Intel Skylake / Kaby Lake

[aigomorla]

See that's the thing, for as much as people bitch about DLC, they still almost all buy it. So much so, that most games make more from the DLC than from the initial game! Its like people bitching about F2P w/ microtransactions, doesn't matter what they say, cash pays.

Thats why i said INB4....
If people end up buying into it, someone in management will suggest, dongle Overclocking, Dongle SLI, and well... pretty much we end up with a DLC laced motherboard which can only be unlocked though pre-orders or buying the Deluxe version with DLC's unlocked!

Actually if you paid attention to the ITX part, you'd understand why the price gap is larger than that.

There are four confirmed ITX AM4 boards, vs only one for X299, similar to X99. All the AM4 boards will be sub-150$, while the AsRock X299 will very likely be 300$.

X99 in the past didn't result in any better boards in terms of build-quality, there are lots of reports of Asus X99 boards failing left and right taking CPUs out along with them.

Overclocking is less appealing in ITX builds, and even then the difference in clocks doesn't translate to difference in processing power the way you described. We don't even know what IPC improvements are there in SKL-X vs BD-E, let alone SKL-S, and we know that Ryzen is BD-E level in terms of IPC.

There's hardly any consumer application that uses AVX2 to its full potential, let alone AVX512, and the dual vs quad channel debate is age-old and we all know how big of an issue it really is. ITX also limits the ways you can use all those extra PCI-E lanes.

The difference is more likely to be 400$, and 500$ if you consider the 1700X at its reduced price-point. So that's enough money for a GTX 1080 if you go with the AMD route.

You really need to take your AMD Fanisim out of this thread.
Its quite apparent you have Red Pride, and well im happy for you, I also own a 1800X system,

But AMD Fanism doesn't belong in this thread.
Were not here to convince someone to get a AMD, were here to talk about the said INTEL processor and platform.

So please take the AMD Fanism out of this thread and save Mark the trouble of cleanup.

 

[beginner99]

So, what are the odds that Coffee Lake's platform will require a "RAID dongle" too? If Intel ports their (firmware-based?) NVMe RAID feature to that platform too. (Only have two M.2 Ultra slots though?)

The chance is slim. They just outright won't offer that functionality (RAID from CPU PCIe lanes) to add some more market segmentation. But I think that is perfectly ok. who really needs raided M.2 drives?

 

[CatMerc]

Really amazing how Intel was able to keep this secret until (almost) Computex. Everyone I talked to just a few days ago reiterated the previous leak, even Intel in the presentation below.

8th Gen Intel Core 'Coffee Lake-S' + Z370 Chipset to Launch in August/September (4C and 6C 'K' Options)

Rest of the lineup, including a dual-core variant (native 2C KBL-S die? cut down 4C CFL-S die?), will follow in Q1-2018.

Almost like this wasn't planned at all and was a last minute decision because they realized AMD is a proper threat now.

Can't leak when no plans exist

 

[CatMerc]

'Worth' is determined in part by cost.

Ryzen engineering motherboards had 2:1 DF:Memclock in debug mode, and it is not known how Snowy Owl and Naples would have it configured. Broadwell-E already showed latency issues with 22+ cores, and my hunch is that the revised cache configuration in Skylake-X is aimed primarily at alleviating that problem. What it'll do in case of lower core counts is anybody's guess.

I don't want to go nitpicking about the details regarding to what extent Naples is NUMA-like(the technical thread has tons of info about that), but historically AMD has been more comfortable with NUMA than Intel since the Opteron heydays.

In fact, the HCC and XCC variants of Intel dies use two or more ringbusses. Which, you guessed it, creates NUMA like behavior

 

[Glo.]

$300? That's actually not bad.

Even if you throw out such things as AVX-512, Quad-Channel Memory Controller, Extra PCIE Lanes, the more quality motherboards and components and the general stability of the Intel HEDT platform, you're still left with a cpu to cpu comparison. Here, base is 3.6ghz vs 4.0ghz. That's 8x 400mhz + HT, translating into 3.2Ghz extra processing power. Turbo is 4ghz vs 4.3ghz, with another 200mhz for a 4.5ghz 2-core turbo. With AMD, you're stuck at 4.0Ghz overclock, while the Intel potentially offers more - certainly, 4.5Ghz should be minimum as it's only 200mhz above turbo frequency. Taking Intel's superior ipc plus all the goodies mentioned above into consideration, the 7820X is more appealing to me.

