Soldato
Wut.
-
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.
OcUK RX5800X3D review thread
- Thread starter CAT-THE-FIFTH
- Start date
Caporegime
@CAT-THE-FIFTH A new review has just popped up for the 5800X3D from Igors lab. AMD Ryzen 7 5800X3D - the only (and last) fighter of its kind as a perfect and very efficient upgrade | igor'sLAB (igorslab.de)
It shows the 5800X3D is the fastest CPU tested, while using the least power of every CPU tested.
It shows the 5800X3D is the fastest CPU tested, while using the least power of every CPU tested.
Caporegime
@CAT-THE-FIFTH A new review has just popped up for the 5800X3D from Igors lab. AMD Ryzen 7 5800X3D - the only (and last) fighter of its kind as a perfect and very efficient upgrade | igor'sLAB (igorslab.de)
It shows the 5800X3D is the fastest CPU tested, while using the least power of every CPU tested.
That power consumption chart is helarious, the 12900KS uses about 80% more power for 2% better performance, that's Intel scaling for you, contrast that with the 5800X3D which is significantly more power efficient for significantly higher performance than the CPU its based on.
I am quite looking forward to seeing how much of an improvement this will be for my 2200G.
They use geomean on the % difference for each game which is the correct way to do it when comparing A to B.
Other methods get wonky results at the extremes or depending on which order the comparison is done.
They use geomean on the % difference for each game which is the correct way to do it when comparing A to B.
Other methods get wonky results at the extremes or depending on which order the comparison is done.
Caporegime
You add up the same numbers, you get the same result no matter which order you add them up in, they are the same numbers, 1+4 = 5, 4+1 still = 5, with that if you divide A 4240 by B 3994 you get 1.061, if you turn that around and divide B 3994 by A 4240 you get 0.941
So the first way is A (5800X3D) vs B (12900K)
The second way is B (12900K) vs A (5800X3D)
So the first way is A (5800X3D) vs B (12900K)
The second way is B (12900K) vs A (5800X3D)
Associate
- Joined
- 7 Nov 2017
- Posts
- 1,923
your method is scientifically flawed is what we are getting at. It doesn't really matter as HUB methods are flawed as well. Only fan boys care about it in the long run.
Caporegime
"Fanboy" right, now i get why you still haven't dropped this, now i get the dogged deterministic mental gymnastics to gaslight the method, what i got wrong was using the individual published slides instead of the full chart, that's a mistake i had already conceded and here you are still trying to gaslight me. Whose the fanboy?
I gave you the result using your method on post #198, to which you never responded, would you like to do that now?
Good point. I think there are a lot of positives with it as long as you don't need more cores. High minimums, low power use, one chiplet, lots of cache.
Caporegime
MLID is predicting 600 watts for the RTX 4090, 700 watts for AIB versions.
This is getting silly now....
This is getting silly now....
@CAT-THE-FIFTH A new review has just popped up for the 5800X3D from Igors lab. AMD Ryzen 7 5800X3D - the only (and last) fighter of its kind as a perfect and very efficient upgrade | igor'sLAB (igorslab.de)
It shows the 5800X3D is the fastest CPU tested, while using the least power of every CPU tested.
Incredible on the power efficiency side - imagine a 'clocks per watt' between the two.
Again...
1fps => 2fps is either +1fps or 100% increase, depending on how you want to look at it.
Do this for 5 games
+5fps or +500% performance, over 5 games is +1 fps per game or +100% per game either way - both ways equate to the same answer.
Once you add in different values, the 2 ways of adding them up give different results.
1fps => 2fps for 5 games AND 100fps => 110 fps for a 6th game comes to either;
(Your way) (2 + 2 + 2 + 2 + 2 + 110) / (1 + 1 + 1 + 1 + 1 + 100) = 1.14 which you'd say is a 14% increase. This is considered the 'wrong' way in statistics because you are heavily weighting the larger fps game which had a smaller relative increase.
This could also be worked out based on the % increases:
(Other way) (100% + 100% + 100% + 100% + 100% + 10%) / 6 games = 85% increase on average
Do you now understand the difference? I know you don't accept the latter for being the correct way, but can you at least accept there is a different way of calculating it?
Caporegime
@Miganto
Repeating the same thing over and over again for the sake of attrition isn't going to change the result, can we just drop this now ^^^ ?
Its the way Intel pushed the power levels to frankly stupid levels just to get another 2% out of the 12900K, its 80% higher, Its looking like Nvidia are going to be doing the same, at what point is this sort of behaviour going too far, once these components heat up the planet by how much infrared radiation they pump out?
Repeating the same thing over and over again for the sake of attrition isn't going to change the result, can we just drop this now ^^^ ?
Its the way Intel pushed the power levels to frankly stupid levels just to get another 2% out of the 12900K, its 80% higher, Its looking like Nvidia are going to be doing the same, at what point is this sort of behaviour going too far, once these components heat up the planet by how much infrared radiation they pump out?
Caporegime
Everyone will have their own mini power station in a couple more generations
My EVGA 650 GS is 6 years old and trembling in fear.
Associate
- Joined
- 7 Nov 2017
- Posts
- 1,923
I have no idea what you are on about.
It makes a difference.
Card A vs Card B with Card A as the benchmark.
Game 1 Card A 100 FPS, Card B 50 FPS, 100*(A/B) = 200
Game 2 Card A 50FPS, Card B 100 FPS, 100*(A/B) = 50
Arithmetic mean is (200 + 50) / 2 = 125 or Card A is 25% faster than card B. This is obviously incorrect.
Geomean is a lot more complicated but just do it in excel and you get a result of 100 or Card A = Card B which is correct.
Any method that does not use a GEOMEAN is flawed.
Any method that sums FPS is weighting high FPS more than low FPS which is the inverse of what you should weight if you are going to apply a weighting at all.
Best method is to take the % difference in each game for the benchmark vs each comparitor. You then take that list of % differences and geomean the result for each comparison. That gives you your ranking. This is the method TPU and HUB both use because it is the correct method if you want an even weighting to all results.
You could argue a method that weighs wins at low FPS more than wins at high FPS (so winning by 100% when the benchmark is getting 30FPS means more than winning by 100% when the benchmark is getting 300FPS) would be even better because it tells you if a part helps more where needed than just inflating the result of games that already run well beyond the max refresh rate of any screen anyway but that is a slightly different topic.