X99/Haswell-E is a scam, and here's why

Well it sort of is, I assume everyone who needs a x99 (i.e. work use) already has a similar platform, and that because they need it (or are an enthusiast) that they upgrade frequently (i.e. not really more than 2 cycles), and that they probably have a 3960/4960k, there really doesnt seem to be any point in upgrading. Now if there was an 8 core 5960k at the same price that would be amazing, but there isn't.

And its not like DDR4 is even advantageous here (except in server usage), because the CAS latency is horrible (meaning it probably isn't really any faster), the price is absurd, and it isn't like the CPUs are able to use larger capacities than its predecessors, which is really disappointing. You can get DDR3 ram now anyway, that is around the same voltage and probably faster, definitely way cheaper.

With desktops people upgrade a lot less often, especially the mainstream, than do people who's livelihoods depend on these platforms, or who are enthusiasts. So its not as bad. But yes generally speaking same thing on the desktop. 

As a every day dude, i cant justify the expense on theoretical tech. Price for performance is hammered which leaves it at something to wait till the price comes down. Fools and their money are soon parted?

First, define 'scam' for me, please.  That would be like saying a Challenger Hellcat is a 'scam'.  It's powerful.  It's over the top, yes.  Does anyone NEED one? No.  But I'm drooling all the same.  Technology must advance, period.  Now it's true that the software isn't caught up with our hardware needs but that has always been true and always will.  No one writes code for hardware that doesn't exist yet.  Furthermore if I want to spend my money on the latest-and-greatest stuff I've the right to do that for bragging rights If I so choose.  Now as for your arguement that 'noone' will use this tech to the limit, I've my doubts about that as well.  Video editors need loads of bandwidth to process each and every frame as quickly as possible in order to give you top quality content on youtube sites like teksyndicate.  Edzel from LTT Uses a monster machine and needs more power to do the intensive processing for all the content that they release.  The saying is 'if you build it they will come.'  That's true of tech as well.  I doubt x99 will fail in any sense at all.  Programming is trending towards parallel processing as traditional serial processing is becoming more and more inadequate since individual clock speeds per generation have almost ground to a halt.  More cores running in parallel is the new hotness and will be for some time.  In the meantime, I've some turbochargers to look at for my Dodge.  Cheers.

I've got a few problems with this post, entirely because it leads me to believe that you are relatively new to the hardware world.

First of:

Gamers do not need them, most professionals do not need them, and your average consumer certainly doesn't need them.

Yea, I agree. However the top end $1000 CPU's are not targeted at people who need them (because they will have likely gone with a strong Xeon setup) but rather at the people who want them as they are enthusiast parts. I certainly do not need them, but it is highly likely I will get one.

AMD's Opteron offerings in the $400-$500 range will blow away all Haswell-E desktop CPU's.

spits coffee across room You seriously, like seriously, over estimate the performance of the opteron's. The exact one you are referencing, in terms of multicore performance barely exceeds the processing power AMD's own Fx-8350 and if people thought Bulldozers singlethreaded/IPC performance was poor, this was worse. In my mind it was an outdated chip when it launched especially for its target market.

Also from what benchmarks have already emerged (although synthetic benchs are not the b-end all of analysis) show that even the 5820K beats out the 4960x. With passmark being a pretty good round up overall performance of a CPU, you can see that the 4960x has around 150% the performance of the Opteron. So a chip that performs the better for the same amount of money is a no brainer.

  • DDR4

Literally useless to anyone who isn't managing a datacenter. All you get are incremental clock speed increases and lower voltages. Unless you're running an APU/GPU-on-CPU-die, RAM speed (when I say speed I mean overall speed, not the clock rate) has been completely irrelevant for a long time now. There hasn't been a real world improvement from RAM generations since DDR2.

This I kind of agree with. But lower voltages are important, for a system that maybe on 24/7 there is a difference. But as with all memory revisions (of the DDR line anyway) although response of the memory hasn't really changed (taking into the account of higher transfer but lower timings) bandwidth has increased tenfold. For all performance applications, the quicker something can be moved the better, this includes gaming. As always though, you won't really notice an immediate benefit to RAM revisions at the start of its life, as its release specs are usually similar to the high end specs of the previous. Look at how much DDR3 has improved over its life, hell it was rare to see something faster than 1333MHz when it released and now its able to get past 3GHz.

  • X99

Really nothing special. They're expensive boards with features that you'll never use and will most likely never get market share. (Read: Thunderbolt 2)

Like what exactly? For me I may never use the thunderbolt standard, but the additional USB3 in the spec, increased number of SATA3. (And as I have seen as major response over all here) the additional PCI-e lanes.

