AMD's Ryzen CPUs (Ryzen/TR/Epyc) & Vega/Polaris/Navi GPUs

Thanks.

Even as an Intel user, I feel AMD has let 'us' down again.

 

Anyway, if I can build a 8-core Zen tower with a GTX 1060 or RX 470 within HK$3500-4000 then I'm all set. Goodbye Intel!

 

Not quite. 4-core will always win in games. So if you are talking multiplier or speed, it will lose in single core. You should properly discuss it against 6 and 8 cores. Raw processing that can use more cores, Ryzen kills an Intel quad core. So your statement needs spelled out more.

Yes and no. So, the pcie lanes are 28 or 32 (still need clarification and there is a chance of variance depending on version of the chip), but this falls far short of the 40 lanes of an Intel 6900K, meaning limits related to gpgpu and gpu intensive programs. It only has dual channel, not quad channel ram. Because the F3 variant had 36/39 boost and the F4 variant supposedly is 36/40 boost, with the IPC being just a little lower clock for clock with Broadwell-E (from early sample benchmarks, with later variants suggesting slightly higher IPC), the higher clocks suggest it beats the 6900K at stock. Meanwhile, we do not know if the headroom for overclock was left or if they timed the chip to the edge, something very important. The 5GHz on a single core on air was limited by the developer's board not having strong enough power capabilities to push all cores further to find out the all core limit. This was with a chip whose base was 3.1 at the time of testing. So, enhancements since and enthusiast boards at retail suggest it may overclock well.

Pricing- the SR-7 black edition (flagship)
8- core will likely be $500, rather than the Intel 6850k which is over $600, but has 40 lanes and quad channel support, and the 6900K, which costs around $1000. If the prefetch works well, it positions the Ryzen 8 core to be a budget power house. If you have Broadwell-E, it is not worth the upgrade, but there is a chance that it has a position in high end computing, even among the upcoming skylake-X (for example, due to pricing and pcie lanes, it beats out the 6-core 20-24 lane S-X so long as Ryzen overclocks very well).


Sent from my SM-G900P using Tapatalk

 

Confirmation bias. If you noticed, the other person has no facts on which to base his/her opinion, and what was said was factually inaccurate. This, to an objective viewer, should have said disregard the statement. Logic brings truth. Please read my post responding to you.

Sent from my SM-G900P using Tapatalk

 

If that APU could hit over 3000 3DMark 11 points and could keep the performance also in gaming, that would be nice improvement! But I think it is only possible if they bringing back the good old 45W TDP limit...

 

Confirmation bias? Possibly. But no less 'factually inaccurate' than your assumptions.

Because that is all they are (still) at this point.

 

False. My assertions are based on factual grounds. I have hedges and multitude of situations where it can vary from my stated opinion. I use facts as to extrapolate data. Skylake from haswell was only around 10%. Kaby lake runs hotter than skylake using less voltage, meaning even though higher clocks are achievable, it has heat limits. Considering this and the most recent benchmarks (not based off early evaluation samples) had potentially 5-10% improvement over Broadwell-e before the higher clock levels of 36/39 or 36/40 were used, this directly states AMD IPC beats broadwell-e (all but proven). You then compare a chip beating a 6900K to the lowest end 6 core chip from Intel, with fewer pcie lanes, even with IPC improvements, it loses at stock (even the 40 lane variant would likely lose, but that is further than I feel comfortable asserting at this point).

Now, if skylake-x implements hexa-channel memory and your programs can fully utilize it, then that changes some, but not all, of my stated opinion.

Instead, you cheer for a loss when all factors point toward a qualified win. Will it outright beat Intel, no. Will it beat certain high end enthusiast chips from Intel? Yes! Will it be able to be built for cheaper than Intel machines? Yes! So, please try analysis and released facts than just assertions.

Sent from my SM-G900P using Tapatalk

 

Yawn. Spec's, BM's, 'scores' and other lies.

I don't care about the theory; I want to see real world results compared: platform vs. platform. There, I highly doubt an AMD 'win'. Even a 'qualified' one... again; for my workloads/workflows.

Not doubting for a second they'll be close - but like I've said before; let's see what Intel answers with...

