First chess benchmarks for the Intel core i9-10980XE emerge

[Zenmastur]

Roughly speaking how many times is i9-10980XE faster than i7-5960X ?

Hard to say with out more benchmarks but less than twice. Maybe 70% faster...

[Joost Buijs]

Why would you use the BMI2 scores for Ryzen when it scores significantly better on popcnt? Aren't you just shooting yourself in the foot? Maybe your doing that on purpose!

You can do really nice things with BMI2 especially PEXT/PDEP but if you are not a programmer it probably doesn't matter.

Benchmarks for the 3950X are out. They have been out since June. I guess you conveniently missed all those leaks. Either that or your not looking for them. Most of them that show comparisons between the two show that the 3950X is faster by over 20%.

I've been looking at all the leaks and most of the benchmarks were synthetic like 3DMark and Cinebench. They were done with totally different GPU's, so you can't really compare things this way.

Fourth, If you buy a 10980XE you will be locked into a 3-year old chipset and MB design with no upgrade path.

I already have two of these X299 mainboards with 3 year old chipsets and outdated design with no upgrade path, for me it is just a matter of swapping CPU's, and maybe adding more and faster memory.

[Joost Buijs]

Roughly speaking how many times is i9-10980XE faster than i7-5960X ?

In some other thread we talked about the i7-5960X performance before. I also have a 5960X at 4 GHz. on an Asus X99 deluxe, it does on the stockfish bench a little bit below 20 mnps, if the leaked benchmark of the 10980XE is true and it really does 60 mnps it is 3 times as fast. I guess we have to wait until it is actually released to get an impression of the true performance, often these leaks are just fake.

[Joost Buijs]

Roughly speaking how many times is i9-10980XE faster than i7-5960X ?

Hard to say with out more benchmarks but less than twice. Maybe 70% faster...

You must be joking.

[Zenmastur]

Why would you use the BMI2 scores for Ryzen when it scores significantly better on popcnt? Aren't you just shooting yourself in the foot? Maybe your doing that on purpose!

You can do really nice things with BMI2 especially PEXT/PDEP but if you are not a programmer it probably doesn't matter.

Benchmarks for the 3950X are out. They have been out since June. I guess you conveniently missed all those leaks. Either that or your not looking for them. Most of them that show comparisons between the two show that the 3950X is faster by over 20%.

I've been looking at all the leaks and most of the benchmarks were synthetic like 3DMark and Cinebench. They were done with totally different GPU's, so you can't really compare things this way.

How does GPU performance affect CPU performance? I thought it was the other way around. So, I don't understand this comment.

Fourth, If you buy a 10980XE you will be locked into a 3-year old chipset and MB design with no upgrade path.

I already have two of these X299 mainboards with 3 year old chipsets and outdated design with no upgrade path, for me it is just a matter of swapping CPU's, and maybe adding more and faster memory.

Then it may makes sense for you to perform a drop in upgrade assuming your MB vendors offer an upgraded bios for your particular motherboards, and, assuming that the new bios doesn't gimp the CPU too much. Or else a new MB will be in order.

It certainly changes the equation for anyone that doesn't happen to own X299 motherboards as they will have to buy a dead end motherboard to support these CPU's.

Roughly speaking how many times is i9-10980XE faster than i7-5960X ?

In some other thread we talked about the i7-5960X performance before. I also have a 5960X at 4 GHz. on an Asus X99 deluxe, it does on the stockfish bench a little bit below 20 mnps, if the leaked benchmark of the 10980XE is true and it really does 60 mnps it is 3 times as fast. I guess we have to wait until it is actually released to get an impression of the true performance, often these leaks are just fake.

While the leaked benchmarks will not be perfect they should be close enough for most purposes.

The numbers I gave are for a general purpose benchmark. i.e. non-chess specific. The numbers given for the 9980XE were used as surrogates for the 10980XE since comparisons between the two show the 10980XE is slightly slower. The numbers where 29494 for the 9980XE vice 15993 for the 5960X. That's what I based my statements on.

[DustyMonkey]

You can do really nice things with BMI2 especially PEXT/PDEP but if you are not a programmer it probably doesn't matter.

So the point flew right over your head?

Use the fastest executable for the platform, or you are a wanker with an agenda.

[Joost Buijs]

How does GPU performance affect CPU performance? I thought it was the other way around. So, I don't understand this comment.

Cinebench and 3DMark are 3D graphical benchmarks, don't you think that the GPU has influence on that as well?

The numbers I gave are for a general purpose benchmark. i.e. non-chess specific. The numbers given for the 9980XE were used as surrogates for the 10980XE since comparisons between the two show the 10980XE is slightly slower. The numbers where 29494 for the 9980XE vice 15993 for the 5960X. That's what I based my statements on.

I don't know what you mean by 'numbers' but the 9980XE is at least 2 times as fast as the 5960X.

Like I already said before we have to wait until both the AMD 3950X and the Intel 10980XE are truly available to make a real comparison, numbers based on leaks with 'engineering samples' are not very reliable to say the least.

And now I will stop with this endless discussion.

[Leo]

Techpowerup quote:

"18 cores, the i9-10980XE is ahead of the 12-core Ryzen 9 3900X in rendering and simulation tests, although not by much (for a chip that has 50% more cores). This is probably attributed to the competing AMD chip being able to sustain higher all-core boost clock speeds. In tests that not only scale with cores, but are also hungry for memory bandwidth, such as 7-zip and Media, Intel extends its lead thanks to its quad-channel memory interface that's able to feed its cores with datasets faster."

[Zenmastur]

How does GPU performance affect CPU performance? I thought it was the other way around. So, I don't understand this comment.

Cinebench and 3DMark are 3D graphical benchmarks, don't you think that the GPU has influence on that as well?

“Graphical benchmarks”, to use your words, are meant to measure graphics card performance not CPU performance. You posted a link to a CPU benchmark. The two are even remotely similar. And to answer your question, NO, I don't think which GPU is used will have any affect on the CPU performance if the test is done properly.

The numbers I gave are for a general purpose benchmark. i.e. non-chess specific. The numbers given for the 9980XE were used as surrogates for the 10980XE since comparisons between the two show the 10980XE is slightly slower. The numbers where 29494 for the 9980XE vice 15993 for the 5960X. That's what I based my statements on.

I don't know what you mean by 'numbers' but the 9980XE is at least 2 times as fast as the 5960X.

I guess that depends on the application, which is exactly why they bother doing benchmarks in the first place.

Like I already said before we have to wait until both the AMD 3950X and the Intel 10980XE are truly available to make a real comparison, numbers based on leaks with 'engineering samples' are not very reliable to say the least.

You were the one that posted the leaked benchmark. You certainly seemed to think it had some significance or why bother posting it. As soon as I posted a comparison benchmark they both suddenly lost their significance. What do you suppose caused that?

And now I will stop with this endless discussion.

As far as endless discussions go this seems to be a rather short one! But, I agree that there isn't much to be had by continuing this conversation.

Regards,

Zenmastur

[smatovic]

Hehe, looks like AMD is really back, people are starting good old flame wars again

--
Srdja