If you go Ryzen 1700 vs i9 7820X, the frequency gap increases dramatically: 8x 800mhz (7.4ghz + HT more processing power) at base, 8x 700mhz at turbo clocks, and 4.5ghz 2 -core turbo the 7820X. Overclocking is 3.7 - 3.8Ghz vs 4.3 - 4.5Ghz. Add ipc and extra Intel HEDT features and again, things don't look as bad as some would have you think. And this is even without the Intel tax

Finally, those $120, feature-limited boards people bring into these comparisons actually doesn't help but reinforce AMD's "cheapo" image in many minds. It is also quite dishonest. HEDT is just that, HEDT, and Intel's HEDT boards are built like tanks, to endure the demands of multicore processors, quad-core and extra IO and PCIE lanes. Proper comparisons would be the X370 and Taichi boards. If I were to build an AMD system, I wouldn't be considering any feature-limited boards. So for me, the point is really moot - i9 7820X all the way! It's price/performance ratio is not as bad as it seems. It's in fact, great value for the performance and features you get when compared to the competition, and that includes the Ryzen 1700.

The problem is this. If you want to build 4K gaming computer, you will not see the difference in performance between Ryzen 7 and 7820X. Because you are GPU bound. None of the features will benefit you. Content creation? Intel does not have that huge advantage in IPC, and all core turbo states are most important here, not Turbo Boost 2.0 and 3.0 which are max's for single core Turbo's.

If you cannot see the benefit of your money invested in Intel platform - why spend it in it in the first place? I cannot justify spending another 300$ for Intel platform if I will not see the benefit of it. And nobody will buy HEDT to play games in 1080p, even with high refresh rates. That is the biggest problem with Intel platform in current state, that I have. It is not flexible, in terms of performance offered and cost. It is great product, but their competition also has great product, that is much cheaper, and I will not see any difference between them in my usage.

The other problem is that I am Intel supporter, and going mITX ASRock MoBo, Quad channel memory plus 7820X is pretty much... well, very tempting. But the money invested and no benefit from it other part of discussion.

Everything we write is just justifying points of view. Keep this in mind also, guys.

 

[Grooveriding]

Just watched that Linus video and spent about 30 minutes unraveling the bifurcation they've put on this new platform. I get that Intel will be offering 18c chips now, and that is 10 cores more than my current chip. Looking at pricing though, it seems it would cost me $4000 now to put together a top end system of the new parts consisting of motherboard, CPU & RAM. Contrasted to the $2000 it cost to do the same a couple years ago.

What happened to getting more for the same money as time went on ? With these new chips at the same $2000 pricepoint I can only get 2 more cores and a newer architecture that will amount to about 10% more IPC and may not overclock as well. I can spend $200 more for 4 more cores, or $1800 more for 10 more cores.

It's nice that this is all reactionary to AMD's surprisingly decent performance with Ryzen and the upcoming Threadripper. Intel still has the better chips if you don't care about price, but they're charging exorbitantly higher prices for what is now a much slimmer advantage over AMD than they held before. I hope Threadripper can be successful and give Intel a black eye, making them curtail what is becoming ridiculous pricing schemes. Linus also has a good point that Intel could have given customers more a long time ago. Now when they're forced to by AMD, they're doing it in a half-assed fashion.

The lack of solder also is BS! I'd like to know if their Skylake Xeons are going to be soldered and the Skylake-X chips are being given the shaft intentionally.

 

[mikk]

The problem is this. If you want to build 4K gaming computer, you will not see the difference in performance between Ryzen 7 and 7820X. Because you are GPU bound. None of the features will benefit you.

Not all games are GPU bound.

 

[Zucker2k]

The problem is this. If you want to build 4K gaming computer, you will not see the difference in performance between Ryzen 7 and 7820X. Because you are GPU bound. None of the features will benefit you. Content creation? Intel does not have that huge advantage in IPC, and all core turbo states are most important here, not Turbo Boost 2.0 and 3.0 which are max's for single core Turbo's.