 

God you're ****ing dense! The pcie lanes won't matter to most people. Nvidia has reduced much support down to only two SLI cards. Two SLI cards, even Titan X Pascal, do not fully saturate pcie 3.0 x8, meaning anyone not using a Xeon phi or commercial GPUs will not saturate the lanes, meaning the average user will see no difference between Ryzen and Intel chips! Same with memory, which saturating bandwidth is so hard, little benefit accrues in everyday workloads for speed over 3600 (benchmarks are mixed and depend on program; doesn't apply in some situations; this point is less true than the pcie discussion as quad channel will double the data throughout, almost (closer to 80%), which is perceived, but cpu speed on all cores will impact more than ram)! Now, if you know you need quad channel and need extra PCIe lanes and aren't just buying them as a gimmick, or are a benchmark number chaser, Intel provides the highest end product. Also, having more lanes dedicated to m.2, which Intel can do, is important to some. But, when you see AMD perform an encoding task clocked at 3.4 and beat a 6900K with boost (meaning the chip was running at 3.5-3.7 all cores) and beat it by 10%, then that chip that won with lower clocks is clocked with an extra 200MHz (3.6) base clock with a boost of 3.9-4.0 (an extra 10%), AMD stomped on Intel's balls! After looking at skylake efficiency and Kaby lake, both of which feed into the skylake-X design, you are looking at 15% improved IPC, better overclocking than broadwell-e, with a lot of extra heat of Kaby lake is an indication, although broadwell-e is a ****ing oven like haswell was, so that is a wash). So, unless specialized uses, average users have AMD as a budget option, so long as it can overclock like a champ. I still wish it had quad channel, but belittling Risen without pointing to facts or reasonable extrapolation of information, means you are a fanboy just here to try to piss on AMD. I've spoken even handedly and am not pointing to internal amd marketing puffery, etc. I'm relying on neutral, publicly available information to show where AMD stands ONLY WITH REGARD TO IT'S 8-core. There isn't anything to analyze 6 and 4 core or APU. So please, actually discuss the topic or leave!

Sent from my SM-G900P using Tapatalk

 

Oh, yes, I have absolutely, look back to the page 5 at this topic. There were few preview benchmarks of Ryzen; video encoding, gaming and power consumption. Those numbers described about what to expect and Ryzner definiately not going to be like Athlon 64 back in 2003.

 

I'm the fanboy?

No; I'm realistic. Nothing points to the "AMD stomped on Intel's balls!" as you want everyone to believe.

If/when I'm wrong; I just admit it and move on.

Right now? You're trying to fit jigsaw puzzles of unicorns together*.

Your unchanging point of view is just as valid as my unchanging point of view and both move the conversation forward.

I think the difference is that I'm willing to change my mind - when actual facts are presented (real world; not BM 'scores' nor 'specs' comparisons), whereas you seem to think this is the hill to die on?

Just try to see it from my point of view - you don't have to concede to it publicly - I've seen many promising 'advancements' that were 'proved' on paper (theory/specs) that just vanished into thin air the minute the platform/component/system in question became available to normal people (cough, cough; if they became available at all...).

You keep claiming you want to discuss facts, yet the biggest fact you ignore: real world doesn't care about theory. Real world just wants something bigger/better/badder than what it already has 'now'.

*TradeMark @ tilleroftheearth 2017

 

So, you are referring to encoding on a system with 3.4 locked for all cores for encoding and an early evaluation samples that showed it losing to 6900K with 3.1 base and 3.4 boost, compared to 6900K chips with boost, then choose to compare single core performance with 6700k, to which even the 6900K losses by a significant margin, to say this chip is worthless. Do you understand how optimizations work? Do you understand performing math for roughly linear performance by adding speed, going from 3.1/3.4 in the early unoptimized sample to 3.4 being locked in the encoding tasks, to 3.6/3.9 in the F3 variant, to 3.6/4.0 in the F4 variant likely to be the final release version of the flagship. The fact you don't look at the blender encoding in August followed by handbrake encoding in December to examine chipset optimization estimates (which are not exactly transferable as blender is openCL giving way to heavy optimization (note-settings and file were given to the market for comparisons on own hardware) and handbrake uses different instruction sets to perform the encoding tasks (also file and settings released)) says you are not trying to actually analyze performance, instead pretending to be an authority and just stating an opinion.

After the speed increase, those numbers no longer reflect performance until you adjust for multiplier increases, which you have not done.

If you are an authority, state your name, job title, and the reasons you are an expert on estimating this sort of performance. Without this, you are a nameless person without the credentials to show why the **** your opinion, without any ****ing analyses, should be heeded.

Sent from my SM-G900P using Tapatalk

 

@tilleroftheearth - I'm somewhat confused at your posts in this thread. You're not really bringing any arguments of value. I say this with the utmost respect (as we've had discussions in the past, and I feel you're a valued member of the NBR community).

On the one hand, you say that when Ryzen comes out, you're skeptical on what it can do. You'll not be putting much stock into Ryzen reviews, actual benchmarks, clock speeds, etc., because it has to work in your specialized environment. Fair enough. But without reading and researching what others are reporting, what are your plans on evaluating if it will be superior with your daily workloads?

