• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Anyone planning to buy the RX 6900 XT, rather than RTX 3080?

nvidiamd said:
That's funny it was 17805 on monday, it is 17873 now but i still haven't proved anything. That is 68 more points in 5 days.
Of course the performance is mostly increased by overclocking but this is how the time spy works. It is mostly increased by overclocking on the 3080 ( and there you have even some LN2 scores keeping the average score up ) and mostly increased by overclocking on the 6800xt . The difference is that on the 6800xt you see only overclocks that anyone can use, the best score is from a reference card. And it is almost 2k above your silly ASUS 3080 example. :)
Click to expand...

Very few RTX 3080's were put on LN2, mostly RTX 3090 got the LN2 treatment. Also again 68 points is like <1 of a FPS average. You need to show proof. All you have is a weak hypothesis. Its a starting point for investigation, not the conclusion for one. You are just repeating a hasty generization, over and over again like its the truth. Like you are right. You dont have a point and you can never win without evidence. Evidence that will take you the next 6 months to get.
 
zx128k said:
Very few RTX 3080's were put on LN2, mostly RTX 3090 got the LN2 treatment. Also again 68 points is like <1 of a FPS average. You need to show proof. All you have is a weak hypothesis. Its a starting point for investigation, not the conclusion for one. You are just repeating a hasty generization, over and over again like its the truth. Like you are right. You dont have a point and you can never win without evidence. Evidence that will take you the next 6 months to get.
Click to expand...
Right now it is 200 points above the 3080. :)
 
eeii said:
Wait tell you get near your 50's you really do't give a duck :D:D:D:D
Click to expand...

You see this is what im talking about. Im reading most of this benchmark stuff about which card is faster in firestrike or which one grows magic wings under LN2 and just don't care. I watched the release day reviews, it's faster than my old card in most things (ill probably keep the 7 for it's pro drivers) and run the 7 for my compute workflow (unless they fix the 6000 series in compute) and have the 6900xt for some games etc. This is what im thinking, they never needed to beat Nvidia for me they needed to beat the Radeon 7 in the things I do.
 
nvidiamd said:
Right now it is 200 points above the 3080. :)
Click to expand...

The main problem is 6800xt bios lets max possible 2750mhz. (2800-50mhz always) That's why 6800xt's 3dmark scores is always lower than it should be. Also amd's wattman is not working properly and always pulls back the card.

When there is no artificial barrier 6800xt scores should be much more than today.

It seems that the best binned 6900xt can hold 2730mhz avg even there is no limit for it. Less core helps 6800xt might hit 2900 or more on air.

edit: 2750mhz is only max possible. In reality it's around 2740mhz or less.
 
Last edited:
xcx xcxvgyt said:
The main problem is 6800xt bios lets max possible 2750mhz. (2800-50mhz always) That's why 6800xt's 3dmark scores is always lower than it should be. Also amd's wattman is not working properly and always pulls back the card.

When there is no artificial barrier 6800xt scores should be much more than today.

It seems that the best binned 6900xt can hold 2730mhz avg even there is no limit for it. Less core helps 6800xt might hit 2900 or more on air.

edit: 2750mhz is only max possible. In reality it's around 2740mhz or less.
Click to expand...

IDK if they are lower than they should be. The fact that you can do 2.5-2.6 Ghz on any reference card, shows it is a good card. You will be a fool not to undervolt/overclock it because it works fine around 2.5 Ghz. The biggest frequencies are just for the show. Not to many people will keep their cards above the 2.6 Ghz.
I am not really interested in these benchmarks, but in the past when the 6800xt beat the record on Firestrike, the Nvidia gang on this forum told us that the 6800xt is not good in Timespy because it was below the 3080 at that time. So yeah the 6800xt might be good on an old DX11 benchmark but the 3080 is the next gen DX12 god. It doesn't look like a god anymore. :)

The 6800xt is kept that low by the frequency limit and the power limit. The 6900xt is kept back by the power limit. We will see higher frequencies once the AIB cards will come because there are some with 3x8 power connectors and most likely a higher power limit. We will also see higher scores in benchmarks but again, this doesn't mean too much but it is interesting to see how high it can go.
But if one can keep the card @ 2.5 - 2.6Ghz on air at a nice temperature, that is already impressive.
 
Yes impressive but can be more impressive, best 6800xt time spy score is less than ln2+3080 just because of bios limitations.

3080&ln2: GT1 141.21 fps Gt2 123.14 fps
6800xt on air: GT1 140.34 fps Gt2 117.1 fps

I believe one of the unlimited 6800xt on air will beat every 3080s soon.

It's hard to understand why AMD created such a barrier.
 
nvidiamd said:
Right now it is 200 points above the 3080. :)
Click to expand...