If you cannot see the benefit of your money invested in Intel platform - why spend it in it in the first place? I cannot justify spending another 300$ for Intel platform if I will not see the benefit of it. And nobody will buy HEDT to play games in 1080p, even with high refresh rates. That is the biggest problem with Intel platform in current state, that I have. It is not flexible, in terms of performance offered and cost. It is great product, but their competition also has great product, that is much cheaper, and I will not see any difference between them in my usage.

The other problem is that I am Intel supporter, and going mITX ASRock MoBo, Quad channel memory plus 7820X is pretty much... well, very tempting. But the money invested and no benefit from it other part of discussion.

Everything we write is just justifying points of view. Keep this in mind also, guys.

You are very likely to see a difference between the 2 systems at 4k gaming with your x16 PCIE lanes on the Ryzen system. You are also bound to run into other bottlenecks eg. single threaded situations, way before the Intel system does. Thanks to Turbo Boost 3.0, the Intel system no longer relies on a single core to handle low threaded situations. It is dual-core now @ 4.5Ghz. Content creation is where Sky-X is going to stretch it's legs the most, in my opinion, with it's high low-thread turbo, many cores, and quadchannel, and AVX-512, it's got content creation covered. I think the gap here could be anywhere from 10% to over 50% based on software code and optimizations. In terms of future-proofing, too, I would rather shell $300 now and enjoy the Intel system for the next 3 to 5 years, satisfied in the knowledge that I am well prepared for any game or software I'm going to run into. That's what HEDT should be about, not cutting corners to save a few bucks and thereby being bottle-necked even in current gen.

 

[tamz_msc]

You really need to take your AMD Fanisim out of this thread.
Its quite apparent you have Red Pride, and well im happy for you, I also own a 1800X system,

But AMD Fanism doesn't belong in this thread.
Were not here to convince someone to get a AMD, were here to talk about the said INTEL processor and platform.

So please take the AMD Fanism out of this thread and save Mark the trouble of cleanup.

I'm primarily a fan of spending money commensurate with the kind of features that I know I'll need.

Merely having features that the other platform doesn't have does not mean it's better unless you know how you'll put those features to use.

I've seen a lot of threads about why the presence AVX512 or quad channel memory on its own means that Intel platform is better, or in the case of the comment I was responding to, the AMD platform lacking these features means reinforcing the "cheaper" mentality of their platform.

Except one or two people, I've never come across anybody who is clear about how they're going to use those features, and yet comments justifying the price premium due to those extra features making the Intel platform inherently better are repeated ad nauseum.

 

[nvgpu]

https://forum.beyond3d.com/threads/amd-ryzen-cpu-architecture-for-2017.56187/page-51#post-1967173

SSE vectors are only 4-wide. AVX adds 8-wide vectors and some new instructions (broadcast, permute, masked move). Also AVX is non-destructive (3 operands), meaning that AVX doesn't need to write the result over one of the inputs. This reduces register pressure and removes need to generate extra mov instructions. On Intel CPUs, you can reach up to 2x higher FLOPS with AVX compared to SSE. Without AVX you can only reach 1/2 peak FLOPS of modern Intel CPUs.

One company with their crappy AVX2 implementation is gonna lose very badly once the industry moves to AVX2. And without AVX-512 support, you simply won't run modern software at all.

 

[VirtualLarry]

One company with their crappy AVX2 implementation is gonna lose very badly once the industry moves to AVX2.

"Once the industry moves to AVX2"... LOL!

Outside specialized number-crunching, and a few particular scenarios, AVX2 hardly applies to most standard code.

Not to mention, Intel has hindered AVX and now AVX2 adoption, by keeping those opcodes limited to their top SKUs, and not providing it top-to-bottom in their CPU stack.

If anything, AMD is finally going to make those opcodes popular, because AMD will have more chips in the field than Intel that support AVX, due to Intel's extreme market segmentation.

(Even AMD's AM1 CPUs support AVX! Yet my Intel Kaby Lake Pentiums, with HyperThreading, do not.)

