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

[Etern4l]

26 minutes ago, tps3443 said:

Youtube will fix the quality momentarily, it’s always bad right after it’s uploaded. I have not touched the E-Cores yet. I just set 45 E-Cores, and 51 cache and left it. I have not tried to go past 45. 

 

Planning to try that, I think I may have thermal headroom. For now just need to hold my horses and check if 55x is stable. Very happy with the repaste so far - clearly better than SYY. I couldn't get 55x stable before (with UV). Once everything tests out over the next few days, I will post separately to warn people against this pseudo-paste.

 

BTW With 61x P-Cores, how badly do you beat these Indigo Bench 4.4.15 and Blender 3.3.0 scores?

 

IndigoBench:

 

 

 

[tps3443]

11 minutes ago, Etern4l said:

 

Planning to try that, I think I may have thermal headroom. For now just need to hold my horses and check if 55x is stable. Very happy with the repaste so far - clearly better than SYY. I couldn't get 55x stable before (with UV). Once everything tests out over the next few days, I will post separately to warn people against this pseudo-paste.

 

BTW With 61x P-Cores, how badly do you beat these Indigo Bench 4.4.15 and Blender 3.3.0 scores?

 

IndigoBench:

 

 

 

I’m not running 6.1 all cores. I’m only running 6.1 boost on (2) cores. As for the E-Cores, I have not even touched them yet. I’ve been working, and I’ve only had a little time here and there to tinker with the cpu. 

[Clamibot]

3 hours ago, electrosoft said:

 

 

 

In many of these cases you have them running Y-Cruncher or similar and just red lining their CPUs for hours in not only pull but equally (or more) important temps. If you want to sit there and have your CPU plow through Y-Cruncher at 88c+ be my guest that's not my cup of tea.

 

For every OC'er that posts a killer OC with their own testing criteria, you will always...ALWAYS have someone yip in about, "Well what about OCCT? Y-Cruncher? P95? Run that for a Bazillion Hours and let's see!" Like some of these OC's are D2D?

When I see those posts I just grimace and move to the next.

 

I may do OC runs for shiggles that momentarily (less than a few minutes) hit high temps but my D2D is always about finding the sweet spot of a good clock at the lowest pull (even stock). This especially extends to my laptop use. The two modes of operation can co-exist and are not mutually exclusive.

 

@tps3443 the nature of your setup helps prevent those sustained, high temp runs that in unison with pull *may* cause degradation.

 

 

 

 

 

Pretty much same here. I prefer finding the max stable overclock for regular use cases as it's actually useful. I respect that heavy benching for hours on end is a hobby for some people and that brings them joy, but I'm interested in increasing productivity by shaving every last second I can off of video renders, code compilation, game builds, etc, and I want to maximize my framerate in games.

 

I do the occasional cinebench test to see how well my settings hold up. If they hold up to at least 3 runs in a row without crashing the system, I say those settings are stable enough for regular usage. I will admit I also do it to show off sometimes.

[Etern4l]

8 minutes ago, tps3443 said:

I’m not running 6.1 all cores. I’m only running 6.1 boost on (2) cores. As for the E-Cores, I have not even touched them yet. I’ve been working, and I’ve only had a little time here and there to tinker with the cpu. 

 

On your video you seem to be showing 6.1GHz on all but one core (that's 58x) while running CB. Something's broken with the sound ATM, so can't hear what you're saying. Anyway, even 5.8GHz on all cores would be great in my little book. Curious how well that would do in Blender at least. Takes just a moment to install and not that long to run, same for IndigoBench actually.

[tps3443]

1 hour ago, Etern4l said:

 

On your video you seem to be showing 6.1GHz on all but one core (that's 58x) while running CB. Something's broken with the sound ATM, so can't hear what you're saying. Anyway, even 5.8GHz on all cores would be great in my little book. Curious how well that would do in Blender at least. Takes just a moment to install and not that long to run, same for IndigoBench actually.

It’s just 6.1Ghz on a (2) core load. It’s just flipping between the cores that are loaded. 5.8Ghz all core load. But yes I can run the test. I’ll download it and report back.

[Raiderman]

Hmmm, I must be missing something, I cant seem to find the like button 😄 Glad to see everyone again. Thanks for the nice replies!!

[Etern4l]

9 minutes ago, Raiderman said:

Hmmm, I must be missing something, I cant seem to find the like button 😄 Glad to see everyone again. Thanks for the nice replies!!

At the bottom there is a Theme dropdown, click and select Dark.

[Raiderman]

Im Running my own custom dark theme, the like button is white, and does not show up on the default theme. I switched it to the dark theme, and voila!

[Reciever]

1 minute ago, Raiderman said:

Im Running my own custom dark theme, the like button is white, and does not show up on the default theme. I switched it to the dark theme, and voila!

We have a dark theme that honestly probably will make the default since Invision community forum never approved my question for viewership. 

 

Be sure to check out some of the howto's, I'm still figuring it out as I go on my spare time. 

