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

[zx128k]

Don't worry about the 6900xt yet, please explain why can't we see any 3080 doing at least 20600 points on water or on air? If it is within the overclocking range, we should see a some of them going over 20600 points but there is none.
How long before Nvidia will release some tricky drivers to beat the Big Navi in 3d mark like they did in the past when they were left behind?
Meanwhile the 6800xt...
https://www.3dmark.com/spy/16240104

RTX 3080 hits a maximum of 2.2GHz on water. https://www.3dmark.com/spy/16203941 Graphics Score 20 573 Its LN2 for higher. This makes the RTX3080 far faster at maximum clock than the 6900xt atm. Remember the RT performance is massively higher. Atm from the data can reach 21k on the 6900xt but this is LN2 level for the RTX3080. The 6900xt is 11k in port royal highest air https://www.3dmark.com/search#advanced?test=pr P&cpuId=&gpuId=1353&gpuCount=1&deviceType=ALL&memoryChannels=0&country=&scoreType=overallScore&hofMode=false&showInvalidResults=false&freeParams=&minGpuCoreClock=&maxGpuCoreClock=&minGpuMemClock=&maxGpuMemClock=&minCpuClock=&maxCpuClock=. RTX 3080 is 13369 air port royal https://www.3dmark.com/pr/579484

Overall maximum overclock the RTX 3080 is faster. Most ppl are going to get overlocks that are random but much lower than these overclocks. The RTX 3090 just dominates once overclocked. Seen videos of 17k Port royal on LN2, the RTX 3080 get 14k.

There is no reason to get a 6900xt, the same with the 3090. The 6800xt overclocks well.

6800xt gets https://www.3dmark.com/pr/632833 10595 in port royal maximum.
6800xt gets https://www.3dmark.com/spy/16240104 20929 maximum in Time spy graphics.
The maximum 6900xt is just a little faster.

 

[nvidiamd]

RTX 3080 hits a maximum of 2.2GHz on water. https://www.3dmark.com/spy/16203941 Graphics Score 20 573 Its LN2 for higher. This makes the RTX3080 far faster at maximum clock than the 6900xt atm. Remember the RT performance is massively higher. You atm from the data can reach 21k on the 6900xt but this is LN2 level for the RTX3080. The 6900xt is 11k in port royal highest air https://www.3dmark.com/search#advanced?test=pr P&cpuId=&gpuId=1353&gpuCount=1&deviceType=ALL&memoryChannels=0&country=&scoreType=overallScore&hofMode=false&showInvalidResults=false&freeParams=&minGpuCoreClock=&maxGpuCoreClock=&minGpuMemClock=&maxGpuMemClock=&minCpuClock=&maxCpuClock=. RTX 3080 is 13369 air port royal https://www.3dmark.com/pr/579484

Overall maximum overclock the RTX 3080 is faster. Most ppl are going to get overlocks that are random but much lower than these overclocks. The RTX 3090 just dominates once overclocked. Seen videos of 17k Port royal on LN2, the RTX 3080 get 14k.

So you are saying that since the RTX 3080 can only do 2.2Ghz on water, it is faster than a 2.7-2.8 6800xt because the 6800xt won't get 20573 if it is clocked to 2.2Ghz? Or what?
The clock is only a part of the problem. Right now on normal conditions (water or air), the 3080 can do less than 20600 and the 6800xt went over 20900 points.
The AIBs 6900xt will go much higher than that. I won't speculate how high will they go right now but i've told you before that the 6800xt will go above the 3080.

Its nice that the 6800xt can get a good score in Firestrike DX11.

Yet there are no 6800xt's in the time spy hall of fame yet. Come on overclockers. There is even a 2080ti at 45 with a 2.76GHz core clock. https://www.3dmark.com/spy/6991533 We all know why the 6800xt is strong in DX11 (requires 1 GB video card memory, what size is the infinity cache? 128MB ROPS etc) but everyone wants to know the LN2 DX12 performance.