 

[R0H1T]

https://forum.beyond3d.com/threads/amd-ryzen-cpu-architecture-for-2017.56187/page-51#post-1967173



One company with their crappy AVX2 implementation is gonna lose very badly once the industry moves to AVX2. And without AVX-512 support, you simply won't run modern software at all.

Right, when was the last time you used a browser that "used" AVX2? FYI there's only one fork (of a fork) that I know used AVX & that fork ran out of steam a couple of years back, also you don't have anything more popular (software) than a browser out there!

 

[Zucker2k]

I'm primarily a fan of spending money commensurate with the kind of features that I know I'll need.

Merely having features that the other platform doesn't have does not mean it's better unless you know how you'll put those features to use.

I've seen a lot of threads about why the presence AVX512 or quad channel memory on its own means that Intel platform is better, or in the case of the comment I was responding to, the AMD platform lacking these features means reinforcing the "cheaper" mentality of their platform.

Except one or two people, I've never come across anybody who is clear about how they're going to use those features, and yet comments justifying the price premium due to those extra features making the Intel platform inherently better are repeated ad nauseum.

I wonder what all those Ryzen users spending money on expensive ram with the hopes of clocking them even higher think they're doing? I'm tempted to think there's some performance to be had in that area? With quadchannel, you have the bandwidth on tap, without breaking a sweat. It's plug and play. AVX2 has proven very useful in handbrake and would probably get adopted more in content creation going forward. It's advantages can't be ignored for too long now.

 

[VirtualLarry]

Interesting point, I think my browser of choice (Waterfox) DOES support using AVX2, mostly for memory-moving for faster scrolling, though they might use it for media-decoding as well. I had forgotten about that.

At least, they did at one point; they've changed compilers several times, I don't know if the current builds support it.

 

[Zucker2k]

"Once the industry moves to AVX2"... LOL!

Outside specialized number-crunching, and a few particular scenarios, AVX2 hardly applies to most standard code.

Not to mention, Intel has hindered AVX and now AVX2 adoption, by keeping those opcodes limited to their top SKUs, and not providing it top-to-bottom in their CPU stack.

If anything, AMD is finally going to make those opcodes popular, because AMD will have more chips in the field than Intel that support AVX, due to Intel's extreme market segmentation.

(Even AMD's AM1 CPUs support AVX! Yet my Intel Kaby Lake Pentiums, with HyperThreading, do not.)

I'm not sure if you're serious. Certainly, AMD doesn't sell more 'AVX2 chips' than Intel?

 

[ryzenmaster]

But AMD Fanism doesn't belong in this thread.
Were not here to convince someone to get a AMD, were here to talk about the said INTEL processor and platform.

So please take the AMD Fanism out of this thread and save Mark the trouble of cleanup.

As this is an Intel specific thread, that's where most discussion should be focused. It is though entirely fair to compare upcoming products to other available products to establish where the new product will land. Comparing to both other Intel products and AMD products as well is to be expected.

Using this one simple trick people can decide which product they should go for: Does it fit your use case and your budget? Buy it!

I fully expect Intel to offer slightly better clock:clock and core:core performance in almost every scenario, but at higher price. Keeping in mind that we are talking about HEDT platform, I think it would only be fair to mainly compare Skylake-X to ThreadRipper rather than AM4. While I'm not planning on getting one, I'm glad to see Intel finally offer 8 core parts at much more reasonable price. It should make things interesting especially for those planning to buy one.

 

[Glo.]

Not all games are GPU bound.

I have to see that games in 4K are not GPU bound to believe in this.

You are very likely to see a difference between the 2 systems at 4k gaming with your x16 PCIE lanes on the Ryzen system. You are also bound to run into other bottlenecks eg. single threaded situations, way before the Intel system does. Thanks to Turbo Boost 3.0, the Intel system no longer relies on a single core to handle low threaded situations. It is dual-core now @ 4.5Ghz. Content creation is where Sky-X is going to stretch it's legs the most, in my opinion, with it's high low-thread turbo, many cores, and quadchannel, and AVX-512, it's got content creation covered. I think the gap here could be anywhere from 10% to over 50% based on software code and optimizations. In terms of future-proofing, too, I would rather shell $300 now and enjoy the Intel system for the next 3 to 5 years, satisfied in the knowledge that I am well prepared for any game or software I'm going to run into. That's what HEDT should be about, not cutting corners to save a few bucks and thereby being bottle-necked even in current gen.

