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

[Mr. Fox]

My 80°F office is definitely not good for air cooling. Even testing with Crysis 3 Remastered, hotspot is 90°C with an undervolt, and TimeSpy hotspot over 100°C, so I am going to stop doing things until I get the block installed as I do not think it is good for silicon to run this hot. A couple of guys on ExtremeHW tell me it is normal for their 6900 XT to have hotspots in the 100°C range and "not to worry about it, been running it that way a couple of years now" but, I still don't think it is a good thing. It can't be helpful, LOL. I attribute at least 10°C to my abnormally high ambient temperatures and lack of humidity. (My 3060 Ti temperatures were also much higher than desirable before I put a block on it.)

 

Compute performance is really good and impressive on this GPU, but ray tracing not so much. Look at the FPS with ray tracing enabled. It stress the GPU severely. I ran Bright Memory Infinite benchmark and only sitting on the benchmark menu and not running the benchmark the hotspot rose to 90°. Without ray tracing the temps are definitely lower and the frames much higher.

 

[electrosoft]

45 minutes ago, Mr. Fox said:

My 80°F office is definitely not good for air cooling. Even testing with Crysis 3 Remastered, hotspot is 90°C with an undervolt, and TimeSpy hotspot over 100°C, so I am going to stop doing things until I get the block installed as I do not think it is good for silicon to run this hot. A couple of guys on ExtremeHW tell me it is normal for their 6900 XT to have hotspots in the 100°C range and "not to worry about it, been running it that way a couple of years now" but, I still don't think it is a good thing. It can't be helpful, LOL. I attribute at least 10°C to my abnormally high ambient temperatures and lack of humidity. (My 3060 Ti temperatures were also much higher than desirable before I put a block on it.)

 

Compute performance is really good and impressive on this GPU, but ray tracing not so much. Look at the FPS with ray tracing enabled. It stress the GPU severely. I ran Bright Memory Infinite benchmark and only sitting on the benchmark menu and not running the benchmark the hotspot rose to 90°. Without ray tracing the temps are definitely lower and the frames much higher.

 

 

Plus that is the best of the best class 6900xt card. The Asrock Formula OC 6900xt/6950xt have the best PCB design of them all (nevermind the beautiful aesthetics).

 

 

[Mr. Fox]

19 minutes ago, electrosoft said:

 

Plus that is the best of the best class 6900xt card. The Asrock Formula OC 6900xt/6950xt have the best PCB design of them all (nevermind the beautiful aesthetics).

 

 

Yes, it is crazy nice. Looks and build quality are superior.I am eager to see how it behaves on a block. I believe this one has higher clocks and voltage than most, which is also going to affect the temps negatively on air cooling.

[Papusan]

2 hours ago, Mr. Fox said:

A couple of guys on ExtremeHW tell me it is normal for their 6900 XT to have hotspots in the 100°C range and "not to worry about it, been running it that way a couple of years now" but, I still don't think it is a good thing. 

 

My 80°F office is definitely not good for air cooling. Even testing with Crysis 3 Remastered, hotspot is 90°C with an undervolt, and TimeSpy hotspot over 100°C, so I am going to stop doing things until I get the block installed as I do not think it is good for silicon to run this hot.

Amd got roasted because this is to be expected with the 7900 series Radeon cards (110C) when they trotted down clocks due faulty cooling.

 

It's a good thing let the chips have the chance and run wild above 100C. But not if the cooling is already good enough.  100C should only be reached with awful cooling and or the chips run well above the TGP specs. 100C and 300-350W load is awful for a high end graphics card.

5 hours ago, Rage Set said:

 

That is an easy answer, because people were crying about the lack of 12/16 core variant during the launch of the 5800X3D. There was a reason why AMD didn't launch those high core processors and they should have stuck with the same logic. So this time AMD decided to rob those people that wanted the 12/16 core X3D parts. The crazy thing is these parts are still going to sell, regardless of what we think. The 7950X3D is sold out everywhere and the 7900X3D is selling better than it should....

Yep, and now they start complaining about hot running, awful priced chips that has to be castrated to reduce the RMA costs. Not so sure all is happy with the castration from 170W down to 120W TDP. The locked out overclocking is only the Icing on the cupcake

 

 

 

Not easy to please them all😎 But I expect many with the newer AMD platform is happy that AMD lowered TDP  for their new gen X3D chips. Lower TDP = Lower max temp. 