Again many of these things don't benefit my gaming rig, but for my productivity rig that has multiple HDD's for scratch purposes, 10G ethernet, Infiniband card, RAID cards (for the more important data) and GPU's for acceleration purposes. The more bandwidth available to these cards, the faster the manufacturers can try and make. If the potential for higher bandwidth is there, yes they can't change their existing stuff, but it will allow them to make even faster add ins.

 

I just feel that you haven't accounted for everyone within the computing world.

AMD has had more PCI-E lanes than Intels Z series.. They are 2.0 though.

X79 had 40 from beginning to end. Half way through their lifespan they acquired 3.0 40 pci-e lanes. If you wanted the lanes, get an X79 board and the cheapest cpu you can find. Unlike X99 you don't have to spend 10-50% more on a cpu to get 40 lanes, nor 60% more on ram that performs roughly the same, and you won't deal with the bugs that will plague the X99 platform for the upcoming months.

That statement is false. There is not much change.

I don't know why tech sites are raving over it, it's not that much better then X79 or AMD's higher end offerings. Its benefits are more USB 3.0 ports, better efficiency, and.. and I can't think of anything else. The TDP wasn't even lowered from X79, so, I don't even know. They at least bracketed it appropriately, the less lanes on the 5820k is a sin to the X series but it adds to its tiering, the 5930k has all the lanes + a little more performance, and finally the 5960X has all the lanes + great performance. Either way a 4790k or 4930k/Xeon (for the same amount of lanes as the X99 platform) can be used for video editing at 4k, it takes more time yeah, but you save so much. Maybe in a few years it'll be worth buying.

But its Intel, and their "first" 8 core processor deserves all this attention and praise I suppose. The 8 core the really the only CPU worth purchasing for X99 right now. Anything else and the 6 core previous gens will perform roughly the same for less cash. I do believe it will be at least a 3 year platform but that's because of used prices and the 5960X will still be somewhere near the top in terms of 8 core usage..

Thanks for your reply. The point I'm trying to make is that very, very few people can justify the price of X99. It's not going to magically give you more bandwidth for your HDD's and whatever other I/O devices you use, so unless you're craving more PCIE lanes there's little advantage. As for the Opterons, I only researched their x264 performance as that's the only thing that concerns me personally.

 

Remember everyone, there is a difference between "I can take advantage of this!" and "I need this so bad I'm willing to spend $1800 on it."

This leads me to a question then, is the x264 encode library developed enough to fully utilise all 16 cores and fully saturate them?

But one critical flaw to this is that we are making a comparison of a CPU on the brim of release is being compared to one released 2 years ago. A CPU which also cost over $1000 when it was released.

It's not going to magically give you more bandwidth for your HDD's and whatever other I/O devices you use

I'm not saying it will magically improve existing hardware, I am saying it will aid in the development of faster hardware. For example, in a RAID card, you can rely on pretty much every system having an 8x slot free. That's great, but bear in mind an PCI-e 3.0 8x == PCI-e 2.0 x16. If vendors can rely on that interface being avaliable, which they can if more are provided by the chipset which makes it more likely they'll appear on boards. They will make the product for it. If you have a HBA card, that externally connects to a SAN. Providing that the SAN can read at a high enough speed, which is going to have the higher throughput?

Anyway, I accept to the average user this is going to have little or no benefit immediately but it all aids in the incremental performance gains across years.

So, I have more money than sense since I work my a$$ off. I probably have no where near a need other than for the fun of it but I totally plan on a x99 build with 32 gigs if ram and two 780ti's. I'll load it with a few tb's of ssd's. And I'll still spend most of my time playing minecraft. 

Now, am I being scammed or do I just enjoy playing with systems and seeing what can do what. At the same time guys like me help afford the next round of chips and research for more. Without guys like me the guys who really need it might see these jumps disappear. 

All great points. I'm more looking towards what Express SSD, PCI technology, and 28/40 lanes will do for AMD counter CPU.

Even knowing I have a FX 8350, and 7970CF I'm far from being a AMD fanyboy. The truth is, this opens up doors for a new line of future PC gaming.

1.6g writes

True 3 way SLI

Thunderbolt

DDR 4 that rumored 4k MHZ

 These are all the new features that us games can look forward to. Saying AMD is the better buy or Intel is to expensive yad yad yad... Means nothing. Have to look at the bigger picture.

You may see some of the new features on the revamped Z97 or AM3 platforms in the future even a 5690K I5 for us games ect..

 

they made a E-6core $380. and they introduced DDR4 (it had to be introduced) and added 2 more cores to the top end.... 20-33% performance boost across the board. usb3, all sata 6Gbps.....

i mean.. what do you want? XD that is a pretty significant boost from the previous platform. what kind of progress would you have liked to see?!?!?!