Where are they now?

good point, amd still weak at DX12 but good at dx11? And they laughed when I said big Navi for old games and ampere for new

Actually the average score in TimeSpy is 17685 (3080) vs 17557 (6800xt)
Since 3080 is an older graphic card and the drivers, overclockers and custom AIB's have already sucked all the juice from the card, we should expect to see the 6800xt in taking the lead pretty soon.

Look where they are now: 17693 for the 3080 and 17810 for the 6800xt.

 

[zx128k]

So you are saying that since the RTX 3080 can only do 2.2Ghz on water, it is faster than a 2.7-2.8 6800xt because the 6800xt won't get 20573 if it is clocked to 2.2Ghz? Or what?
The clock is only a part of the problem. Right now on normal conditions (water or air), the 3080 can do less than 20600 and the 6800xt went over 20900 points.
The AIBs 6900xt will go much higher than that. I won't speculate how high will they go right now but i've told you before that the 6800xt will go above the 3080.


Where are they now?

Look where they are now: 17693 for the 3080 and 17810 for the 6800xt.

The top 100 scores are cherry picked exceptions to the rule. Many custom RTX 3080's will hit 19k air cooled. Thats the 6900xt range. So will the 6800xt. The RTX 3080 will have masssively faster RT performance over both the 6800xt and 6900xt.

Maximum score are not representative of a distrabution. They are outliers, so treat them for what they are (exceptions to the rule). Like people that jump out of planes and their parachute fails. Then manage to survive hitting the ground at terminal velocity.

NVidia went with more cores, lower frequency and AMD went with less cores and higher frequency. If a game does not max out the AMD 6800xt cores then the frequency will make it faster. Once AMD's hardware is max'ed out, then Nvidia's design will pull ahead. This is why no one can say Nvidia are slower at 4k. NVidia's gpu's can't clock higher because its like comparing a 64 core to an 32 core.

6800xt went over 20900 points for only one card. There are only three 6800xt's over 20600. This makes the 3080 and the 6800xt on power but thats not the whole story. At maximum clocks 6800xt is more or less equal to the 6900xt. https://www.3dmark.com/search#advanced?test=spy P&cpuId=&gpuId=1348&gpuCount=1&deviceType=ALL&memoryChannels=0&country=&scoreType=graphicsScore&hofMode=false&showInvalidResults=false&freeParams=&minGpuCoreClock=&maxGpuCoreClock=&minGpuMemClock=&maxGpuMemClock=&minCpuClock=&maxCpuClock=


https://www.3dmark.com/search#advanced?test=spy P&cpuId=&gpuId=1353&gpuCount=1&deviceType=ALL&memoryChannels=0&country=&scoreType=overallScore&hofMode=false&showInvalidResults=false&freeParams=&minGpuCoreClock=&maxGpuCoreClock=&minGpuMemClock=&maxGpuMemClock=&minCpuClock=&maxCpuClock= enter the gpu you want to look at and then change it from overall score to graphics. Then you will find the following:
All of the top 100 RTX 3080 are above the 20k graphics score in time spy. With the 6800xt its only the top 43 that are above 20k. With the 6900xt its only the top 46 cards that are above 20k. Then you have the masive lead the RTX 3080 has in RT.

 

[nvidiamd]

You still don't understand how it works. AMD users had a lot less time to test their cards. Nvidia users had more time. AMD cards are almost unexistent ATM. That's why i told you before that the 6800xt will go over the 3080 and we will see even a bigger gap. Of course there will be more 3080's to go over 20k because there are more Nvidia owners and more people who buy Nvidia are interested in their 3d mark performance. But it looks like no normal cooled 3080 ( at least so far ) is able to go over 20600. We will see if one will do it in the future.
Again i don't want to speculate too much but soon you will see 6900xt's going over 22k on air or water. I don't know where they will stop but they will go much higher than the 6800xt or the 3080.
It is wrong to think that this is the 6900xt performance, as it was wrong when you laughed about the 6800xt performance. It was clear from the results available at that time that the 6800xt will get far better scores on TimeSpy. Once the custom 6900xt will come, we will see they are way better.