 

And AMD claims they always go after high efficiency over max performance for their Cpu and Gpu chips, so... It match what AMD's gaming div boss Azor want.

[Mr. Fox]

25 minutes ago, Papusan said:

It's a good thing let the chips have the chance and run wild above 100C. But not if the cooling is already good enough.  100C should only be reached with awful cooling and or the chips run well above the TGP specs. 100C and 300-350W load is awful for a high end graphics card.

My impression is that it is caused by the amount of default voltage AMD has to dump on the chip to clock it high enough to deliver competitive performance. While it is a very nice GPU that delivers respectable performance, (ray tracing excluded,) we see a scenario that generally applies to AMD CPUs and GPUs. They have to be pushed harder to compete with Intel and NVIDIA, and it should come as no surprise that managing the thermals in that scenario is going to present a challenge. It is too bad they did not use GDDR6X on the Radeon 69XX and 79XX series. They would not have needed to push the core clock as hard and the performance would have been greatly enhanced with the faster memory and increased bandwidth without a lot of effort on AMD's part.

[Papusan]

6 hours ago, Mr. Fox said:

On a much sadder note, it appears the firmware on this 6900 XT is incompatible with Windows 7. I have spent hours trying to make it work. (Literally stayed up all night trying to get it to work, took a shower and started work.) I tried a half dozen Windows 7 Macrium Reflect images, manually changed to a few different bootx64.efi files (including a custom one that allows pure UEFI mode for Windows 7) and even tried a clean install. 100% failure on all attempts. Windows 7 setup refuses to run with the 6900 XT installed. Tried USB and DVD... both the same. It hangs before it reaches the first screen of the setup GUI. If I put the NVIDIA GPU back in it boots right up like normal. So, it has to be the AMD firmware.

If you can't find a solution... Maybe try ask Chispy for help. He is an retired oc'r from hwbot. Also very helpful person from what I have seen. You find him over on Techpowerup forum. I think he still is active there .

 

He used Win 7 and 6900XT https://hwbot.org/submission/4857577_chispy_3dmark11___performance_radeon_rx_6900_xt_64690_marks

 

The vbios version in the pict

 

Also Nata use Win 7

 

1 hour ago, Mr. Fox said:

It is too bad they did not use GDDR6X on the Radeon 69XX and 79XX series. They would not have needed to push the core clock as hard and the performance would have been greatly enhanced with the faster memory and increased bandwidth without a lot of effort on AMD's part.

AMD go after costs over anything to try compete. They could have used GDDR6X if they wanted. And they could offer higher TGP if they put "more money" into the cooling. 

[Raiderman]

I wonder what the difference  in the firmware is between my 6800xt and that 6900xt, besides the obvious? I know brother @Mr. Fox will have his hands full if he gets it running on windows 7. Not sure if the final AMD windows 7 driver is crippled a bit or the newer drivers for windows 10 are more optimized for the card, but I am a good 1000 points shy with win 7 in 3dmark 11. The better physics score in win 7 won't make up for the better graphics score in windows 10. At least that's what I found.

[Mr. Fox]

43 minutes ago, Raiderman said:

The better physics score in win 7 won't make up for the better graphics score in windows 10. At least that's what I found.

I have never seen an example of that with 3DMark 11 before. Something must be wrong with your Windows 7 installation.  That, or since my last benching spree UL have released a software/system info malware update to deliberately mess with Windows 7 so that Windows 10/11 don't get molested severely. Also, check your Windows Update vomit trail. If KB2670838 is installed, remove it. It will ruin 3DMark 11 performance and cause problems with some games and benchmarks. You have to have that update for IE11, but nobody cares about that anymore. If you have the GUI mods, getting the GUI back to stock will also all Windows 7 to shine bright.

[Papusan]

On 3/9/2023 at 12:28 AM, Papusan said:

What is this? Nvidia pushing out new 4090 silicon revision. Tried manipulating the curve in many ways but even locking the voltage preference does not allow you to go further than 1.07v. Tried on two different windows installations. Used MSI Afterburner 4.6.5 (Beta 4). 

 

Anyone else with Founders Edition with die shots? I contacted Nvidia Customer Care to ask about this and if they had a bios update to enable 1.1v limit.