Then on the other hand, you post that even if Ryzen does meet or exceed expectations computational expectations of everyone else, then you're still not interested. Instead you will wait for (BTW... who knows how long) an intel response, which may or may not exceed the current release from Ryzen. Again, fair enough, but this is where I become confused.

You don't want to discuss the current reviews, rumors, benches, etc. of the upcoming CPU release because that doesn't work for you. You don't say how you will evaluate once it is public, and even if it does become public, you're not interested in that, but rather the next big thing from intel. So, I'm a bit perplexed why strike up an argument with @ajc9988 to begin with?

 

My thoughts exactly.

 

I wouldn't get too excited about Ryzen just yet. Wait till it releases... remember the RX480 hype? Patience, young Padawan.

 

The difference is the rx480 was said to need two to compete with one 1080. That was disclosed at the unveiling. Now, the issue of pulling too much voltage from the pcie slot was unforeseen, as was losing to the 1070, but in that regard it was priced appropriately.

The big question is whether the voltage supply was a matter of the VRMs, as detailed by canard pc, or if it is a pin delivery failure, preventing enough voltage from reaching the CPU (as Intel did which requires a mod to do extreme overclocking on specific chips). But, at stock, I will stand by my assessment of the chip based on currently available information. At most, 20% improvement between HW-E and skylake-x, if the improvement scales as on performance chips AND Kaby lake improvements are integrated into skylake-x. As such, and adding the speed improvements from early Ryzen samples, unless quad channel provides a significant boost, I will put Ryzen above broadwell-e, lower than skylake-x, with same core count, at stock. Overclocked is another matter, though....

Sent from my SM-G900P using Tapatalk

 

I intentionally take everything with a grain of salt, but with that being said, I am looking forward to Ryzen. I think it'll be a real game-changer in terms of the CPU market. Intel basically has the CPU market cornered, hopefully Ryzen can turn things around in the consumer's favor.

 

Here is a description of the infinity fabric:
http://wccftech.com/amds-infinity-fabric-detailed/

The implications of it are truly remarkable if it works well...

 

I don't know how much more clear I can make my point(s) - especially after my post #114?

Common sense tells us that nothing is 'leaked' or 'posted' by a manufacturer that won't show them in the best light. That tells me that platform setups, 'tests', 'scores' and other trivia is hand picked to create a buzz for the as yet to ship product.

Like Tanner@XoticPC states (full quote below); patience is required here.

Not just because this is AMD with their less than stunning track record over the last decade or so... but maybe; especially because of it...

jclausius, it's not that my environment is specialized - actually, it is the 'specialized' nature of the hand picked platform/scores that are presented that is my issue - those are nobody's environments unless all they do is run the exact same benchmarks with their platforms...

ajc9988, I agree; this isn't about me (never said it was...) but I sure can't stand by and take marketing fluff being passed off as 'facts'. Even by a well meaning member like yourself.

The litmus test for 'better' is what 'I' can achieve in 'my' workflows. And here, 'I' is every 'I' here; you and me and everyone else included - and based on each of our specific real world workflows and workloads.

Not marketing fluff sprinkled with theory, multiplied by anticipation and added to years and decades of yearning to see Intel knocked off their decade plus year old throne...

Like I said; where I'm at (platform-wise) where I can afford to wait to see Intel's 'answer' to Ryzen. What happens after that is anyone's guess (no matter how well educated that guess may be...).

But my take based on the years of years of 'proof' that AMD has shown us over and over so far? I'll be using Intel for a good long time.

For AMD to get a possible ~40% or higher increase in productivity is easy right now (that's how far behind they are...). To complain about Intel's 'mere' 5% to 10% increases per new platform/cpu generation just shows how little people understand TVM equations - and - how they correspond to 'productivity' and 'computing' in general.

My main point here is really just wait to see what AMD actually delivers. Because what they say and what they do is not jiving for a very long time... And my gut feeling still says that even if they catch up and ultimately surpass Intel with Ryzen... it won't be a position they are able to maintain/sustain for the long term (and I vote with my wallet for long term relationships - even as I'm adapt at minimizing the cost to me and my business with products/platforms/components that ultimately fell short of my expectations for them - even after all my due diligence, too...).

In the '70's and '80's I would read tech articles similar in style to your posts here. Most of those promised/anticipated/and eagerly awaited 'advances' never saw the light of day... and the 'science' and the 'analysis' of those ancient ideas/potential products was of a much higher caliber too (with all due respect).

All I can say to AMD is what Nike says' to would-be athletes all over the world; JUST DO IT!