 

[zx128k]

You still don't understand how it works. AMD users had a lot less time to test their cards. Nvidia users had more time. AMD cards are almost unexistent ATM. That's why i told you before that the 6800xt will go over the 3080 and we will see even a bigger gap. Of course there will be more 3080's to go over 20k because there are more Nvidia owners and more people who buy Nvidia are interested in their 3d mark performance. But it looks like no normal cooled 3080 ( at least so far ) is able to go over 20600. We will see if one will do it in the future.
Again i don't want to speculate too much but soon you will see 6900xt's going over 22k on air or water. I don't know where they will stop but they will go much higher than the 6800xt or the 3080.
It is wrong to think that this is the 6900xt performance, as it was wrong when you laughed about the 6800xt performance. It was clear from the results available at that time that the 6800xt will get far better scores on TimeSpy. Once the custom 6900xt will come, we will see they are way better.

You need to prove all that. Every point, when you cant understand why you are wrong. You dont have the information to make that kind of point.

Why you could be wrong. The data could imply that the RTX 3080 more reliably reaches 2.1GHz and hits 20k. Whereas the 6800xt/6900xt more rarely reaches its higher clocks, thus has fewer cards above 20k in time spy. Based on the available data. Anything else cant be proven. In future with more data one could change the conclusion. Atm both the 6800xt and RTX 3080 are sold out everywhere. The time spy data is what it is and the sample size is large enough for an opinion. You cant get data for every gpu released. Numbers sold and the overclocking potential. You just have to accept the data available and avoid wishful thinking or special pleading. You always have the data you have got and have to work with it.

 

[JediFragger]

****, thought we got rid of this ****.

 

[nvidiamd]

You need to prove all that. Every point, when you cant understand why you are wrong. You dont have the information to make that kind of point.

Why you could be wrong. The data could imply that the RTX 3080 more reliably reaches 2.1GHz and hits 20k. Whereas the 6800xt/6900xt more rarely reaches its higher clocks, thus has fewer cards above 20k in time spy. Based on the available data. Anything else cant be proven. In future with more data one could change the conclusion. Atm both the 6800xt and RTX 3080 are sold out everywhere. The time spy data is what it is and the sample size is large enough for an opinion. You cant get data for every gpu released. Numbers sold and the overclocking potential. You just have to accept the data available and avoid wishful thinking or special pleading. You always have the data you have got and have to work with it.

Of course i can prove it but i don't know you can understand.
Here you see all the test made with the 3080. And the average score for all the cards tested:

And here you see all the tests made with the 6800xt. And the average score for all the cards tested

So what can we see here? The 3080 owners have done more than 30 times more Timespy tests than the 6800xt owners. Of course you have more 3080 cards over 20k it is pure mathematics. In fact you should have up to 30 times more 3080 cards over 20k points, than 6800xt.
The fact that 6800xt wins on average, shows that it is already a better overclocker. The gap grows bigger and bigger every day. And the maximum overclock is more than 300 points ahead for 6800xt on cards that run on water/air.
It is not hard to look at the numbers and understand what's going on. I told you that the 6800xt will go over 3080 when it was 130 points behind on average. It was obvious that once the AIB will release their cards, AMD will get better scores.
I am telling you again now: You will see the 6900xt going over 22k in less than 4 weeks. On water/air.

 

[zx128k]

Of course i can prove it but i don't know you can understand.
Here you see all the test made with the 3080. And the average score for all the cards tested:

And here you see all the tests made with the 6800xt. And the average score for all the cards tested

So what can we see here? The 3080 owners have done more than 30 times more Timespy tests than the 6800xt owners. Of course you have more 3080 cards over 20k it is pure mathematics. In fact you should have up to 30 times more 3080 cards over 20k points, than 6800xt.
The fact that 6800xt wins on average, shows that it is already a better overclocker. The gap grows bigger and bigger every day. And the maximum overclock is more than 300 points ahead for 6800xt on cards that run on water/air.
It is not hard to look at the numbers and understand what's going on. I told you that the 6800xt will go over 3080 when it was 130 points behind on average. It was obvious that once the AIB will release their cards, AMD will get better scores.
I am telling you again now: You will see the 6900xt going over 22k in less than 4 weeks. On water/air.

Thats within error margins for the score averages. Few 100 points here and there is nothing. You can get that level of change from a change of driver version. Saying the gap grows day on day would need a graph showing the trend which is impossible from a bell curve. Stating day on day is wrong, driver version dont come out day on day and overclocking is not the norm. You leave yourself in an impossible to prove place. Stop with the sophistry. You need to show a day on day graph showing the trend which is an increase in graphics score. Please provide a value for this increase in the slope. https://www.youtube.com/watch?v=MYCH7gswI4k and the equation for the graph. y=mx+c for example.

If you look at the bell curve for the 6800xt you can see that the average is higher than the mean. The mean or most common value is the peak in that bell curve, https://www.investopedia.com/terms/b/bell-curve.asp much lower than the average. The skewness is more to the left than the 3080 or more towards less peroformance. You can even see in the gpu core clock values that it is wider and most values are towards the low end of frequency. Most values for the 3080 are toward the high end of frequency with less toward the low end.

 

[nvidiamd]

Ok let's stop here because it seems like you don't understand what the average is. Otherwise you would not say things like " it is within error margins" or "You can even see in the gpu core clock values that it is wider and most values are towards the low end of frequency. Most values for the 3080 are toward the high end of frequency with less toward the low end." That is some BS right there.
If the 3080 was 128 points ahead on the November 22nd (17685 (3080) vs 17557 (6800xt) ) and it is 112 points below now (17693 vs 17805 ), it means AMD score has improved in these 3 weeks by a lot, while the 3080 only improved its score a little (8 points on average). Btw, the AMD average is 17815 right now. So there you have the proof that the gap grows bigger and bigger every day.

AMD Navi2/RDNA2 event thread

 

[zx128k]

Ok let's stop here because it seems like you don't understand what the average is. Otherwise you would not say things like " it is within error margins" or "You can even see in the gpu core clock values that it is wider and most values are towards the low end of frequency. Most values for the 3080 are toward the high end of frequency with less toward the low end." That is some BS right there.
If the 3080 was 128 points ahead on the November 22nd (17685 (3080) vs 17557 (6800xt) ) and it is 112 points below now (17693 vs 17805 ), it means AMD score has improved in these 3 weeks by a lot, while the 3080 only improved its score a little (8 points on average). Btw, the AMD average is 17815 right now. So there you have the proof that the gap grows bigger and bigger every day.

AMD Navi2/RDNA2 event thread

With a bell curve the mean is the peak and the average can be different. That statement is in the pictures you posted. Just stating what can be see. Your own "proof" implies what I stated. Yet you have yet to prove your conclusion. Post a graphic showing day or day increases in the 3d mark time spy average. Show were this data came from to prove you did not manufacture it. I know you cant. Everyone knows you just made the whole thing up. Scores do increase with time but driver releases would not fuel day on day increases in performance. A few hundred point is not faster. Its like 0.1 fps more or 1 fps more. You dont get massive points increases with drivers normally.

 

[nvidiamd]

With a bell curve the mean is the peak and the average can be different. That statement is in the pictures you posted. Just stating what can be see. Your own "prove" implies what I stated. Yet you have yet to prove your conclusion. Post a graphic showing day or day increases in the 3d mark time spy average. Show were this data came from to prove you did not manufacture it. I know you cant. Everyone knows you just made the whole thing up. Scores do increase with time but driver releases would not fuel day on day increases in performance. A few hundred point is not faster. Its like 0.1 fps more or 1 fps more. You dont get massive points increases with drivers normally.

I know what a bell curve means but it makes no sense when you have a lot of tests. It makes sense on small numbers, it makes no sense in huge numbers. You can't have most of the cards doing better than the average because that will mean the worst cards are doing something like 10k or 5k to be able to drag the average score low.
I already posted a link to my post from 22nd of November. When 6800xt was 128 points below the 3080 on average. Yesterday it was 112 points above the 3080. And yet, you ask me to show you evidence that the gap grows bigger every day. I posted today's score earlier which was 17815. It is 17820 now, while the 3080 has the same average score.