 

I was trying to figure out why I could not max 1.1v and after taking a look at the Bios and rechecking my teardown pictures. This is a newer die I believe

 

 

Me... This reminds me a lot about when Nvidia remowed the 83mv headroom from the voltage slider in all OC software for my Zotac 980Ti Amp. Extreme (same for all Maxwell cards). Nvidia is a master in manipulating and trickery to max out profits.

Yep, can't be more disgusting than this!!! Maybe the 4090  AIC graphics cards will be the next one with voltage cap at 1.070v. All roads lead to Rome. But Nvidia will keep guard at the the city gate of Rome and none will come out again with more than 1.070v.

[Raiderman]

1 hour ago, Mr. Fox said:

I have never seen an example of that with 3DMark 11 before. Something must be wrong with your Windows 7 installation.  That, or since my last benching spree UL have released a software/system info malware update to deliberately mess with Windows 7 so that Windows 10/11 don't get molested severely. Also, check your Windows Update vomit trail. If KB2670838 is installed, remove it. It will ruin 3DMark 11 performance and cause problems with some games and benchmarks. You have to have that update for IE11, but nobody cares about that anymore. If you have the GUI mods, getting the GUI back to stock will also all Windows 7 to shine bright.

I'll have to check that, plus I did notice that it runs full screen 720p in win 7 as opposed to a cropped/centered 720p in win 10. I wonder if I accidentally ticked a box in the settings of 3dmark.? Yes I run the windows classic gui, and set it to performance in advanced system properties. Looks like windows 2000 when it's all configured.

[Raiderman]

8 hours ago, electrosoft said:

 

Plus that is the best of the best class 6900xt card. The Asrock Formula OC 6900xt/6950xt have the best PCB design of them all (nevermind the beautiful aesthetics).

 

 

Haven't checked the 6900xt, but the 6950xt XFX card has ~50mhz higher boost clock 2324 vs 2368 compared to the Asrock card.

[electrosoft]

 

 

Just now, Raiderman said:

Haven't checked the 6900xt, but the 6950xt XFX card has ~50mhz higher boost clock 2324 vs 2368 compared to the Asrock card.

 

The actual PCB design and power delivery of the Asrock trumps all other cards including XFX. Factory issued boosts never tell the actual story of the silicon or board design.

[Raiderman]

3 minutes ago, electrosoft said:

 

 

 

The actual PCB design and power delivery of the Asrock trumps all other cards including XFX. Factory issued boosts never tell the actual story of the silicon or board design.

The asrock card has a P bios, and a Q bios listed in the specs, with the P bios having a boost clock of 2495mhz

[Papusan]

Some progress in GPUPI 3.3.3 - 32B

 

https://hwbot.org/submission/5222483_papusan_gpupi_v3.3___32b_geforce_rtx_4090_57sec_243ms

 

[electrosoft]

1 minute ago, Raiderman said:

The asrock card has a P bios, and a Q bios listed in the specs, with the P bios having a boost clock of 2495mhz

 

The actual design of the PCB of the Asrock is the best of all the 6900/6950xt cards. Over-engineered past what was really needed actually for most use cases (See: Suprim vs Trio or bottom barrel Gigabyte 4090). If Gigabyte slaps a higher boost on their 4090 vs a Suprim or HOF that doesn't mean its better. That is just marketing and sales tactics.

 

Same applies here for XFX. They build some tank like cards but the actual PCB design is inferior to the Asrock. Even if it boosts 300mhz+ higher (exaggerating) the design still is inferior on a hardware level.

 

Does that mean anything in the end? Maybe when pushed to the absolute extremes but it is what it is.

 

[Etern4l]

2 hours ago, Papusan said:

Yep, can't be more disgusting than this!!! Maybe the 4090  AIC graphics cards will be the next one with voltage cap at 1.070v. All roads lead to Rome. But Nvidia will keep guard at the the city gate of Rome and none will come out again with more than 1.070v.

 

For clarity's sake, what was your query to NVidia? I guess what they are saying is that NVIDIA doesn't offer BIOS updates for FE cards, but not sure how that's relevant to you. 

[Papusan]

1 hour ago, Etern4l said:

 

For clarity's sake, what was your query to NVidia? I guess what they are saying is that NVIDIA doesn't offer BIOS updates for FE cards, but not sure how that's relevant to you. 

Yep, not relevant for me.... But all know how Dell deal with their customers. All Nvidia AIC cards is at same risk get same treatment. Remember Nvidia did something similar to my Zotac 989Ti Amp. Extreme. They just removed the extra voltage headroom (+87mv) with their newer drivers. So no more use of the voltage slider in any of the OC software.