So you have a small sample size and a fluctuation in the average. Were is your graph showing that this fluctuation has an upward trend? Your analyses this will lead to statistically significant increase in performance? ~1fps average increase in performance up from 140fps, why would I care?

Also as SAM/PCIe4 is a factor, show that the performance increase is just the graphics card and not down to the number of people using a Ryzen 5000 series cpu increasing in numbers in the data.
 
Last edited:
xcx xcxvgyt said:
I'm not sure their bios let them to give more power than AMD limitation. I'm might be wrong.
Click to expand...

Yep, its confusing, for me too.

The way I think its been done is that the 'Power Limit' and the 'Total Board Power' are separate values, the PL is software clipped and the TBP is hardware clipped.

It's hard to speculate with the 6900XT but the 6800XT i think goes like this:

Reference 6800XT - 300W Average TBP - 330W peak
Red Devil 6800XT - 'over' 400W Average TBP - i assume the peak is approx 450W.

Which means that the reference card can go over 300W TBP and provide a peak of 330W but the firmware/circuitry will reign it back to 300W or under. Which simply means to me that if you set +15% Powerlimit on the slider, on the Red Devil, is that you will be allowed to pull +15% over the maximum TBP peak of the reference 6800XT's 330W TBP.

This works out as 330W+50Watts which, handily, is the observed max peak power draw of the 6800XT Red Devil when using PCAT - approx 380watts.

From what i have read, is that the Red Devil does not clip power from the 6800XT @ 380watts but the clock speed limit of 2753mhz max clock is there as it was demonstrated to be the case, this is probably a software limit. Why that clock limit is there, only AMD know.

What's does that mean to the 6900XT ? Well, from what i understand the reference 6900XT Reference TBP is exactly the same as the 6800XT Reference. Why AMD did this, only they will know but the public reason is that the 6900XT is a binned 6800XT and through the binning process, will draw the same or less power eventhough the chip has more enabled cores.

The million dollar question is whether the 6900XT AIB boards will provide over a 380watt average TBP to the chip. It really needs to be over 400 watts average TBP and 440 watt peak to fully feed the chip.

I don't personally believe that any of the 6900XT AIB cards fitted with an air cooler will be configurable to do it (440 watt peak) out of the box and 380 watts will be the limit. And this is probably the reason AMD was not sure whether AIB's should do 6900XT cards at all.

That said, AIO/Liquid cards like the Asus 6900XT Strix LC, Liquid Devil etc. should be fitted with 400w+ firmware so we need to wait and see.

The peak TBP of the 5700XT Liquid Devil when under OC was .........approx 360 watts which was well over what the Red Devil was pushing therefore my prediction is that we could see a 2753mhz 6900XT Liquid Devil at 425watts and at that sustained clock speed, it puts the 6900XT in the 22-24k Timespy range - no shunt mod, no BIOS crossflash.
 
Last edited:
Why wouldn't you trade a 6900 xt for a 3090? You would have to be quite brand focussed not to IMO.

Why did the youtuber want the 6900 xt btw ?
 
Going back to clock limit talk, it seems unless there is a problem with high clocks killing chips, then it is only harming AMD to be limited on these top end cards.

IMO they NEED to show they can hang and potentially dominate in the raster games.
 
Besty said:
The peak TBP of the 5700XT Liquid Devil when under OC was .........approx 360 watts which was well over what the Red Devil was pushing therefore my prediction is that we could see a 2753mhz 6900XT Liquid Devil at 425watts and at that sustained clock speed, it puts the 6900XT in the 22-24k Timespy range - no shunt mod, no BIOS crossflash.
Click to expand...

https://www.3dmark.com/fs/24397663

How is that possible? There must be some kind of magic.
 
Besty said:
Yep, its confusing, for me too.

The way I think its been done is that the 'Power Limit' and the 'Total Board Power' are separate values, the PL is software clipped and the TBP is hardware clipped.

It's hard to speculate with the 6900XT but the 6800XT i think goes like this:

Reference 6800XT - 300W Average TBP - 330W peak
Red Devil 6800XT - 'over' 400W Average TBP - i assume the peak is approx 450W.

Which means that the reference card can go over 300W TBP and provide a peak of 330W but the firmware/circuitry will reign it back to 300W or under. Which simply means to me that if you set +15% Powerlimit on the slider, on the Red Devil, is that you will be allowed to pull +15% over the maximum TBP peak of the reference 6800XT's 330W TBP.

This works out as 330W+50Watts which, handily, is the observed max peak power draw of the 6800XT Red Devil when using PCAT - approx 380watts.

From what i have read, is that the Red Devil does not clip power from the 6800XT @ 380watts but the clock speed limit of 2753mhz max clock is there as it was demonstrated to be the case, this is probably a software limit. Why that clock limit is there, only AMD know.