To say that few hundreds points is not faster when we talk about the average score, shows that you have no idea what you are talking about. Even an average of 50 points lead can prove that a card is a better overclocker than the other. It is not only about drivers, it is more people having access to the hardware, and custom cards coming to market with better cooling solutions.

There you have it, 15 points more in one day vs 0 more points for the 3080. You can check that yourself.

 

[zx128k]

I know what a bell curve means but it makes no sense when you have a lot of tests. It makes sense on small numbers, it makes no sense in huge numbers. You can't have most of the cards doing better than the average because that will mean the worst cards are doing something like 10k or 5k to be able to drag the average score low.
I already posted a link to my post from 22nd of November. When 6800xt was 128 points below the 3080 on average. Yesterday it was 112 points above the 3080. And yet, you ask me to show you evidence that the gap grows bigger every day. I posted today's score earlier which was 17815. It is 17820 now, while the 3080 has the same average score.

To say that few hundreds points is not faster when we talk about the average score, shows that you have no idea what you are talking about. Even an average of 50 points lead can prove that a card is a better overclocker than the other. It is not only about drivers, it is more people having access to the hardware, and custom cards coming to market with better cooling solutions.

There you have it, 15 points more in one day vs 0 more points for the 3080. You can check that yourself.

You are not proving your point. You need a graphic of weeks, for each day. I am sorry but you were the one running your mouth. Made big conclusions you could never possibly prove. Posting this weak reply does not cut it. You stated that there was a trend of increased performance but lack the evidence for the increase in the trend for performance. Also you have not isolated the reason for a performance increase. Also a few 100 points is likely less than a single fps increase in performance.

Basically you are stating that the score increased by 17815 to17820 and this is increased performance. Lets look at that.
Graphics Score 21 560 to Graphics Score 21 566, https://www.3dmark.com/spy/14041262 vs https://www.3dmark.com/spy/14039907
thats 141.68fps vs 141.21fps first test.
thats 122.72fps vs 123.14fps second test.

Thats your 5 points.

Say the graphics score droped to Graphics Score 21 517
thats 141.06fps first test.
thats 122.74fps the second test.

Say we drop ~100 points. between 20711 and 20590 for example https://www.3dmark.com/spy/16285093 and https://www.3dmark.com/spy/16283981
Graphics Test 1
135.49 fps
Graphics Test 2
117.07 fps

vs

Graphics Test 1
136.35 fps
Graphics Test 2
117.7 fps

See how very tiny that increase is. Your whole conclusion is a farce. Fractions of a fps for 100 points, not even 1 fps. I have run time spy and get your level of increase and decrease in performance based on driver version changes (not always tha graphic card drivers). Lets ignore the fact you can buy a RTX 3080 like the Asus rog strix 3080 gaming oc and get 18.5 graphics score easy at stock settings. Overclock it and get over 19k and match a 6900xt without effort.

 

[eeii]

So I take it your not buying a 6900xt then

 

[Gerard]

6900xt Cards from amd direct come wrapped with a mousepad.

 

[Grim5]

6900xt Cards from amd direct come wrapped with a mousepad.

Thats so lame, where is the box

 

[Besty]

****, thought we got rid of this ****.

unfortunately not.

Fortunately the ‘other forum’ has 300pages of 3080 owners showing how they got these 3dmark scores. buying half a dozen cards to beat the silicon lottery, double Solder on the pcb boards and cross-flashing BIOS.

 

[Gerard]

Thats so lame, where is the box

Obviously it has a box, its wrapped up in that inside the box.

 

[Dontrocktheboat]

7,000 benches on 6800xt
250,000 benches on 3080



Joke launch

 

[Grim5]

7,000 benches on 6800xt
250,000 benches on 3080



Joke launch

Common man, don't do AMD like that. AMD owners just don't like running benchmarks

 

[TheRealDeal]

7,000 benches on 6800xt
250,000 benches on 3080



Joke launch

How many of them are repeat benches trying to beat the 6800xt to save there Epeen. I would hazard a guess its a very high number.