6850k Broadwell-E 6850K benchmarks leaked

[Curufinwe_wins]

1 minute ago, Lurick said:

 

Yah that's what I'm thinking, it's an EVGA board and it doesn't look like they even released a BIOS for the new CPUs yet.

Well as uncomfortable as it may seem to be, if you post a thread on LTT about it when push comes to shove, we can see if someone is near where you live and help you out.

 

As seen in my description, I live in Madison, WI, so anyone within say a 90 minute drive or so I could help out with this problem.

[Technous285]

Ahh flipping heck, this sorta thing pops up almost right after I order the parts for my upcoming Monsuta Rig (5820K/MSI X99A SLI Krait/32GB (4x8GB) Corsair Dominator Platinum/980Ti). Almost as if Intel ~AND~ Nvidia are trolling me for being a poor Aussie who dumped like half a year's worth of saved cash (~$3400AUD + S&H) on parts for a new system!

Of course, it's unlikely that if/when the 6850K launches here in Aus it'll be around the same price range of $560-600 AUD as the 5820K currently is, so I'll be probably good with the 5820K for at least 3-4 years.

[done12many2]

1 minute ago, Technous285 said:

Ahh flipping heck, this sorta thing pops up almost right after I order the parts for my upcoming Monsuta Rig (5820K/MSI X99A SLI Krait/32GB (4x8GB) Corsair Dominator Platinum/980Ti). Almost as if Intel ~AND~ Nvidia are trolling me for being a poor Aussie who dumped like half a year's worth of saved cash (~$3400AUD + S&H) on parts for a new system!

Of course, it's unlikely that if/when the 6850K launches here in Aus it'll be around the same price range of $560-600 AUD as the 5820K currently is, so I'll be probably good with the 5820K for at least 3-4 years.

You're fine and that's going to be a very strong build.  Enjoy it.  The 5820k is no slouch. 

[done12many2]

7 minutes ago, Curufinwe_wins said:

Well as uncomfortable as it may seem to be, if you post a thread on LTT about it when push comes to shove, we can see if someone is near where you live and help you out.

 

As seen in my description, I live in Madison, WI, so anyone within say a 90 minute drive or so I could help out with this problem.

Broadwell-E BIOS update for the x99 Deluxe was pushed out weeks ago.  Did you get one for your Sabertooth? 

[79wjd]

Why is their such a large boost from Haswell-E? Consumer Broadwell wasn't this much of a bump over consumer Haswell. 

[Cheddle]

1 hour ago, Lurick said:

I got an interesting scenario, lets say someone picked up an x99 board but no CPU as they are now just waiting on Broadwell-E instead to be released. Since a BIOS update is needed for Broadwell-E to work how much of a bind do you think they might be in?

 

Some ASUS boards can BIOs update with no RAM or CPU installed.

[Cheddle]

2 minutes ago, djdwosk97 said:

Why is their such a large boost from Haswell-E? Consumer Broadwell wasn't this much of a bump over consumer Haswell. 

consumer broad well i7's were gimped. 6mb of L3 cache vs 8mb of L3 in Sandy, Ivy, Haswell, Skylake... 

 

As I mentioned before the 5775c is basically an i5 with hyper threading - where as normally an i7 also usually has more L3 cache than the i5

[done12many2]

1 minute ago, djdwosk97 said:

Why is their such a large boost from Haswell-E? Consumer Broadwell wasn't this much of a bump over consumer Haswell. 

I don't know if many of you caught the podcast where the Intel guys were getting overly excited about sharing the Broadwell-E news, but those guys definitely gave you the impression that this very thing was going to happen.

 

Meanwhile everyone to include myself had it at 5% at best. 

[Technous285]

2 minutes ago, done12many2 said:

You're fine and that's going to be a very strong build.  Enjoy it.  The 5820k is no slouch. 

I know, I know.

The biggest fault I probably made with my selection of parts, it's likely going for a 850W PSU (though it IS a 80+ Platinum with 94% efficiency at 50% load) and cutting it close for power when estimated power consumption for my end-result rig (64GB (8x8GB) Corsair Dominator Platinum, 2-way SLI 980TI's being the main changes from the ordered parts so far) is going to be around 770W. And that's before adding in the juice to be drawn by the custom LC loop I'm going to put the 980Ti's under, or the juice they'll pull when OC'd, though that shouldn't add much more power draw either way.

[done12many2]

2 minutes ago, Technous285 said:

I know, I know.

The biggest fault I probably made with my selection of parts, it's likely going for a 850W PSU (though it IS a 80+ Platinum with 94% efficiency at 50% load) and cutting it close for power when estimated power consumption for my end-result rig (64GB (8x8GB) Corsair Dominator Platinum, 2-way SLI 980TI's being the main changes from the ordered parts so far) is going to be around 770W. And that's before adding in the juice to be drawn by the custom LC loop I'm going to put the 980Ti's under, or the juice they'll pull when OC'd, though that shouldn't add much more power draw either way.

Well I don't know if this helps or hurts your cause, but my setup in my signature pulled a peak of 1075 before I added another pump and some additional fans so I'd call it 1100.  That was at the wall and was with a 5960x at some stout voltage.  Maybe that can help with future planning. 

[Curufinwe_wins]

15 minutes ago, done12many2 said:

Broadwell-E BIOS update for the x99 Deluxe was pushed out weeks ago.  Did you get one for your Sabertooth? 

No. I don't update bios's on operational systems unless absolutely necessary. I am running the version from last June.

Just now, done12many2 said:

Well I don't know if this helps or hurts your cause, but my setup in my signature pulled a peak of 1075 before I added another pump and some additional fans so I'd call it 1100.  That was at the wall and was with a 5960x at some stout voltage.  Maybe that can help with future planning. 