(Then, we'll be having a conversation about more than marketing fluff).

 

Evidently you don't know how canard pc came about their CPU for review. They purchased it off of someone outside of Canada and were not subject to the NDA of the ES that gagged other reviewers. That means this wasn't a controlled leak nor was it to show AMD in the best light. It was to break a story ahead of the rest of the media, which the rest of the media was gagged to show fluff. As you haven't even researched the source of the largest leak to date on the product, you make many unfounded assumptions. Note, could you or I have run the platform through its paces more? Potentially. But that is what gives the information provided more credibility.

Now, with that being the case, you extrapolate performance going from 3.1/3.4 on an early evaluation sample to 3.4/unknown for the show in December to 3.6/3.9 F3 variant to a 3.6/4.0 variant which is assumed to be the release flagship. That is a 10-17% increase over the numbers presented by the massive leak, while the 6900K was under 9% performance over the 3.1/3.4 early sample, suggesting that AMD has a winner. This is further compounded with the chipset not being optimized as well as other aspects of the architecture (the tests were performed in August, while the wait time for release of benchmarks was likely double checking imputation of NDAs, checking gray market importation, and waiting for maximum impact (considering the proposed announcement was pushed from October to CES, AMD holding a press conference to show it off in December, etc., lend the time frame of the leak in late December to reach the largest audience, after all, this is a tech magazine business first and foremost)). So, to assume this was a controlled, managed situation as all corporations do, you are wrong. In fact, the first article was very critical of AMD, while the second article explained the circumstances surrounding their acquisition, information on it being an early sample, etc., which sounded softer on AMD and likely an agreement found between the companies to avoid a lawsuit (give the surrounding context to lighten the negative publicity of the first article). But, once again, you speak without context (which you had ample time to research before responding, showing you just want your opinion to be pushed on others).

Instead, you are here as a cheerleader for Intel, the master of marketing fluff (skylake will be the greatest leap since core 2 then delivers around 10% over haswell while broadwell was a **** stain). Get over yourself.

Not only that, Intel hit a wall. Everyone did. It has to do with EUV lithography which was not ready in time for when Intel planned to use it. This, in addition to the market slowing on purchasing PCs because the performance reached the level it could allow longer than three years for expected replacements and everyday people are still reeling from the recession, caused Intel to bleed money developing on the tick tock with incremental improvements and causes a paradigm shift in their approach. But, once again, you do not examine the full picture on what led to now, including IP restrictions for use of SMT, also called hyperthreading (HT TM by Intel).

So, after factoring in improvements on yields by SMTC, GloFo, IBM, and Samsung, with the cross-licensing between Samsung, GloFo, and IBM on different upcoming design, manufacturing processes, etc., everyone meets at the same place from which we move forward. The figurative traffic jams formed by a railroad crossing or ferry crossing, at which all must wait to pass, leaves the situation where you cannot predict who will break out in front moving forward. Not only that, AMD has the first set of looming debt due next year, which Intel wants them to meet for antitrust reasons (which the tin foil hat suggests Intel may be making a move similar to how M$ helped Apple to avoid antitrust in the 90s and early 2000s). But this extends to 2021-22, meaning Intel's anemic release schedule looking forward may be correct (tin foil hat theory), or that they had already designed the chips without EUV (or have begun without it), so intend to ride that out, then try to jump ahead after. This gives an alternative, at least as plausible, to prevent collusion charges with AMD. But what do I know...

So, instead of opening your mouth to put down AMD and tout Intel on an AMD thread, like a fan boy, why don't you simply state what features you are dubious of, what you want to know more about, etc., instead of saying you will wait and buy Intel anyways, base assertions on an amorphous workflow no one cares about or knows, etc. Just move on!

Edit: I also forgot to mention the rise of ARM chips and Intel losing to mobile devices and trying to develop the IoT to sell more chips, which would help moving forward as this sector will grow being the PC market and can potentially be more profitable, which has also caused a paradigm shift at Intel.

Sent from my SM-G900P using Tapatalk

 

Agreed 100%.

But their track record is why I am 'rooting' for them.

Actions speak louder than words. Always have. Always will.

At the very least, Intel delivers what it 'promises' - and many times even exceeds them. Even if those 'promises' seem to be artificially low for many here...

 

"This wasn't a controlled leak nor was it to show AMD in the best light..."

Okay... so you say.

Bottom line: all your theorizing and analysis will have no impact on the final product they deliver. And furthermore, will have little semblance to reality of what Intel answers with (if they have too, of course...).

The last Intel 'update' is almost a year old now. Yawn. Has nothing to do with what AMD will be releasing in a few months time - and is currently being compared to... (get it)?