[tps3443]

Testing 6.2Ghz Boost with 5.9Ghz all core

UPDATE, this is my new daily profile.

I see a lot of people running this test, so here goes. Geekbench 5. Yes, its SUPER STABLE! I love it. This is a Pre- 13900KS all the way. CRAZY FAST single threaded performance, that's for sure. 



 

[Papusan]

17 hours ago, Mr. Fox said:

The elite chosen are an entitled group. In fact, I honestly don't believe CENS wanted the CPU from @tps3443for himself as much as wanting to remove it from "retail" circulation because extraordinary silicon samples diminish the fake sense of mystery and talent the chosen need to validate their position. Even the comment by Splave exposes the subtle arrogance, although it probably was not intended to send that message. He may have well said, "Nice to see a pion accidentally get lucky with retail garbage that wasn't reserved for them." Guys like you and me and @johnksss could be a very dangerous threat to the elite overclockers' status if we got our hands on silicon samples like those @tps3443has lucked out with, and they know it. They're not worried about silly gamer kids that have no idea what they have, nor should they be. They're going to be too busy bragging about their low voltage and low power draw rather than chasing world records. Even if we didn't go down the LN2 path and take their crowns away, the gap between their spots on the leaderboard and the "not chosen" riff raff would be too close for their comfort. 

Haven't the silicon winner as the elite usually get their hands on but... Me, my 13900K and H2O, baked in between LN2 cooled processors. It's also about efficiency. Higher boost clocks doesn't always help if the rest of the package don't keep up. Hwbot Global ranking wPrime - 32m 24xCPU's.......

 

 

Bro @Mr. Foxcould you do me a favor? 

 

1th. Run Cinebench R20 or R23 with one and two pumps. Or even a 3rd test if you have 3 pumps. Would be nice to see a possible temp difference with added pumps.

 

Stock clocks or your daily driver is just fine for all the different tests.

 

2nd. Check the water flow vs different water temperature. Run a bench with the chiller then a bench with your normal custom water cooling. The water flow should decrease when the water become quite cold. Hence testing with and without the chiller. Test with only one pump. Thanks 🙂

 

[Etern4l]

8 hours ago, tps3443 said:

Testing 6.2Ghz Boost with 5.9Ghz all core

UPDATE, this is my new daily profile.

I see a lot of people running this test, so here goes. Geekbench 5. Yes, its SUPER STABLE! I love it. This is a Pre- 13900KS all the way. CRAZY FAST single threaded performance, that's for sure. 



 

 

GeekBench 5 has its uses, but the main one is that its cross platform and tests GPUs a bit. The whole test barely fits within a boost period of macbook, before it inevitably starts throttling. I've noticed a lot of variance in the results due to the short duration. 

[Mr. Fox]

Stock 3090 KPE: https://opendata.blender.org/benchmarks/d8bb4b44-e566-4908-9be6-633bb6335451/

Stock 3060 Ti FTW3: https://opendata.blender.org/benchmarks/7b290602-6130-4b33-b952-a7fad73f32e3/

[Papusan]

Nice with new added professional features on gamer cards. They can be used for everything. Yeah and activating it slows down the VRAM. No need to let the vram breath. Added to the Hwbot rules today. Maybe hwbot should delete all "bugged scores" so everyone can compete under equal conditions? Or is it so that older results can have its own place at the top of the ranking board? Will/should Hwbot throw away all scores on the leaderboards from 3090 Ti and upwards because people don't want such fancy features enabled? That would be would push an outcry 🙂 Everything was a lot easier in the older days. New is always better. In all ways (sarcasm).

 

[Etern4l]

1 minute ago, Papusan said:

Nice with new added professional features on gamer cards. They can be used for everything. Yeah and activating it slows down the VRAM. No need to let the vram breath. Added to the Hwbot rules today. Maybe hwbot should delete all "bugged scores" so everyone can compete under equal conditions? Or is it so that older results can have its own place at the top of the ranking board? Will/should Hwbot throw away all scores on the leaderboards from 3090 Ti and upwards because people don't want such fancy features enabled? That would be would push an outcry 🙂 Everything was a lot easier in the older days. New is always better. In all ways.

 

 

Sorry, could you clarify which GeForce cards support ECC, and how to enable this? 

[Papusan]

57 minutes ago, Etern4l said:

 

Sorry, could you clarify which GeForce cards support ECC, and how to enable this? 

Haven’t the new cards but mean it’s 3090 Ti and above. Most likely a switch in NCP. 
 

No gamers should bother with this. And now you get a performance penalty if you bench for glitter and glory. 

[electrosoft]

1 hour ago, Etern4l said:

 

Sorry, could you clarify which GeForce cards support ECC, and how to enable this? 

 

21 minutes ago, Papusan said:

Haven’t the new cards but mean it’s 3090 Ti and above. Most likely a switch in NCP. 
 

No gamers should bother with this. And now you get a performance penalty if you bench for glitter and glory. 