 

So your golden AIC vbios with voltage adjustment can be rendered useless!! I can get away with a custom bios with fixed over-voltage, but all those with newer graphics cards will be stuck with what Nvidia think your cards should have. No tweaks can fix that. They can even reduce TGP if they want so. Remember noting in specs say 600W for 4090

 

3225mhz on the cores.

https://hwbot.org/submission/5222501_papusan_gpupi_v3.3___1b_geforce_rtx_4090_1sec_14ms

 

https://hwbot.org/submission/5222485_papusan_gpupi_v3.3___32b_geforce_rtx_4090_57sec_208ms

 

 

@Mr. Fox Have in your back-head that AMD often offer loads of non WHQL drivers that UL/Futuremark won't approve. They are the king of beta drivers.

 

 

Btw. Anyone that like this fix? Not the best looking cable salad in your pc box. Sad he never posted pict of the cable bend 🙂 

 

From the link.... Hey guys. I went to a pc technician to help me build my pc but he pig tailed the pci-e cables' connection to the 16 pin adapter which is not recommended.

This technician doesn't want to fix his error and I myself am 2 left handed, so my questions are these: Is this safe?

[Mr. Fox]

So, got the block installed. I think it looks awesome @Rage Set. The kit did not have screws for the backplate (had short screws for naked installation). Screws should be here from Amazon in the next hour or so. Will post more photos once the back plate is installed. It requires M2x10-12mm (have both coming) and the kit included M2x6mm. I ordered stainless Allen screws so they will look nice with the black backplate.

I am hoping having pads behind the core will help hotspot temps (still quite high, but better). I have not touched the firmware yet. I will dive into MPT later today or tomorrow, once I get the temps as good as I can get them without the chiller. It will get better, as it is hitting the power limit and remaining there throughout all of the graphics tests. Totally maxed out with the stock firmware limits.

 

CPU not being pushed hard due to high temps with the EKWB direct die. The SCDD was supposed to arrive yesterday, so hopefully USPS will bring it today.

 

https://hwbot.org/submission/5222705_ | https://www.3dmark.com/3dm/90656314#

 

[Papusan]

Me.... On the way to buy my first K|ngp|n card ever. Me... Have checked the used market the last years for interesting tech things. Now I'm on the way grab it😎 EVGA GeForce GTX 980 K|NGP|N ACX 2.0+ P/N: 04G-P4-5988-KR 

 

$120 for the old. But I enjoy play with the old. More fun than with the new modern. Have a MSI GTX 980 but that low end Gamer card isn't the best one out there. And all should know my feelings about MSI, so....

 

Yep, now I'm forced to buy a 2nd PSU. This old bad boy need more than 2x 8-pin connectors. I don't want swap in and out the dual 12vhpwr pci-e adapter cables for playing  with my older cards. Fun times... Got my first HOF card. Now my first K|ngp|n card. At least the latter is a lot cheaper.

 

 

 

 

 

@electrosoft @ryan Are you ready to see some pict of well used all black crocs?😁 I think even bro Fox will enjoy them. 

[electrosoft]

18 minutes ago, Papusan said:

Me.... On the way to buy my first K|ngp|in card ever. Me... Have checked the used market the last years for interesting tech things. Now I'm on the way grab it😎 EVGA GeForce GTX 980 K|NGP|N ACX 2.0+ P/N: 04G-P4-5988-KR 

 

$120 for the old. But I enjoy play with the old. More fun than with the new modern. Have a MSI GTX 980 but that low end Gamer card isn't the best one out there. And all should know my feelings about MSI, so....

 

Yep, now I'm forced to buy a 2nd PSU. This old bad boy need more than 2x 8-pin connectors. I don't want swap in and out the dual 12vhpwr pci-e adapter cables for playing  with my older cards. Fun times... Got my first HOF card. Now my first K|ngp|n card. At least the latter is a lot cheaper.

 

 

 

 

 

@electrosoft @ryan Are you ready to see some pict of well used all black crocs?😁 I think even bro Fox will enjoy them. 

 

Ugh Crocs 2.0 🙂

 

Personally, one of my favorite things of this thread (amongst many) is you benching old GPUs on new hardware and seeing it pushed to their max.