My homie, I think you forgots to do some reading: http://ark.intel.com/products/77780

http://ark.intel.com/products/77779/

All the newer 2011 cpus support 3.0 40 lanes and the older ones 3820, 3930, etc.. All support 40 lanes of pci 2.0..

Would you please present evidence of this? I know for a fact they are not limited to 16, otherwise 3 cards in crossfire or sli would present the same frames.. but the Z97 lags behind.. Or the really high end 295X2 in crossfire would present the same fps as Z97, but it does not, the x79 has higher fps.. Or people couldn't run two video cards and a x8 pci ssd at their full potential at the same time (it'd be x8 x4 x4).. But people do and have the same benchmarks as people using just that device...

But if you can prove that they don't, I'll believe you. And you'll crush every single gamers benchmarks that uses tri or quad gpus with your singular proof.


I know gaming isn't a perfect answer, but, it shows that the cards do use the bandwidth..

There are scenarios where using X99 makes sense. I'm looking into it myself because I want to be able to run VMs for software development during the day and play a game or two during the evening. I also really like the idea of having a powerful machine that can virtualize anything I want instead of having to buy appliance-like hardware.

I understand OP wants to justify his own purchase and grab the attention of like-minded people to circle-jerk with. Don't forget the same companies that make your budget parts are being sponsored by the enthusiast that want a premium product and are willing to spend extra. There's already too much stagnation with the shift to mobile first and centralized services. I want my personal cloud.

Also, not everyone wants to build a wannabe peasant box just to play games. I would rather have some headroom so I can try new and interesting things. Like running a 4 headed Linux gaming machine with virtual machines that each have their own dedicated graphics card.

Well, you don't really need an X99 rig to run some httpd instances in containers, do you? A Raspberry Pi will do that just as well.

If you're going to run really heavy compute applications, X99 is not certified for 24/7 operation, it's basically made to idle in the most expensive way conceivable.

Another thing to consider is that X99 is consumer-oriented, unlike C-chipsets for instance. X99 is an "enthusiast" marketed product, not a "professional" product.

For a personal cloud, X99 is a hell of a power hog. As I said, a Raspberry Pi can do that just fine, or maybe a small Atom SoC board is you want Intel because you want to be sure it has "NSA-approved" random number generators and encryption microcode.

X99 is an expensive beta testing program for possible future interesting hardware. It is what it is, 25% more bits and pieces of everything that was already there for 100% or more price premium to give a performance boost of maximum 20%. It's not even known yet if the X99 mobos support the virtualization features, because they're pretty much all consumer ("gamer" or "SOHO") boards.

The general rule of thumb history has taught us about Intel, is that the more noise Intel makes when it launches a product, the less interesting the product is. There has never before been more noise made about a product that only very few people potentially buy, than now with X99. Look at how much samples Intel sent out to generate those YouTube videos... wow, that's not what they do with Xeon CPU's is it?

My conclusion: X99 is like a Porsche 924/944 or like Windows NT/XP/Vista/7/8/Server, a really innovative old NSU/Audi model that Audi didn't think was interesting enough, or an operating system that was declared end-of-life by IBM, but that can still be sold as an expensive superfluous luxury product with a lot of marketing effort.

I run an 7 old year budget board 24/7 without any problems. I do worry about the power draw. Why would a premium board that goes through server-grade testing not be able to do this? The shift to DDR4 also gives a 40% reduction in per draw.

I thought the whole point of this platform was to combine the ability to run efficiently but still be able to push it when needed and thermal constraints allow it.

If X99 is an old Porsche, what's the equivalent of a BMW i8?

I use my dual GPUs and dual NICs very often. I render on one GPU, while I game on another GPU, dual NICs are bridged and allow a faster dl/ul rate. I honestly have no issues with my CPU while rendering or compiling and I have no need to upgrade to a X99 platform. I'll most likely update my GPUs 2 or 3 more times and then hopefully there will be a new AMD platform I can move to. Not a fan boy just like the red and black better then the blue. 

Yes, x264 will fully utilize every core you throw at it at all times.

You're right: we are comparing tray price to a used street price. I don't think it's a very flawed argument though, because I highly doubt Haswell-E prices are going to get a significant drop anytime soon.

So basically your argument is that if enough people buy new HW, HW manufacturers will be more likely to utilize new HW since the market share would be better. That sounds logical but IMO you're throwing your money on hopes and dreams that will never happen. What it really boils down to is too many individuals and businesses are content with what they have now.

The older ones used 16 lanes that were multiplexed to 40, they weren't truly open and independent...