I'm talking about existing products, if you are incapable of understanding the context of the discussion that is.
Its much faster than the 6950 it replaces and it expands the portfolio. Plenty reason to replace it. I dont hink oc plays an important part as seen by the tim and why bin for a lesser margin product?
Better than you might think http://edgeup.asus.com/2016/06/17/broadwell-e-overclocking-guide/
Most workloads, even "MT," do not cause the CPU to draw 100% of TDP. Server BIOSes are typically configured to allow the CPU to run at the highest frequency that remains under TDP, which is the full turbo frequency for almost every workload. Heavy vector workloads can cause fallback to base or AVX-base frequencies, but why complain about doing 4x, 8x, etc. the work in each cycle?
That's all dandy but people who want both high clocks and more cores don't go for Xeons, especially now when BCLK overclocking is restricted. That was my point when I said in reply to the other poster that Xeons and HEDT aren't interchangeable, unless you are going to save a couple of hundred dollars looking for a particular SKU.
Where do you think HEDT CPUs come from? Intel is using the same TIM on every product in their line, including 260 W Xeon Phi parts. At 145 W, Broadwell-EP can already achieve 2.8 GHz. With OC to 260 W and some process improvements, i9-9280XE will be reaching 3.6-3.7 GHz. In fact, I see an 18C Xeon-SP SKU at 3.0 GHz base frequency and 205 W TDP, so it is pretty much confirmed. There is not even a regression in single-thread performance thanks to Turbo Boost 3.0, which makes it easy for Intel to claim 4.5 GHz turbo on every part. In fact, the more cores there are, the easier it will be for Intel to find two that will run at high frequency.
For every negative user experience you find about Ryzen, I can find a positive one. There was a post here in these forums by someone who somehow managed to kill three 6850Ks. Does that make Broadwell-E a crap product? Even someone like JayzTwoCents who used to be like the biggest Intel fanboy ever has said that he would never have got his 6950X if it was not sponsored and has been very impressed with his 1800X.
All I'm saying that given the 6950X which does 4.3GHz at 1.3V which is soldered, 4.5GHz which is the Turbo 2.0 speed of the 7900X, meaning that it will at least OC to that on all cores, won't possibly be as stable and the culprit might turn out to be the use of TIM.Where do you think HEDT CPUs come from? Intel is using the same TIM on every product in their line, including 260 W Xeon Phi parts. At 145 W, Broadwell-EP can already achieve 2.8 GHz. With OC to 260 W and some process improvements, i9-9280XE will be reaching 3.6-3.7 GHz. In fact, I see an 18C Xeon-SP SKU at 3.0 GHz base frequency and 205 W TDP, so it is pretty much confirmed. There is not even a regression in single-thread performance thanks to Turbo Boost 3.0, which makes it easy for Intel to claim 4.5 GHz turbo on every part. In fact, the more cores there are, the easier it will be for Intel to find two that will run at high frequency.
Are you just complaining that delidding will be needed to have 5 GHz on every single core? Nobody actually needs that, and the only reason to OC that far is to make some kind of statement, sort of like the folks that modify diesel engines to maximize pollution.
4.5 GHz is the Turbo 3.0 speed. Turbo 2.0 is 4.3 GHz. Again, I see no need to have 4+ GHz on every single core. It serves no useful purpose when you can just get more cores instead. The worst I forsee is that enthusiasts will have to get used to having different OC frequencies for different active core counts, just like how the stock CPU runs. Not even solder is going to make 18 cores at 5 GHz a useful operating scenario (unless you just want to roll coal).
If power is not being consumed at the CPU package, it does not matter for cooling, so why are you bringing it up? As far as your claim goes, I see power consumption at 145 W in Tom's Hardware. That is wall power, so the package power is actually below TDP. If you are going to make some inane energy efficiency claim based on overclocking, I can also point out that you can force a i7-7700K to draw over 1 kW on liquid nitrogen.
IPC has increased by as much as Broadwell-U/S to Skylake-U/S. Why would you expect anything else?
Better tell Google how bad the IPC is on their new chips. Some overclocking clown in Taiwan knows something they don't (hint: uncore)!
I'm making no claims about energy efficiency, just that there are people who do overclock and the TIM might be a problem. Why is it so hard to understand.
Pretty sure that Google doesn't pay ark prices for their chips. And Uncore does nothing for the most part except waste power on Broadwell-E.
Facebook
Twitter