1950X is embarrassingly slow in 3D Mark

>>61752495
I'd say only in Timespy it's kinda underwhelming

>>61752528
100% more coarz compared to r7, only 15% better CPU benchmark results

>>61752495
>glued together

>>61752495
>>61752528
Yeah seems like new hardware optimization issue, as usual not everything is going to be perfect.

>>61752554
should tell you something about software at that point

>>61752573
Can you somehow explain superior 7900x score then?

>>61752585
>different programs work better on [x]

Wow, this needs further investigation!

File: 1500085542560.jpg (663KB, 5000x2813px) Image search: [Google]

663KB, 5000x2813px

>>61752554
Most likely just down to inter die and CCX latency per usual.
Does anyone know of a benchmark that compares a 1700/1700x/1800x with a single CCX and SMT enabled(so 4+0 in bios) vs the 1500x? I'm curious to see exactly how much better the arch would perform if they could get IF latency down.
Really the modularity of zen is incredible and caused nvidia to nearly immediately publish that whitepaper about modular GPUs(even though we've known navi would be modular for ever) and I can't wait to see how they get the IF speed up or latency down in future revisions.

>>61752585
Look at the video for yourself. https://www.youtube.com/watch?v=decz1N9YpOw

>>61752605
It is important to note that 7700K was curbstomped by 1800x, yet 1950x still lost to 7900x. Something strange is going on here.

>>61752619
What's cool about Zen is that you can use something like Process Lasso to pin a program to specific die/ccx with or without SMT, and get better performance just by optimizing this.

>>61752627
Yes, 3DMark whateverspy is a bad CPU benchmark. This shouldn't even need explanation but apparently OP (who is a faggot) decided to make this thread to bait people into replying.

File: 1491010800039.jpg (20KB, 420x418px) Image search: [Google]

20KB, 420x418px

>>61752495
Pic is from : https://www.youtube.com/watch?v=decz1N9YpOw

> The system was running a pre-production Threadripper processor that doesn't perform like the production CPUs. Confirmed here :
https://www.youtube.com/watch?v=JjVP76zz8ZU

Furthermore, The threadripper system was running on DDR @2666 while the intel was running @3200. Confirmed here :
https://youtu.be/decz1N9YpOw @ 4:00min in.
The ram alone accounts for the difference+some. The pre-production processor difference = blow out.

Glad I could provide source sauce to your otherwise pointless shill post

File: 1494387422977.jpg (181KB, 888x1120px) Image search: [Google]

181KB, 888x1120px

>>61752557
>>61752585
>>61752605
See :
>>61752679

The shills are busy today.

>>61752679
Yea but shouldn't 16 cores win 10 cores even at a tad slower RAM? I mean 16 cores is still 16 cores provided timespy is able to scale that well.

>>61752495
>Forgetti professional work. Muh gayms

>>61752706
>Alienware
>Professional work

>>61752679
delet