Like I said; you can extrapolate, analyze, use past facts to get your best guestimate of what will land in the not too distant future from AMD - but as you're neglecting to do the same for Intel? We see the bias you have. All good!

But don't try to tell me how to read the cards that are on the table. I know this game better than most.

 

See, you ignore my analysis of Intel's answer also, ranging from 7-15% over its current broadwell-E offering, using known data from the quad core skylake and Kaby lake and assuming implementation of certain aspects of Kaby in their skylake-x lineup, as well as more mature yields as the 14nm noise has been in use for years now.

Meanwhile, no. Ryzen is currently being compared with Broadwell-E, not Intel's answer, skylake-x, which isn't really an answer so much as a planned release that has been pushed off for over 6 months to a year (if you add up all release push backs from haswell-e on). You have until the second half of this year before Intel's next release, if they release on time!

Also, Zen+ will use the same socket, as all will, until 2020. That was confirmed. Once we reach that, it will change with the release of DDR5 and incorporation of PCIe 4.0. But you don't bother watching the market or expanding your understanding of the market. You, instead, push Intel regardless of other facts.

This shows you don't pay any attention to facts nor can you read the cards. Talking, without substance, is very transparent. Your sophistry is betrayed by the facts on the table, including my prior analysis of Intel in this thread and showing skylake-x will likely outperform, but that if you need something now and on a budget, amd wins. You ignore everything to say I'm wrong and use only your assumptions. You say I'm wrong without saying why, including calling a release of two articles, which I described accurately, as a false release, without ever reading them. In fact, it is the painstaking detail given in the second, softer article, that makes it more believable due to the discussion of legal principles and the first half not discussing the chip at all, instead trying to build the record for a legal defense and to help shape public opinion in the magazine's favor. Instead, you go to your head and historical data. Let's examine another case like that: LCDs.

So, there was coordination between every LCD manufacturer before 2007. It was a price fixing scheme. But they used brand notoriety to say anything except them is cheap and not worth your money. LG busted on the scene and started unloading way under the market price, followed by other brands like vizio. They undercut the market and that gave way to an investigation proving price fixing collusion. So, to say any of the companies involved in price fixing before the fraud was exposed were the best and nothing else could stand against them on quality would have been considered true, just as you keep pushing Intel. If something disruptive happens, even in the budget category, then you want to say it isn't until it happened, then you further push it should be ignored even if it does occur. You are promoting ignoring future facts in a realization event (the happening of an event that makes a prediction of the future true in the present)! That is how much you cannot except a happening, you've forbid the future acceptance if the event occurs!

This makes you transparent and baseless! Do you own stock in Intel? Must have been a hard couple years!

Sent from my SM-G900P using Tapatalk

 

Unfortunately

• no 6 core raven ridge apus scheduled to be released
• wouldn't meet minimum vr spec
• the apu is supposed to be PS4 level graphics (around 2 Tflops, same as 1050Ti)

but would great mainstream style laptops (think xps 15)

http://www.thebitbag.com/amd-raven-ridge-apu-much-powerful-ps4-graphics-engine/219991

 

What a BIG wall with words one of you will loose Or put it the other way... All of us will win when the competition starts Intel and Nvidia is a bit scared now. Intel loosing also on the smaller platforms. See the new ****y Jokebooks who will be pushed out. Yeah we will see other times now!!

 

Can't wait for 'other times'.

But that time is 'not yet'.

 

"For device makers, competition between ARM chips and x86 is a good scenario. It gives PC makers more choices to add to their product mix and gives them more leverage in negotiations with Intel"

"Intel will still offer x86 PC chips for low-cost PCs and continue to dominate that market. Another threat for Intel is coming from AMD's Ryzen chips in the high-performance x86 PC market"

http://www.pcworld.com/article/3161...ttle-intel-in-chromebooks-and-windows-10.html

 

I have absolutely zero interest in the consumer Zen due to dual channel memory. I am curious about their HPC offerings.

 

AMD might not release a hexacore Ryzen
http://www.guru3d.com/news-story/amd-might-not-release-a-hexacore-ryzen.html

 

This I need to correct.

It's not that cards "saturate" the bandwidth, but rather the available bandwidth between cards. PCI/e 3.0 x8/x8 scales worse than PCI/e 3.0 x16/x16 in quite a few situations, mainly related to higher resolution or games/engines/tech that are AFR-unfriendly due to requirements from previous frames. The High Bandwidth bridge was intended as a (stupid, band-aid-ish) solution to this problem that has affected even 980Ti SLI cards. x16/x16 is even the difference in positive or negative scaling when forcing SLI on some titles. It's what's holding back some engines from flat out adding mGPU support at a base level, because they don't want to (they indeed can, though) add a SLI profile that only triggers if x16/x16 PCI/e 3.0 is detected... AMD mGPU would have no problem with the bandwidth once it's XDMA.