What's does that mean to the 6900XT ? Well, from what i understand the reference 6900XT Reference TBP is exactly the same as the 6800XT Reference. Why AMD did this, only they will know but the public reason is that the 6900XT is a binned 6800XT and through the binning process, will draw the same or less power eventhough the chip has more enabled cores.

The million dollar question is whether the 6900XT AIB boards will provide over a 380watt average TBP to the chip. It really needs to be over 400 watts average TBP and 440 watt peak to fully feed the chip.

I don't personally believe that any of the 6900XT AIB cards fitted with an air cooler will be configurable to do it (440 watt peak) out of the box and 380 watts will be the limit. And this is probably the reason AMD was not sure whether AIB's should do 6900XT cards at all.

That said, AIO/Liquid cards like the Asus 6900XT Strix LC, Liquid Devil etc. should be fitted with 400w+ firmware so we need to wait and see.

The peak TBP of the 5700XT Liquid Devil when under OC was .........approx 360 watts which was well over what the Red Devil was pushing therefore my prediction is that we could see a 2753mhz 6900XT Liquid Devil at 425watts and at that sustained clock speed, it puts the 6900XT in the 22-24k Timespy range - no shunt mod, no BIOS crossflash.
Click to expand...

Basically the curent maximum overclock for the 6900xt. This result is at an average frequency of 2,710 MHz for the 6900xt. The graphics score is 21563. This disproves that a 2753MHz 6900xt can reach 22-24k time spy. Remember if power limits are reached then the clock frequency will drop. This has to happen because the GPU must be provided with enough power to stop voltage dropping for the core. If voltage drops because of the power limit then frequency must drop. Or the gpu core will become unstable. Also just because you increase the power limit does not imply that the core can reach higher frequencies. The 6800xt has a hard BIOS set limit of 2.8Ghz. The 6900xt has a hard bios set limit of 3Ghz. https://phonemantra.com/the-radeon-rx-6900-xt-has-a-crazy-gpu-frequency-limit-bios-limited-to-3-ghz/ This is the extreme maximum for these cards, you would need to bin the gpus to find out how well they overclock. Time Spy shows that the 6900xt and 6800xt overclock to the same maximum frequency. https://www.3dmark.com/spy/16240104 2,716 MHz 6800xt Graphics Score 20 929

There is now evidence of much more overclocking in the 6800xt and 6900xt results. Most of the 6800xt's are now near 20k overclocking wise. Same with the 6900xt which is approx 20k for all the top 100. As both the AMD cards have low sample sizes this could cause an increase in the average. Remember the average value is not the same as the card you will buy. Reviews show a much lower stock score (thats the score you should expect when buying a card). As more cards are sold and more scores are benchmarked the average should become more accurate.

Remember the mean in the bell curve is the most common value for performance. If you look at both AMD cards the average is more to the right. The 3080 less so.
 
Last edited:
zx128k said:
This disproves that a 2753MHz 6900xt can reach 22-24k time spy.
Click to expand...

There is no 'proof' or 'disproof' because nobody has benched a 400W+ 2753mhz 6900XT on air yet, its all theoretical.

It's fairly simple mathematics, 10% quicker than a top-bin reference 6800XT on air is approx 23025. No warranty voiding cross-flash, no liquid chiller, no hardware shunt-mod.

The reference 6900XT's cooler can't cope, the power-limit can't cope, its all a bit of a mess for that card but its fairly clear that the chip+vram has the theoretical legs to get there.

We just need the unicorn tears to show up in the OCUK loading bay, it will be fun to find out.
 
Besty said:
There is no 'proof' or 'disproof' because nobody has benched a 400W+ 2753mhz 6900XT on air yet, its all theoretical.

It's fairly simple mathematics, 10% quicker than a top-bin reference 6800XT on air is approx 23025. No warranty voiding cross-flash, no liquid chiller, no hardware shunt-mod.

The reference 6900XT's cooler can't cope, the power-limit can't cope, its all a bit of a mess for that card but its fairly clear that the chip+vram has the theoretical legs to get there.

We just need the unicorn tears to show up in the OCUK loading bay, it will be fun to find out.
Click to expand...

Someone has already run time spy with a 6900xt with an average frequency of 2.7GHz. I posted the top result for the 6800xt and its not 23K in time spy. 6900xt average frequency 2.71GHz https://www.3dmark.com/spy/16436136 Graphics score 21563 The score could increase with a higher power limit but its not reaching 23k-24k in time spy.

The 3090 could get 23.539k on water in time spy. Example https://www.3dmark.com/spy/16076861

Expect lower clocks in port royal. Like 2.6Ghz water for the 6900xt. Average temperature 34 °C
https://www.3dmark.com/pr/638640 This too is the highest OC which net someone a score close to a 3080 stock.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom