X99 Platform upgrade path

Lepton87

Platinum Member
Jul 28, 2009
2,544
9
81
I bought the mobo that I liked the most when I was faced with the motherboard choice for the i7 5820K as that CPU seemed and still seems like the best value from all the Intel's HEDT's CPUs.
What's more its value is not even that far of from the 4790K and maybe even matching it if MT performance is the only criterion.
That is with taking into account the whole platform cost along with the RAM cost. If the DDR4 prices would have fallen down then that would have made the HEDT platform a much more attractive buy than it has been since its introduction. There seems to be a downward trend in the DDR3 memory prices but it's making the price premium of the faster memory eroding very fast, which is quite understandable as the faster memory don't provide much in the way of boosting performance, that and maybe the technology is getting more mature and the faster RAM speeds are getting more common, soon DDR4 over the speed of 2800MHz will become a dime a dozen and to think I payed so much for my 2666MHz modules for a nice premium over the regular 2133MHZ but that was over six months ago but still I got lot more value and possibly even more oomph from the additional money I spent for the faster memory than if I would have gone with the 5930K instead of the 5820K.

And now that would have been a monumental waste of money and it still is because the 5930K costs a 600$ and what it offers for that money over the 5820K? Pretty much nothing, 3.5 GHz stock clock
Max Turbo Frequency of 3.7 GHz, compared to 3.3/3.6GHz that 5820K does. So that is a 5.7% frequency uplift in base frequency and a "whooping" 2.7% frequency increase in the turbo frequency. And all of that for a very "modest" increase in the price only up from the BOX price of $396.00 to the BOX price of $594.00. That's a 50% increase in the price tag for the indistinguishable performance uplift.

The only saving grace of the 5930 is its full 40 lanes of PCI-E express connectivity but I'm not convinced it justifies this huge price discrepancy because it only becomes very useful when you are dealing with more than 3 graphics cards and somewhat useful with 3 cards. Even three cards are mostly fine with a 28 PCI-E express lines but that does not mean that there is no benefit at all to 40 lines for three cards but that benefit is small enough not to warrant that kind of expenditure. Such miniscule difference in performance don't
don't make any noticeable differences when actually playing games so the main purpose for it is beating the records. Showing and flashing the e-peen is always a huge turn-on for lots of enthusiasts which I consider a good thing at least up up until the point the participants in the conversation resort to naming call, twisting facts and other despicable practices I also hate how much of an impact good marketing can have on people, some are preparing to protect and try to shrug off any of the wrong-ding their favourite company did, even if it was a blatant fraud or some other despicable behavior. It is the kind of behavior I would expect lovers to engage in and not someone who doesn;t even work for a given corporation and that corporation doesn't even give a damn about its fans. This is mostly curious. What primal instincts come into play in such interactions? Of course there are exceptions.
I'd say that 5820K is already an attractive buy as it's always have been since Intel launched it. When I decided to buy my motherboard I disregarded the cost of it because it was suppose to last two CPU generation.

Once you take into account that such a high-end motherboard can last for two CPU generations the cost gets split over into two CPUs from different generations. So that allows to spread the cost of the motherboard into to CPUs generations and after accounting for that it's not that expensive after-all. But there's a thing that I think just might put a wrench in all of that my plans. That is the lack of a new CPU that would be a worthy successor that would make sense to buy it and replace the
CPU that I currently have, If things are to remain on their not so favorable course BW-E/EX might end up severely late to the market maybe so much that the prospective buyers would just be better off waiting a few months (3-4) and just skip that delayed CPU and went straight to the Skylake.

Some people say that the delay of Broadway is not going to impact the release schedule of the SKY LAKE CPUs at all, somehow I'm very skeptical about this but I'm not saying it's impossible, but I'm not competent and knowledgeable enough to say either way for certain. If that's the case then maybe BW-E is just not worth it at all, don't get me wrong, not worth it at all but only when you already have HW-E and not not worth it at all in a vacuum. And even that is a bit too harsh, not worth it is a bit too much, just not very cost effective, that's it. Which is quite sad to me because it was always in my upgrade plans.

Here are some of my phantasies: A non-extreme 8C BW-E SKU that would overclock to 4.8GHz for 650$ or maybe even something cheaper but also very compelling to me and I imagine to a lot of enthusiasts: 6C/12T overclock-able to 5GHz-5.2GHz with custom water such as mine or 4.8GHz on good air like Noctua NH D15, that would be great if the price was about 500$.
Well, I guess that 8 core is out of the question for that price range because it competes with Extreme Edition CPU. I can only imagine that if the Extreme CPU gets more cores like 10 or 12. Still that 6 cores version would be great, but is the process really up for this? That is really clock that much higher than 22nm?



