AMD Ryzen 9 3950X beats Intel i9-10980XE in 3DMark Firestrike by almost 25%

We should put this in the Lost Order thread as well.

 

I can't speak for others but i don't think it makes Intel greedy. The point of all business is not make money. AMD is no different. But i do think Intel over the years became over confident and took their fans for granted. Lack of competition definitely influenced some of their decisions (potentially : high price, no HT on core i5, staying on 4 cores for so long, new socket about every new generation, ...). You can't really blame people for hating that and wanting to send a message with their pocket. AMD would do the same as Intel in the same position. But they are not in the same position ... Competition is good. It drives companies to improve. Without it it's all about making more money..

 

@blkspade
except it allows to compare systems by hardware and able to not have variations just because the scene (game) changed a bit.
might not mirror real world use, but is nice if i want an idea of the difference in perf i will get between different chips/arcs etc.

@nevcairiel
i tend to call companies greedy when they add features like HT to i3 (i5?) without any redesign,
meaning they had the capability already, but were not giving it to customers all this time,
until amd started doing it.

 

people been quite OOT
not to drag it more but regarding socket (chipset)
why u think intel get all the profit ? do u ever think/consider hardware/mobo-maker ?
in selling such mobo, the one that getting more profit is mobo-maker, no? cmiiw

partnership in business can means alot in supporting each other... for intel, it can mean those company preferable using intel-product more rather than other-else
for intel, it will be more profitable going with same chipset/socket as long they can sell more of cpu + chipset

 

The headline says the cpu is 25% faster, period. That is false and misleading bs and you know it. They did it that way so that the vast majority of people who just look at headlines will start repeating this false information. And web search engines will find that headline and repeat it. If they were being honest they never would have put that as the title of the article.

The title of the article should have been "2080Ti beats 2070 in 3dmark firestrike by almost 25%"

 

So amd had 2080ti vs intel 2070?

Guru your so much better then this..

 

a 2080ti is roughly 35% faster than a 2070 fyi

and as I just found out by swapping G.Skill for "good looking" rgb corsair vengeance ram all those benchmarks heavily rely on memory
cinebench r15 g.skill 3600mhz 2137-218
cinebench r15 corsair 3400mhz 1919-215
cinebench r20 g.skill 3600mhz 5217-525
cinebench r20 corsair 3400mhz 4625-511
firestrike corsair 3400mhz 20758 physics
firestrike g.skill 3600mhz 25196 physics
timespy corsair 3400mhz 9963 physics
timespy g.skill 3600mhz 12119 physics
even without the gpu thing considering the +18% I get with +200Mhz with a 534Mhz delta it would be an unfair comparison

edit : on a 9900k btw
and something else I found out, if you have rgb memory or (corsair) aio rgb block don't run them at max led brightness or better kill the rgb when you benchmark, my AIO coolant was +3°C at idle ! and the memory sticks +7°C only because of the leds ...what ? omg...
Edit, problem found :
hey guys and thanks everyone for confirming that my scores were broken
after a few hours I found the culprit, it was my motherboard apparently my 1st slot closest to the cpu has a defect
with 4 sticks in I only had 1 test "failing" Aida64 memory and cache bench : L2 Cache write was around 430Gb/s when with the G.Skill I had like 650Gb/s
after trying all the sticks and slots I found that the 1st MB slots closest to the cpu has some defect but not enough to completely break the system...weirdly G.Skill just "brute forces" through it
if I socket only a single stick in the 1st slot both brands and all 8 sticks fail...
really weird problem I actually played encoded videos benchmarked and streamed like this through the year
amazing how modern systems are tough ! from what I remember in the old days they just refused to boot with bad memory

 

Cinebench doesn't depend at all on memory speed (differences within margin of error)...
Firestrike depends -a bit-, but not much...
TimeSpy also doesn't depend on memory speed...

... unless severely choking the CPU with extremely slow memory or single channel.

Those results do not make any sense.
You must have done something wrong when installing the Corsair memory (wrong slots probably so you're running in single-channel mode).

Check CPU-z and do the Aida64 memory/cache test

 

We are talking about Physics benchmark inside firestrike.

This test heavely depend on memory speed and tweaks. That's why we are overclocking memory and tweaking 2. And 3. Timings for this test alone

 

firestrike corsair 3400mhz 20758 physics
firestrike g.skill 3600mhz 25196 physics
timespy corsair 3400mhz 9963 physics
timespy g.skill 3600mhz 12119 physics

Those results seems to not make sense.
20% increase adding 200mhz to memory?
Even if the test was heavily relying on memory bandwidth, 100% capped from memor bandwidth, 200 mhz on 3400 is around 6%.
It should not give you a 20% speed increase.
Unless you went in single channel mode or you set timing super strict on 3600 and super relaxed on 3400.

 

thx for the comments, Nizzen I'll try to tweak the timings Wavetrex I'll do the cpu-z thing

sadly no there's no mistake on my part, that monster difference is just from the ram itself
I have 4x8Gbs each set (on 4 possible slots) and both are on QVL from my motherboard both using XMP profile
the only thing I see is that the Corsair ram is defective in some way but well since I can run benchmarks and aida64 stress test fine RMAing would do nothing it "works"

reassuring that it's not normal because it would mean those benchmarks are completely useless

 

Show us you're 3dmark firestrike physics score

 

Pls read again the topic we are talking about physics score between 3950x mainstream cpu 16 core vs i9 10980XE HEDT 18 core the gpu has nothing to do with physics score but the memory does dual vs quad do you have enything to say about that?And how effecting the overall score the bigger memory bandwidth?.For many years Intel was the undefeated champion well no more dear friend times have changed.Sorry for grammar English isn't my native language. And of course ty Guru3d for your Profesional work all this years καλησπέρα από Ελλάδα
Haters gonna hate.

 

@MaxBlade
how about the fact that running a 2080ti definitely puts the bottleneck towards the cpu,
or the fact that the intel is quad channel?

 

It's a PURE cpu/memory test. Latency and bandwidth is the key for the memory + high frequency on all cpucores.

 

Remember, remember the fifth of November,
Intel Inside treason and plot.
We see no reason
Why Intel Inside treason
Should ever be forgot!

 

Again, as other stated in the thread: CPU score is what is compared. Though not easily compared anymore, since intel bench is gone.

 

hey guys and thanks everyone for confirming that my scores were broken
after a few hours I found the culprit, it was my motherboard apparently my 1st slot closest to the cpu has a defect
with 4 sticks in I only had 1 test "failing" Aida64 memory and cache bench : L2 Cache write was around 430Gb/s when with the G.Skill I had like 650Gb/s
after trying all the sticks and slots I found that the 1st MB slots closest to the cpu has some defect but not enough to completely break the system...weirdly G.Skill just "brute forces" through it
if I socket only a single stick in the 1st slot both brands and all 8 sticks fail...
really weird problem I actually played encoded videos benchmarked and streamed like this through the year
amazing how modern systems are tough ! from what I remember in the old days they just refused to boot with bad memory

 

Again, you can pretend that the cpu score is what this is about. It's not. Read again, I will quote:

"
AMD Ryzen 9 3950X beats Intel i9-10980XE in 3DMark Firestrike by almost 25%"


That is pure bs headline that is 100% false and cannot be defended. It didn't beat the intel, the 2080ti beat the 2070.

If they were being honest and if you were being objective the headline would have read:

"AMD Ryzen 9 3950x TIES intel i9-10980XE"

 

Oh...this is bad.




Searching for i9-10980XE score...