referring to it as ASIC quality. It is/was a real thing that some people cared about. Not remarkably different conceptually than an Asus ROG SP rating 

Exactly. And I'm sure Nvidia still have tools and could offer the SDK to the GPU-Z author.

[ryan]

yeah I used to overclock years ago, now I just benchmark casually with no clue as to what im doing. I don't get pascal up cards. too reliant on temps. I remember my 260m overvolting it and overclocking the crap out of it. I broke a few records with each gpu I bought but yeah those days are over as it seems more like a lottery. less skill more luck. and now the big question is? will nvidia and amd as well as intel snuff out this skill and make everything more like dumb blocks(lego-computers)>?

[Mr. Fox]

31 minutes ago, ryan said:

yeah I used to overclock years ago, now I just benchmark casually with no clue as to what im doing. I don't get pascal up cards. too reliant on temps. I remember my 260m overvolting it and overclocking the crap out of it. I broke a few records with each gpu I bought but yeah those days are over as it seems more like a lottery. less skill more luck. and now the big question is? will nvidia and amd as well as intel snuff out this skill and make everything more like dumb blocks(lego-computers)>?

Pretty much. Sums it up well. Overclocking Intel CPUs is the only thing left that bears a healthy resemblance to normalcy, as long as you don't do the stupid TVB overclocking for idiot sheeple wusses thing. That is not normal and if it ever becomes normal I will want no part of it.

 

[ryan]

I honestly think LEGO computers will become a thing. they got rid of MXM and made overclocking basically automatic and therefore safe. its just greed fueling these trends. less risk to them= more money. people would wreck cards and return them and nvidia/amd/intel would lose money. its kinda like a food bank having to supply meat, expensive and when some of it goes bad they have to throw it out and people lose. so they do the obvious, stop giving out meat. but is that the ethical and right descision? no.....freeking people need meat. people need customization. people need modularity. this is just fricken dumb. STOP

[Mr. Fox]

9 minutes ago, ryan said:

I honestly think LEGO computers will become a thing. they got rid of MXM and made overclocking basically automatic and therefore safe. its just greed fueling these trends. less risk to them= more money. people would wreck cards and return them and nvidia/amd/intel would lose money. its kinda like a food bank having to supply meat, expensive and when some of it goes bad they have to throw it out and people lose. so they do the obvious, stop giving out meat.

That is not real overclocking. It is allowing the CPU to behave normally based on a temperature regulated algorithm predetermined by the manufacturer. It is overclocking for limp-wristed pansies that whine about their turdbooks being too big, thick and heavy. These mama's boys are the target market though. They aren't smart enough to figure anything out on their own. It is easy to sell trash to them. All you have to do is get a shill on YouTube to say it is awesome.

[ryan]

I hear the undertones. I get that. But when the effect is random im afraid alot of people dumb and smart fall into that catergory. and all the new school overclockers are doing/saying is PI is exactly 3. wrong/right. no proof of concept.

 

for example if it was skill, and not random. what guide would you introduce for people using pascal up chips too get world record scores, because if it was skill it would be replicable.

 

as it stands

 

overclock 50

overclock 50 more

and yet again 50 more

artifacts back down 10 untill it goes away.

 

then lower temps

 

then cross fingers.

 

Ive literally tweaked the crap outta my laptop and when it was bloated scored 9200 gpu in timespy now my max is around 8900, same drivers ect. it just isn't replicable. the tiniest change can have a profound impact.

 

[electrosoft]

27 minutes ago, ryan said:

I hear the undertones. I get that. But when the effect is random im afraid alot of people dumb and smart fall into that catergory. and all the new school overclockers are doing/saying is PI is exactly 3. wrong/right. no proof of concept.

 

for example if it was skill, and not random. what guide would you introduce for people using pascal up chips too get world record scores, because if it was skill it would be replicable.

 

as it stands

 

overclock 50

overclock 50 more

and yet again 50 more

artifacts back down 10 untill it goes away.

 

then lower temps

 

then cross fingers.

 

Ive literally tweaked the crap outta my laptop and when it was bloated scored 9200 gpu in timespy now my max is around 8900, same drivers ect. it just isn't replicable. the tiniest change can have a profound impact.

 

 

Exactly....in a dynamic environment such as an actively running OS, the tiniest changes CAN have a small or even profound impact.

 

Moving the sliders is the culmination of other steps taken from OS tweaking to going custom water, outlier cooling solutions (dragging your PC outside, using a portable AC unit, dunking your rads, pelt, LN, etc...) to binning chips, motherboards, memory, fine tuning thermal solutions, lapping CPU and GPU and more...

 

There's so much more to it.

 

Take your laptop which has many more constraints than a desktop.

 

Create another OS partition that is tweaked for benching

You can get under the hood and analyze your heatsink to tweak and enhance it

Install or mod in custom fans

use a portable AC unit for better cooler air intake

add water cooling even

Test various thermal compounds and solutions with end game being a perfect LM application

lap the CPU and GPU if applicable

 

It depends on how serious you are and how far you want to take it.

 

[Rage Set]

33 minutes ago, ryan said:

I hear the undertones. I get that. But when the effect is random im afraid alot of people dumb and smart fall into that catergory. and all the new school overclockers are doing/saying is PI is exactly 3. wrong/right. no proof of concept.

 

for example if it was skill, and not random. what guide would you introduce for people using pascal up chips too get world record scores, because if it was skill it would be replicable.

 

as it stands

 

overclock 50

overclock 50 more

and yet again 50 more

artifacts back down 10 untill it goes away.

 

then lower temps

 

then cross fingers.

 

Ive literally tweaked the crap outta my laptop and when it was bloated scored 9200 gpu in timespy now my max is around 8900, same drivers ect. it just isn't replicable. the tiniest change can have a profound impact.

 

 

Luck only comes in when buying a component. The skill is extracting the most performance out of that component. Variables will always be a determining factor when overclocking a component, but the skill/experience is knowing how to identify the variable(s) that can cause an outcome. 

 

I know TimeSpy/Firestrike so well, I know by heart what a 1080 TI to a 3090 should score based off their overclocks. Here is an example I did on an MSI 2080 TI stream. I have it at the exact spot where I guessed correctly where the score would land based off my skill/experience. 

 

 

[Papusan]

1 hour ago, ryan said:

I honestly think LEGO computers will become are already a thing.

 

 

When you're finished. Please post bench results😁