And why would I need more than 16 PCie lanes for single GPU? Secondly, why do you believe that Ryzen CPUs are not future proof? I do not understand. 2 months ago, when everybody praised Price to performance ratio of AMD CPUs - Intel fans were saying its pointless, because no application is working well with more than 4 cores. Today, its Intel platform which is Future proof, but not AMD. What the hell is wrong with you guys? Why do you value your brand preferences and attachment of your mind to brand feel, over your money?

The problem with Intel platform is... Ryzen. AMD pushed HEDT CPU core count into mainstream with its prices, and Intel is still selling 8 CORE CPUs as... HEDT CPUs, with according price point, and more expensive platform. THAT is the problem here. Mainstream for AMD: 4, 6, 8 Cores. For Intel, 2 and 4. HEDT for AMD - 12 and 16 cores. For Intel from 4 to 18 cores. The platform looks like a mess on Intel side. Yes you get more features, but those features look like they were forced to bring premium feel, and to force you to pay more for it.

Guys. There is just absolutely no justification in todays world for prices Intel asks for their platform, at least for up to 8 core CPUs. Others - that is different story.

https://forum.beyond3d.com/threads/amd-ryzen-cpu-architecture-for-2017.56187/page-51#post-1967173



One company with their crappy AVX2 implementation is gonna lose very badly once the industry moves to AVX2. And without AVX-512 support, you simply won't run modern software at all.

So Broadwell-E will also be useless in AVX-512. But hey, it doesn't matter because we cannot hate on forum on it.

 

[Sweepr]

I see there's 'rumors' about Coffee Lake being delayed to 2018 circulating in clickbait websites. The original article from WinFuture.de only provides info about mobile products and a 65W 'S' part (and speculates about the rest), saying Kaby Lake Refresh comes in September, Gemini Lake in late 2017 and Coffee Lake in February. Basically nothing new because it was already known that mobile Coffee Lake was a Q1-2018 product. To clear up any confusion, the only parts launching in August are unlocked 65W/95W 4C/6C desktop models + Z370 chipset, everything else including traditional locked 65W desktop models, mobile variants (U, H), Xeon E3 and new chipsets will only be available next year. This is the latest roadmap from Intel itself pre-Computex:

And here's an article from DigiTimes published today:

ASMedia's revenues will move upward along with the release of Intel X299 series chips in June, AMD X399 series chips is July as well as Intel Coffee Lake CPUs, and Z370 chips in August, Lin noted.

ASMedia has begun developing USB 3.2 controller chips in order to cope with Intel's planned release of 300 series chipsets and expects to roll out USB 3.2 host controller and device controller chips in 2018, Lin revealed.

http://www.digitimes.com/news/a20170603PD201.html

 

[itsmydamnation]

I'm not sure if you're serious. Certainly, AMD doesn't sell more 'AVX2 chips' than Intel?

So how many years does that AVX2 adoption take...... Cuz you know its only been 6.....DX 9,10,11,12 in that time period.......

Or how about some Linus B Torvalds
http://www.realworldtech.com/forum/?threadid=168200&curpostid=168587

Are they? The reason AVX512 seems to be particularly bad is just the frequency change, not because the 512b accesses would be fundamentally slower.

That said, we've seen the "bigger accesses are slower when unaligned" pattern so many times. It turns out that trying to be very wide is just a bad bad idea.

For example, because cache reads (of all sizes) are so critical, only a complete moron would design their L1 cache to only do one maximally wide access at a time, since it's much more common that you see two narrower accesses that can be concurrent. So the L1 interface might be 64 bytes wide, but in reality any good L1 is always going to be able to do two 32-byte reads at a time, because that's actually the much more important case.

End result: there's basically no upside to doing maximally wide accesses. A good cache will do the two narrower (and more generic) accesses in parallel anyway. And the maximally wide access will always be really painful for alignment issues, while the narrower access is much more likely to "just work" (ie you can often split one unaligned narrower access into two aligned narrower accesses and still do it in one cycle, while a unaligned full-width access will take two cycles - but now re-aligning it will be much more painful).