Rainbow Six Siege, Doom 2016 with Nightmare Shadows, Ark: Survival Unoptimized and other UE4 titles (not all) are games that benefit GREATLY from it, and from the HB bridge with older drivers, and it gets even more dependent on it at higher resolutions. Yes, I did say "older" drivers, as newer drivers force all the bandwidth improvements from the HB SLI bridge into frame pacing so things look smoother. I.E. they're wasting the bandwidth. AMD does not have this problem with Crossfire any longer, since XDMA grants full PCI/e bandwidth access. Even a PCI/e 2.0 x8/x8 configuration grants more bandwidth than the High Bandwidth bridge and PCI/e 3.0 x16/x16, I believe (it certainly surpasses the HB bridge on its own, by a factor of about 700MB/s). But AMD has crap for driver profiles and no NVPI alternative, so they can't force profiles where they'd actually be beneficial, so the benefit is pretty much moot, unfortunately.

Oh, and nVidia has XDMA-style tech, mind you. They just refuse to include it in their GPUs by default, because it doesn't make them extra money. NVLink is literally XDMA on steroids. Its primary focus is improved inter-GPU connection bandwidth, with a smaller focus on improving bandwidth to the processor as well. Since they can't sell NVLink (replaces PCI/e) to consumers as it'd be proprietary, they just decided to half-ass it with the HB bridge and cut loose 3-way and 4-way mGPU support. They could have simply used the XDMA-style connection with PCI/e to a lesser extent than NVLink provides, and the world would be happy, but screw that! Competition anyone? When I complain about AMD's drivers all the time, I'm saying it because if they were an actual blasted alternative worth purchasing, we wouldn't have to "settle" for these kinds of things.

This one though, I wholeheartedly agree with... but mGPU is no longer for the average user. The percentage of new mGPU buyers (for gaming purposes! Benchmarkers-only don't apply here!) that know what they're doing to the point where they'd be hunting x16/x16 for mGPU will only increase if things don't change, or maybe mGPU will die altogether first.

 

@D2 Ultima - thank you for the more detailed analysis! I'm hopeful that the integration of L2 cache on the new GPUs (Nvidia has had for awhile) and AMDs use of infinity fiber (which allows for heterogeneous memory architecture, originally developed by AMD somewhere between 2011-13) will allow for some performance boosts and that the the new cards will be able to compete with the 1080 Ti and Pascal refresh. They have made efforts in driver optimization and working with companies, but still have a long way to go!

This is went I support them. They developed new cache systems, memory tech, etc., over the years, of which they had to shelf until now because no one utilized it in coding. When Intel and Nvidia control so much of the relevant markets, it is hard to force adoption. This is where the creation of mantle was extraordinary! It became the basis for Vulkan and Direct X 12. Also, their advancements in async computing has been amazing, although they needed work on the prefetch for the cache (which I've mentioned throughout the thread). They truly look to the future, but, I believe, had the wrong person at the helm until recently. Su pulled a lot of tech off the shelves, separated the development of the CPU architecture from the iGPU (the iGPU was being forced to be designed around the CPU and effected other design aspects of dGPU, to a degree), Radeon was spun off to be a subsidiary (protection for debt in the future), etc. Unfortunately, because of over promising, failed forced adoption efforts, needing to further develop other aspects until recently, etc., has caused the looming problem that they have a year and a half to pay debt, as well as the problem of overly skeptical eyes by consumers...

Sent from my SM-G900P using Tapatalk

 

Oh yeah. I'm supposed to be researching more about vega but I'm a bit lazy for now xD

But here's to hoping!

 

Decent article. First I heard about the L2 cache utilization was like 3-6 months ago (I want to say August, but no sure). But this discusses some of the stuff from this month, including quite a bit of comparison to Nvidia's way of doing things, etc...

http://www.anandtech.com/show/11002/the-amd-vega-gpu-architecture-teaser/3

Edit: Here is another article - http://semiaccurate.com/2017/01/17/amd-talks-vega-high-level/

 

Citations please.

 

AMD Ryzen CPUs to support Windows 7 with drivers http://www.guru3d.com/news-story/amd-ryzen-cpus-to-support-windows-7-with-drivers.html

"Microsoft is not supporting Intel’s latest Kaby Lake range of CPUs on any OS below Windows 10, kind of weird right?"


Six Core AMD Ryzen Processors Might Still Possible http://www.guru3d.com/news-story/six-cores-ryzen-processors-possible.html

 

Ryzen 8 cores is probably not going to be as cheap as you think...
And the lack of quad channel is really questionable....