[tps3443]

On 3/8/2023 at 3:30 PM, Mr. Fox said:

I have not heard of it before. Sound interesting. I wonder how much overhead it requires that would impact performance. I generally use Macrium Reflect to put my system back the way I wanted it before the Redmond Reprobates did something to mess it up.

Wow, that's pretty dog-gone fast internet, bro.

---

The 6900 XT has arrived. Thanks, Brother @Rage Set
I will install it after work.

 

Awesome. I always wanted one of those! As well as the matching Z590 (2) Dimm formula OC motherboards. 
 

These older high end GPU’s are freaking bull-dogs with overclocking though. 

I’m so content with my overclocked 3090KP right now, if I popped in a 4090 I wouldn’t feel the difference unless I turned on a frame counter. 😁

 

Thats good enough for me. I only play 2560x1440

 

[Mr. Fox]

OK, got the screws and now the backplate is installed. Time to do a bit of quick studying and get cracking with the link Brother @Rage Setgave me for the firmware tweaking.
Edit: hotspot reading still shows 92-94°C and at this point I am questioning if the sensor is even accurate.

 

Edited March 10, 2023 by Mr. Fox
comment on hotspot temps

[ryan]

38 minutes ago, Papusan said:

Me.... On the way to buy my first K|ngp|in card ever. Me... Have checked the used market the last years for interesting tech things. Now I'm on the way grab it😎 EVGA GeForce GTX 980 K|NGP|N ACX 2.0+ P/N: 04G-P4-5988-KR 

 

$120 for the old. But I enjoy play with the old. More fun than with the new modern. Have a MSI GTX 980 but that low end Gamer card isn't the best one out there. And all should know my feelings about MSI, so....

 

Yep, now I'm forced to buy a 2nd PSU. This old bad boy need more than 2x 8-pin connectors. I don't want swap in and out the dual 12vhpwr pci-e adapter cables for playing  with my older cards. Fun times... Got my first HOF card. Now my first K|ngp|n card. At least the latter is a lot cheaper.

 

 

 

 

 

@electrosoft @ryan Are you ready to see some pict of well used all black crocs?😁 I think even bro Fox will enjoy them. 

CROCS CROCS CROCS WE WANT TO SEE THE CROCS

[Mr. Fox]

My latest review finally got published. It was finished about 10 days ago and was waiting for final approval. I really like it... way more than I expected to. Check it out... text and video.

 

CORSAIR RM1200x SHIFT Power Supply Review

 

2 minutes ago, Mr. Fox said:

CORSAIR has been involved in computer technology for roughly thirty years. They have become one of the most respected, world-leading providers of computer components and peripherals designed for PC enthusiasts. They celebrated their 10th anniversary selling computer power supplies in 2016, at which time that had sold over ten million PSUs. CORSAIR have remained at the forefront all of these years. As a result, innovation is something we have taken for granted with CORSAIR. They were kind enough to provide us with their latest flagship power supply from the RM series for review. It is a distinct honor and privilege to introduce you to the CORSAIR RM1200x SHIFT power supply.

 

 

 

[Rage Set]

5 hours ago, Mr. Fox said:

So, got the block installed. I think it looks awesome @Rage Set. The kit did not have screws for the backplate (had short screws for naked installation). Screws should be here from Amazon in the next hour or so. Will post more photos once the back plate is installed. It requires M2x10-12mm (have both coming) and the kit included M2x6mm. I ordered stainless Allen screws so they will look nice with the black backplate.

I am hoping having pads behind the core will help hotspot temps (still quite high, but better). I have not touched the firmware yet. I will dive into MPT later today or tomorrow, once I get the temps as good as I can get them without the chiller. It will get better, as it is hitting the power limit and remaining there throughout all of the graphics tests. Totally maxed out with the stock firmware limits.

 

CPU not being pushed hard due to high temps with the EKWB direct die. The SCDD was supposed to arrive yesterday, so hopefully USPS will bring it today.

 

https://hwbot.org/submission/5222705_ | https://www.3dmark.com/3dm/90656314#

 

 

I am not happy they didn't include the backplate screws...which makes zero sense. Now I really wish I had blocked that card when I first got the waterblock in. Unfortunately, there is not much I can do since I purchased that block over a year ago. I am still going to e-mail the manufacturer.  

 

EDIT: I am wondering if those temps are due to the increased power limits of this model. My blocked MSI 6800XT (stock voltage) doesn't go anywhere near those temps.