Not to mention that the narrower thing is more likely to be aligned in the first place.

So in neither the aligned nor in the unaligned case do you actually win much of anything from the wider access width to the cache. You only buy more pain.

On the write side, where you often do have just one write port, the situation might be different, but even there the write buffer might effectively hide most of the difference between the full-width and 2x narrower case.

So honestly, AVX-512 just doesn't sound like a big win - not even if the frequency didn't screw you, which it clearly does.

Sure, you will find cases where simply doing eight double FP ops (or 16 single) in parallel will make you look really good. But they are damn rare, and they won't really sell all that many CPU's.

If AVX-512 was free, that would be another thing entirely. But it's not. I'm biased (since I don't care about FP much anyway), but I'll take 128-bit vector units any time if I get more of the rest of the CPU instead.

Linus

 

[tamz_msc]

I wonder what all those Ryzen users spending money on expensive ram with the hopes of clocking them even higher think they're doing? I'm tempted to think there's some performance to be had in that area? With quadchannel, you have the bandwidth on tap, without breaking a sweat. It's plug and play. AVX2 has proven very useful in handbrake and would probably get adopted more in content creation going forward. It's advantages can't be ignored for too long now.

There are two reasons why faster RAM matters in case of Ryzen. One is due to the increased bandwidth in itself - even Kaby Lake 4C can become bandwidth starved. Second, and more importantly, faster RAM means faster inter-CCX communication because of the way the Infinity Fabric works. That is why it is recommended to get the best RAM you can buy if you're buying a Ryzen.

AVX is, very loosely speaking, a way to use the vector SIMD ALUs on the CPU instead of GPGPU. Encoding is limited by the block sizes of the algorithm that you're using, and so the performance uplift is only 20-30%, and that is mostly restricted to H.265.

Are you doing parameter estimation in a 50-60 dimensional parameter-space? That's where you'll see the 2X performance uplift that AVX2 brings, and AVX512 on top of that.

Do you visualize extremely large data sets by feeding >300MB text files into MATLAB? That's where quad-channel would come in handy.

You won't find these use cases discussed in tech reviews, and neither are they of particular interest to the majority of users, including the most heavy content creators(except the storage and expandability options, for which AMD is preparing an option).

 

[Glo.]

I wonder what all those Ryzen users spending money on expensive ram with the hopes of clocking them even higher think they're doing? I'm tempted to think there's some performance to be had in that area? With quadchannel, you have the bandwidth on tap, without breaking a sweat. It's plug and play. AVX2 has proven very useful in handbrake and would probably get adopted more in content creation going forward. It's advantages can't be ignored for too long now.

Yeah, instead of paying 40$ more for faster RAM on Ryzen platform and paying 500$ for CPU+MoBo combo, lets pay 900$ for CPU+MoBo combo to pay 40$ less for Quad channel memory setup vs dual channel/faster RAM.

What the hell is wrong with you guys?
Ryzen 7 1700X - 349$ + mITX MoBo 149$ + G.Skill 3200 MHz 16 GB 14 Ns - 173$. Around 670$ total.
Core i7 7820 599$ + mITX MoBo at least 199$, more likely 299$ + 16 GB 4 channel memory 2666 MHz - 136$. Around 1030$ total cost.

And there are rumors about AMD lowering the prices after Intel brings X299 to market, and AMD bringing Threadripper to market. What if then you will be able to buy Ryzen 7 1700 for 250$? Price to performance will be complete and utter bloodbath for AMD.

 

[Zucker2k]

I have to see that games in 4K are not GPU bound to believe in this.

And why would I need more than 16 PCie lanes for single GPU? Secondly, why do you believe that Ryzen CPUs are not future proof? I do not understand. 2 months ago, when everybody praised Price to performance ratio of AMD CPUs - Intel fans were saying its pointless, because no application is working well with more than 4 cores. Today, its Intel platform which is Future proof, but not AMD. What the hell is wrong with you guys? Why do you value your brand preferences and attachment of your mind to brand feel, over your money?