 

3090ti, 4080 and 4090 use ECC. Rest of Ampere use EDR.

 

 

[Mr. Fox]

27 minutes ago, Papusan said:

Haven’t the new cards but mean it’s 3090 Ti and above. Most likely a switch in NCP. 
 

No gamers should bother with this. 

[ryan]

Quick question?

 

I'm buying 64gb of ram for my hp omen 16-b0010ca.  I'm looking at gskill 3200 ddr4 ram or corsair vengence ddr4 3200...they both support xmp but only the vengence says it will Automatically overclock.

 

What's the best 32gb and up ram I can get for my laptop. is 3200 the max speed and will my omen 16 overclock the 32gb set to say 3600 Automatically I'm about to buy the 64gb gskill it's in my cart but the overclocking of the 32gb vengence has me interested...I don't mind stepping down to 32gb over 64gb if I get faster ram..

 

Advice?

[Etern4l]

50 minutes ago, Papusan said:

Haven’t the new cards but mean it’s 3090 Ti and above. Most likely a switch in NCP. 
 

No gamers should bother with this. And now you get a performance penalty if you bench for glitter and glory. 

 

Wouldn't ECC allow you to overclock higher without the risk of artifacting, as some minor ones might be autocorrected? 

[Etern4l]

33 minutes ago, ryan said:

Quick question?

 

I'm buying 64gb of ram for my hp omen 16-b0010ca.  I'm looking at gskill 3200 ddr4 ram or corsair vengence ddr4 3200...they both support xmp but only the vengence says it will Automatically overclock.

 

What's the best 32gb and up ram I can get for my laptop. is 3200 the max speed and will my omen 16 overclock the 32gb set to say 3600 Automatically I'm about to buy the 64gb gskill it's in my cart but the overclocking of the 32gb vengence has me interested...I don't mind stepping down to 32gb over 64gb if I get faster ram..

 

Advice?

 

Not really a desktop question, but it's a matter of compatibility with the particular laptop. Gskill can be more of a hit and miss. Usually can't go wrong with HyperX, but timings are more conservative. Ensure there is a good returns policy. 

 

Edit: apologies for the double-post. Spilt milk. 

[Papusan]

54 minutes ago, Mr. Fox said:

Isn't it nice. "If" UL allow you to run with ECC unchecked and Hwbot force you to use it... What score do you want to pull up from the hat if others on the web want you to show your score vs theirs? 🙂

31 minutes ago, Etern4l said:

 

Wouldn't ECC allow you to overclock higher without the risk of artifacting, as some minor ones might be autocorrected? 

I expect the new tool (system info) + that you are forced to enable ECC will always affect max scores. There is a performance panalty up to 5 or 6%if you enable the new feature in NCP regarless what the ECC autocorrect.

[Etern4l]

4 minutes ago, Papusan said:

Isn't it nice. "If" UL allow you to run with ECC unchecked and Hwbot force you to use it... What score do you want to pull up from the hat if others on the web want you to show your score vs theirs? 🙂

I expect the new tool (system info) + that you are forced to enable ECC will always affect max scores. There is a performance panalty up to 5 or 6%if you enable the new feature in NCP regarless what the ECC autocorrect.

 

OK, sounds bad from pure benching perspective, since presumably graphics benchchmarking software doesn't verify correctness of rendering (i.e. artefacts are fine as long as a run completes). 

[Papusan]

1 hour ago, Etern4l said:

 

OK, sounds bad from pure benching perspective, since presumably graphics benchchmarking software doesn't verify correctness of rendering (i.e. artefacts are fine as long as a run completes). 

Or put it the other way... People run their hardware at the edge of stability to max out scores or frames pr. sec (gaming). If the autocorrect has to kick in everytime and you will see lower scores or fps regardless how you see it. I expect this nice professional feature in NCP also will affect your max performance in gaming.

 

Edit: I haven't such cards so I can't test it with a decent overclock. 

 

Ps @Mr. Fox I didn't think this was possible. UL is the Futuremark owner and run certification business. And still can't make futuremark any better than what a single person without the skills can run a small shop. 

https://www.3dmark.com/compare/pr/1890894/pr/1864506/fs/28868586

[Etern4l]

17 minutes ago, Papusan said:

Or put it the other way... People run their hardware at the edge of stability to max out scores or frames pr. sec (gaming). If the autocorrect has to kick in everytime and you will see lower scores or fps regardless how you see it. I expect this nice professional feature in NCP also will affect your max performance in gaming.

 

Edit: I haven't such cards so I can't test it with a decent overclock. 

 

Yeah, so obviously EDR is painful, but ECC should in theory have much less overhead, if any at all. Would be good to see some detailed benchmarks for the 3090Ti or 4090.

 

Back in the day it wasn't very clear:

 

https://www.techspot.com/article/845-ddr3-ram-vs-ecc-memory/

 

The fact that the ECC option exist is very appealing for professional applications, and makes sense given that these are "pseudo-Titan" class cards prosumers might like.