I'm running a similar system (all moderate-high overclocked naturally) with a 5820k on a 850W power supply without issues, although I have had overpower shutdowns when I tried F@H on all my components at once.

[Technous285]

1 minute ago, done12many2 said:

Well I don't know if this helps or hurts your cause, but my setup in my signature pulled a peak of 1075 before I added another pump and some additional fans so I'd call it 1100.  That was at the wall and was with a 5960x at some stout voltage.  Maybe that can help with future planning. 

If the worst comes to the worst, I can grab something like the Phanteks EATX (Extended ATX) Power Supply PH-PWCOB_2P1M (thanks Kyle/Awesomesauce Network) with the appropriate extension cables and install my second Antec HCP 850W Platinum (Beast Rig's PSU) in the Thermaltake Core X71.

Just means I'll have a system sitting Judas without a PSU if I have to make Monsuta a dual-PSU rig, oh well!

[DXMember]

8 hours ago, NumLock21 said:

That's the price of a 5820K, excluding MicroCenter. Until actual msrp comes out, I'll take wccf's price with a grain of salt. IMO the 6850K's should be put against the 5930K, since they're both at the same price range.

umm... I mean for the people that already own a 5820K or something... it's not worth a $400 upgrade

[DXMember]

8 hours ago, done12many2 said:

Break it down and you can see that new larger heat spreader doing work.  That's a hell of a difference between minimum, average and peak temps between the two.

 

All that while producing more.

 

 

 

heat spreader might be improving the temps but it's not made of magic, the broadwell chip also runs on lower voltage and is a smaller process node, and given that they have comparable transistor count, they're spread further apart which helps with thermals

[FloRolf]

Damm! 

Especially the temp improvements sound really good for ITX peeps like me. But I hope there will be more ITX boards available after launch. 

So this, or a used 5820k later? 

[done12many2]

7 hours ago, DXMember said:

heat spreader might be improving the temps but it's not made of magic, the broadwell chip also runs on lower voltage and is a smaller process node, and given that they have comparable transistor count, they're spread further apart which helps with thermals

 

That's interesting.  So your saying that due to the shrink, Intel was able to space/spread the cores further apart?  I'm looking forward to learning more about that.

[79wjd]

16 minutes ago, done12many2 said:

 

That's interesting.  So your saying that due to the shrink, Intel was able to space/spread the cores further apart?  I'm looking forward to learning more about that.

Or just make the die smaller (which is what they normally do -- although the transistor spacing may not be the same).

[Curufinwe_wins]

11 hours ago, djdwosk97 said:

Or just make the die smaller (which is what they normally do -- although the transistor spacing may not be the same).

Making the die smaller at the same power consumption makes it harder to cool. required thermal disipation density increases. Fick's law

 

 

But lower voltages, better metal ihs (thermal diffusivity increase), moving more components to the back of the die... 

 

Any and all of the above could have happened here to improve things.

[done12many2]

12 minutes ago, Curufinwe_wins said:

Making the die smaller at the same power consumption makes it harder to cool. required thermal disipation density increases. Fick's law

 

 

But lower voltages, better metal ihs (thermal diffusivity increase), moving more components to the back of the die... 

 

Any and all of the above could have happened here to improve things.

 A little more intel.  Pun intended. 

[done12many2]

http://www.overclock.net/t/1599263/6950x-benchmarked

 

Oops.  Forgot the link. 

[Curufinwe_wins]

6 minutes ago, done12many2 said:

 A little more intel.  Pun intended. 

I made a fuller explanation on this thread. I really don't feel like going over the 3d differential equations, although relatively speaking they are pretty straight forward.

 

 

[done12many2]

11 minutes ago, Curufinwe_wins said:

I made a fuller explanation on this thread. I really don't feel like going over the 3d differential equations, although relatively speaking they are pretty straight forward.

 

 

Thanks, but I was actually trying to share more intel.  I just forgot the link that's in the post following my error. 

 

Great read though.  You're one of them smart guys! 

[Curufinwe_wins]

17 minutes ago, done12many2 said:

Thanks, but I was actually trying to share more intel.  I just forgot the link that's in the post following my error. 

 

Great read though.  You're one of them smart guys! 

oh. oops. Well smart is relative...

[patrickjp93]

On May 4, 2559 BE at 8:34 PM, djdwosk97 said:

Why is their such a large boost from Haswell-E? Consumer Broadwell wasn't this much of a bump over consumer Haswell. 

It's all about the cache. Not only did the 5775 C have a cut-down L3 cache to start with, but another chunk of the 6MB was used for tag bits to map data in L4. In other words, the L3 cache of Broadwell-C was kneecapped enough to negatively impact performance for Cinebench. Further, cache misses at L4 in Broadwell C are much more expensive than L3 misses on Broadwell E. That's why every discussion of CPU performance programming must start at the cache line level.

[Prysin]

9 minutes ago, patrickjp93 said:

It's all about the cache. Not only did the 5775 C have a cut-down L3 cache to start with, but another chunk of the 6MB was used for tag bits to map data in L4. In other words, the L3 cache of Broadwell-C was kneecapped enough to negatively impact performance for Cinebench. Further, cache misses at L4 in Broadwell C are much more expensive than L3 misses on Broadwell E. That's why every discussion of CPU performance programming must start at the cache line level.

well, Broadwell C L4 was a victim cache, which added uneccessary cycles in CPU workloads, but was excellent for iGPU workloads (which re-reads data lots of times).

 

GT4e is rumored to have a proper, non-victim L4.