The problem with Intel platform is... Ryzen. AMD pushed HEDT CPU core count into mainstream with its prices, and Intel is still selling 8 CORE CPUs as... HEDT CPUs, with according price point, and more expensive platform. THAT is the problem here. Mainstream for AMD: 4, 6, 8 Cores. For Intel, 2 and 4. HEDT for AMD - 12 and 16 cores. For Intel from 4 to 18 cores. The platform looks like a mess on Intel side. Yes you get more features, but those features look like they were forced to bring premium feel, and to force you to pay more for it.

Guys. There is just absolutely no justification in todays world for prices Intel asks for their platform, at least for up to 8 core CPUs. Others - that is different story.

So Broadwell-E will also be useless in AVX-512. But hey, it doesn't matter because we cannot hate on forum on it.

Can we agree future-proofing is simply about eliminating bottlenecks as far into the future as the hardware is capable? Good. So let's go over the numbers again; AMD relative to Intel:

* Gives up 400mhz to 800mhz on base clocks
* Gives up 300mhz to 700mhz on turbo clocks
* Gives up 500mhz/1.0ghz to 700mhz/1.4ghz on single/dual thread turbo
* Lacks Quadchannel (Already bandwidth starved - a bottleneck)
* Only 16 PCIE lanes
* Lacks AVX-512
* Overall platform quality not quite up to Intel standards

You can't give up all these and consider yourself futureproof, relative to your competitor. Your memory subsystem is already buckling. Why buy into this kind of headache, much alone believe you're ready for the future? It's quite clear to me core count alone is not enough here. You need more here, and Intel has the superior platform. They could do better, yes, but show me another platform that offers the same value for the money and then maybe I'll consider it.

 

[Glo.]

Threadripper, X399.

This is the problem which you do not get guys. You compare HEDT platform, because Intel said so, to AMD Mainstream platform, when it goes for 8 core CPUs, which completely borks the perspective. AMD pushed HEDT core counts to Mainstream, Intel does not. Whats more, they want to bring Mainstream into HEDT with Kaby Lake-X. Its a problem with identity of platforms. AM4 is the same mainstream platform as is 1151 for Intel!

AMD still can bring 8 core high clocked, 150W CPU for Threadripper platform for less money than Intel, if they want to, but Im pretty sure its not so easy from marketing perspective. We all would be complaining the same way, about this decision that we are complaining about offering 8 core CPU in HEDT platform for the according prices, by Intel.

Do not get me wrong. If you compare 8 cores vs 8 cores, from both sides - its a win for AMD all the way. Its not worth to pay Intel the prices they ask. If you need more than 8 cores - that is different discussion.

 

[Atari2600]

The chance is slim. They just outright won't offer that functionality (RAID from CPU PCIe lanes) to add some more market segmentation. But I think that is perfectly ok. who really needs raided M.2 drives?

Me.

RAID0 on M.2 would be really beneficial as a scratch area for large FEM runs. Particularly if the M.2 slots are well placed to stop thermal throttling.

 

[Edrick]

and the dual vs quad channel debate is age-old and we all know how big of an issue it really is.

There are two reasons why faster RAM matters in case of Ryzen. One is due to the increased bandwidth in itself - even Kaby Lake 4C can become bandwidth starved. Second, and more importantly, faster RAM means faster inter-CCX communication because of the way the Infinity Fabric works. That is why it is recommended to get the best RAM you can buy if you're buying a Ryzen.

So, in one hand, quad channel RAM (increased bandwidth) doesn't matter because of past experience (pre-Skylake). Completely disregarding the fact the Skylake was the first Intel CPU to memory bandwidth starved, and saw large gains in performance with greater memory bandwidth. On the other hand, you state Ryzen requires the as much memory bandwidth as possible because it is different from past AMD CPUs.

So have you seen information/test results regarding how Skylake-X (X299) handles quad channel memory that the rest of us have not? I generally do not like to assume, but I think it is a safe bet that the test results will differ from Haswell/Broadwell simply from the fact that Skylake was much different.

After reading all your posts, it is clear that you are an AMD fanboy and will do your best to continuously move the goalposts to fit your agenda.

Insulting other members is not allowed.
Markfw
Anandtech Moderator