 

What basis are you using for that? So far, the rumors put the SR7 at $500. Even if $650 with good IPC, it is priced well. I'll agree, in the past, price rumors have been low, so I'm just wanting to know the basis.

As to quad channel, that is a ball buster, admittedly. But, if doing memory intensive work, you weigh the cost versus the hit to productivity. In 4 weeks, all will be revealed...

Sent from my SM-G900P using Tapatalk

 

If SR7 is as good as 6900k, it wouldnt be 500, probably 700-800 make sense. Might sell a 8c/8t around 500-600. Dump the 4c/8t at 300.

If you are AMD, why would you not sell it higher since it got the performance?

 

Because of marketshare and certain other features missing, such as 28-32 pcie lanes and dual instead of quad channel ram. Also, no mention of support for cross point, etc. We know that only effects certain people in the market that needs it, but good IPC isn't enough to get some. The second you cross over above $730-40, you're paying just under 75% for a chip without those features. Paying $250-300 more for that starts making sense. You have to undercut the price to the point it doesn't make sense to purchase the Intel chip over the AMD chip. Even in the $600 range, it makes sense to get the AMD chip.

Then, if you compare to the older AMD chip prices, you see a willingness to hit that threshold for price/performance. That is why I think you are stating higher than what I think, but agreeing the stated $500 for the flagship being potentially low.

Sent from my SM-G900P using Tapatalk

 

Forgot about fx9590? AMD can price outrageously as intel.

AMD is trying to brand itself differently than before with ryzen. SR7 wont be a product thats for price vs performance crowd.

 

I disagree, to an extent. Su has said they plan on targeting the enthusiast, high margin cross this round with both Ryzen and Vega, but that they are still devoted to price/performance. I agree with your point, as seen with the 295x2 (which rumors have it that a Vega successor is on its way). Even the fx9590 is a great point there. But it is a balance. This is why I think it will be lower than your prices, but higher than the $500 price point, especially if the rumored 4.0GHz boost is for all core performance. . . .

Sent from my SM-G900P using Tapatalk

 

The price must be correct, if AMD shall manage to steal a lot of customers from Intel. A better performance + slightly lower price isn't enough.

 

All my speculations are ofcourse just my opinions but I doubt the price will be as low as you think.

AMD can sell the top end part for 750 usd fine. It will likely be 10%-15% slower than 6900k clock per clock though so that is a good price. AMD will be competitive with intel at that price point.

Tbh though, people looking for top end chips are going to lean towards intel's 6900k/6950x anyways. What AMD does is competing against 6800k and 6850k.

 

I hear you, but they only have a year and a half to pay a large debt or refinance. Doing that won't take as much of the market, which means cuts into profits. This is why they cannot do that this round. They need everything to be flawless and to sell the most at the highest margin. $750 doesn't do that at that price, if competing with the 6800K and 6850K, people will buy those for $480-650. This isn't pricing for a normal year, they need a blowout success. That is why I disagree with your numbers. That would increase the margin, but you'd lose so many sales it isn't worth it!

Sent from my SM-G900P using Tapatalk

 

No, you use the 8 core/8 thread version for 500. Sell the top end at 750usd.

Its less about individual prices rather price segmentation. Something intel is very good at and AMD is bad at.

 

Why buy an 8-core/8-thread when you get 6-core/12-thread, quad channel memory, and more pcie lanes for $150 more? You have to remember about ancillary features. If you do $420-60 on the 8/8, then $640 on the 8/16, considering you are pricing to compete against the 6800K and 6850K, same price, two cores difference to directly offset the other features, etc., is why I chose that price point. Even then there is a risk of not capturing some gamers considering above 6/12 barely any games can benefit by the extra cores, so it comes down to comparative performance based on speed at that point. Also, at that price point, if the chip is good enough, no one would automatically recommend Intel if 8 core is needed. Note these prices are both over the prices estimated, which are $320 (or so) for a lower model (guessing 8/8) and $500 for the flagship. I do think those leave margin on the table, but if the margin is already large enough, getting people on the system for future releases is more important. Also, you can make up for lost margin in price by volume.

Considering those focused on history instead of facts of the chip, you may need additional value in lower pricing to get some on board...

Sent from my SM-G900P using Tapatalk

 

Eh.... If core per core performance is there, I highly doubt its would 500 for the full version. That pricing structures doesnt leave room for the 4c/8t version nor the 6c/12t version.

I dont see the 4c/8t version cheaper than 250-270. Probably settles at 300 at release.

 

So, I wanted to collect different estimates and put them up for comparison. So bare with me...

Forbes ( http://www.forbes.com/sites/antonyl...s-release-date-and-motherboards/#6458c397a954):

"
The top end 8-core parts, similar to Intel's $1,000 Core i7-6900K, are expected to retail between $580 and $720, I've seen rumors of 6-core parts with 12 threads, similar to Intel's $600 Core i7-6850K retailing for around $250 and the SR3 - potentially a 4-core part with 8 threads has been rumored to land at around $150, with Intel's current equivalent being the Core i5-7600K, which retails for $100 more.

The SR7 prices are a fairly concrete, but for the rest, notice how many times I said rumor - they will likely cost a little more, but the hope is that at the very least, they'll offer an alternative to Intel - personally I'm hoping for some genuine competition and lower price points for the same performance."

Christianpost ( http://www.christianpost.com/news/a...ch-6-core-variant-spotted-in-new-leak-174056/):

"The AMD Ryzen eight-core CPUs are expected to come with a price range of about $580 to $720. Meanwhile, the six-core variants could cost about $250."

"According to WCCFTech, there are at least four versions of the AMD Ryzen that have been entered into the benchmark database.

One AMD Ryzen eight-core model with the label "ZD3406BAM88F4_38/34_Y" has 16 threads of simultaneous multithreading (SMT) with a base clock speed of 3.4 GHz and a boost clock speed of 3.8 GHz. It features 16 MB of level 3 (L3) cache and 4 MB of level 2 (L2) cache. Furthermore, the CPU has a thermal design power (TDP) of 95 watts and supports overclocking via an unlocked multiplier.

Meanwhile, another eight-core variant was seen earlier this year during the Consumer Electronics Show (CES). This specific CPU has a "1D3601A2M88F3_39/36_N" codename and features a 3.6 GHz base clock speed and a 3.9 GHz boost clock speed. There is also a third eight-core model, "ZD3601BAM88F4_40/36_Y," with a 3.6 GHz base clock speed and a 4.0 GHz boost clock speed.

Finally, one AMD Ryzen six-core model with the label "ZD3301BBM6IF4_37/33_Y" was also spotted with 12 SMT threads, 12 MB of L3 cache, 3.3 GHz base clock and 3.7 GHz boost clock speeds and support for overclocking."

PCGamesN ( https://www.pcgamesn.com/amd/amd-zen-release-date-specs-prices-rumours/#price):

"Current rumours have the Zen processors arriving in three distinct tiers - SR 7, SR 5 and SR 3 - following Intel's Core i7, i5, i3 structure. The top SR (Summit Ridge, innit?) tier will be the eight-core, 16-thread CPUs, and we're being told to expect the highest clocked version to retail for around $500 with a slightly slower octo-core costing around the $350 mark."

"I’d also expect to see similar pricing between the Core i5-6600K and the quad-core, eight-thread variant of the AMD Summit Ridge range. At the moment that would mean a price of around $250 (£215). It looks like AMD will again be betting on an increased thread count to tip the balance against Intel chips which will still have a higher straightline speed."

PC Advisor UK ( http://www.pcadvisor.co.uk/new-prod...or-release-date-price-specs-features-3643552/):

"However, we'll stick our neck out and predict a few price points:

• Ryzen Quad Core: £200 inc VAT
• Ryzen Six Core: £300 inc VAT
• Ryzen Eight Core: £400-600 inc VAT
• APU Ryzen Quad Core: £250 inc VAT

We predict Ryzen coming in four initial models. However, we do see a dual-core APU (£40), a 16-core (£250) and even a 32-core server-type (£400) processor being released too. These predictions are purely based on previous generation models and based upon its competition - Intel.

As soon as we get more concrete prices, we'll update this article."

So, your prices are the high end, but both of our estimates are around the prices many are expecting.

 

I would not pay 300 for a 4c / 8t version. 300 for a chip that performs the same as Intel's Core i5-6700K, Core i5-7600, and Core i5-7500 is a big no. The same, would others say What is the purpose to skip Intel for AMD, if you do not get anything back?
Read more that http://hothardware.com/news/amd-ryzen-processor-lineup-leaked-17-skus#kjlWMCz1wihtHBjk.99

 

Because is 50 dollars cheaper and you can upgrade to a 8 cores. If its on par with the intel quad cores i7.

And where are you getting that the 4c/8t version is on par with the I5s?

I mean, the website mentions a i5-6700k..... its questionable.

Also if a 8c/16t is competing with 6900k and another one is competing with a 7700k, I have a feeling it wont do well against a 6900k@ 4.2ghz.

It could be AMD pulling a fx9590 and taking all the OC room.