There's still hope, but this is just wishful thinking but the 8C 40 PCI-E 3.0 LANES BW-K that would OC to around 4.6GHz would be great, alternatively 6C/12T overclock-able to 4.9GHz for 500$ (overclock under custom water similar to mine) would be great and I wouldn't even think about it, I would just buy it, but both of those are wishful thinking and it's not gonna happen. What do we know about HEDT Broadwell-E/EP? Do we if it's going to offer any uplift in frequencies and most importantly how high frequencies can it reached while overclocked? I'd be interested in both overclocking using the air cooling as well as the best WC custom loops around.

ps. sorry I didn't look at my post's length until I posted it, it really came up long... Good that at the very least I used some paragraphs, I've been reprimanded on this forum for using a wall of text before, good I actually took that to heart. It would have been neigh to impossible to read that without any paragraphs.
 
Last edited:

escrow4

Diamond Member
Feb 4, 2013
3,339
122
106
So what are you asking? Broadwell-E will be a small clock bump and that is about it, just like 3930K to 4930K - 3.2GHz | 3.8GHz to 3.4GHz | 3.9GHz. Sure some new extensions and a node shrink but meh for gaming. I bought a 5930K because I could, the price was irrelevant, and I'd prefer to a have a full CPU with full PCI-E connectivity not a gimped 5820K. I'd take 6 cores anyday now for gaming over 4.
 

Lepton87

Platinum Member
Jul 28, 2009
2,544
9
81
I guess it was obvious from the post, basically two things at first but I whittled it down to one.

1. Clock speed potential. There were many posters that sympathize with Intel who claimed that the situation with the transition form 32nm to 22nm was a one time reduction in top clocks achievable and from now on it should return to normal with every die shrink providing a significant boost in clocking headroom. Anyone remember the transition from 180nm to 130nm? It was really awesome, the clock speed was basically doubled. Contratry to what Intel enthusiasts say I don't believe clock speeds will increase much and that explanation that 32>>22nm regressed our clocks because of switching might be true to a degree but I don't believe we will see much in terms of clock-speeds gains but I would love to be proven wrong. A lot of people is of the opinion that the regress in clock speeds are one time things because we went from planar to FinFet but somehow I doubt we will see the clock speeds ramping up aggressively. I'd be happy if we returned to our enthusiast CPUs clocking at around 5GHz quite commonly a slight improvement from SB which demanded a very good chip to reach those frequencies. What are your thoughts about the clocking potential of 14nm unlocked SKUs from Intel? 5GHz with an 7% IPC improvement for about 500$ and I might bite. But I guess that's wishful thinking. What I realistically expect is 4.6GHz with 6% IP improvement for 550$. Well, after thinking about it there's no room for a 8 core CPU under 1K USD so I'm not even speculating about it, besided I don't really need an 8C/16T CPU right now and 1K is just to much for 2 additional cores and in most probability worse clock-speed.

So what are you asking? Broadwell-E will be a small clock bump and that is about it, just like 3930K to 4930K - 3.2GHz | 3.8GHz to 3.4GHz | 3.9GHz. Sure some new extensions and a node shrink but meh for gaming. I bought a 5930K because I could, the price was irrelevant, and I'd prefer to a have a full CPU with full PCI-E connectivity not a gimped 5820K. I'd take 6 cores anyday now for gaming over 4.
You are contradicting yourself. 4930K is also a "gimped" CPU and both support 6 cores with the same amount of cache. The additional PCI-E connectivity is only useful when you use 4 graphics cards, if you do then it's a more reasonable choice than the 4820K otherwise they are identical after overclocking. I have a ASUS maximum ROG IV with NF 200 chips with PCI-E splitter and even 6990+2x6950 worked just fine and the difference between native design was purely academic a few percent at most. You know that your choice wasn't dictated by rational thought but by your psychology, right? I used to be like that when I was a little kid.
I have that worked gimped when referring to a perfectly good CPU, for all we know it the silicon was probably fine and only laser cut to meet market demands. It's such a small part of the GPU that the chance of it being broken with the rest working flawlessly are infinitesimally small. Marketing is doing a good job instilling to the gamer awareness such expressions as gimped, die harvested, a discarded die etc. It works surprisingly well even among supposedly tech savvy users. . There are 16 Cores Xeons that were originally meant to be 18 cores variants, are those CPUs also die harvested trash? Really that kind of argument is ridiculous. If anything I'd prefer to have more dark silicon to help me with the cooling requirements.
 
Last edited: