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

Yep, living in the mountains, we got snow while trees already have their leaves. So it wreaked havoc!

Sent from my SM-G900P using Tapatalk

 

Toms hardware tests the Ryzen 5.

http://www.tomshardware.com/reviews/amd-ryzen-5-1500x-cpu,5025.html

 

So, SA put out the cliff notes on the AMD earnings report. I'll dig into the 10-Q after I get home internet back:
Net operating loss declined to $29M from a year-ago loss of $68M; net loss overall narrowed to $73M from a year-ago loss of $109M. On a non-GAAP basis, operating loss fell to $6M from $55M, and net loss to $38M from $96M.
Strong demand for Ryzen CPUs as well as the company's graphics processors drove an 18% increase in revenue, says CEO Lisa Su.
Revenue by segment: Computing and Graphics, $593M (up 28.9%); Enterprise, Embedded and Semi-custom, $391M (up 5.1%).
For Q2 it sees a 14-20% increase in revenues from Q1; at the midpoint, that would be 12% gains Y/Y.

Sent from my SM-G900P using Tapatalk

 

On that note the stocks took a big dive today.

 

People are stupid. Basically, the losses first quarter are normal and they meet expectations. Further, Ryzen was sold primarily in only one of three months. But, this shouldn't last. It is also in part to leaked Vega benchmarks and them not in context. So, to stabilize, AMD needs to leak something comparable to the Ti/Titan in the next week. Either that or more on Thread Ripper.

Sent from my SM-G900P using Tapatalk

 

Natural cycle.

They should rise after Q2 finishes.

 

My look at it is it might now be a huge buy!

 

This isn't a dive; it is a market correction.

This isn't a time to get a 'deal' on the stocks; unless you're planning to hold them till retirement (and you've just cashed your first pay cheque).

Stoking the marketing/BS machine won't help here in the long run either - what is needed is for AMD to let their products speak for themselves (when they're actually delivered). If you brag about yourself; that's all you are - a braggart. If others brag about you; then you're actually something worth talking about.

Stock prices isn't how I rate a (tech) company (because those prices can be manipulated...) - what they actually deliver in my workflows/workloads is what counts.

AMD made some huge promises with Ryzen. They even surpassed many of those promises. What they didn't do though was grab the brass ring and be declared the (overall) winner.

Is this market correct a reaction from their biggest fans (i.e. those that actually own their stock)? Or, is this a market manipulation from ??? to slingshot their stock prices even higher than they were?

Either way; it won't be the last correction/manipulation in the near or long term for AMD.

Unless they just deliver (products) first and then talk/discuss later.

 

It isn't exactly a market correction as it is paying on passions, not fundamentals (going with the speculation/manipulation you mentioned).

As for letting others report results, what do you think a leak is? You work it out to give an exclusive in a roundabout way to third parties that are impartial for their "reputation." Also, allowing benches to appear in online benchmark database is even used by Intel. They did it not too long ago with the Geekbench leak.

But, there is a double standard, with some reason. Any leak by Intel, and knees to the sky. Any leak by AMD, you either hear it isn't good enough, or if it is, it was cherry picked. As I said, some historical basis for it, but there is a reason Intel says you get 15% per generation when most independent testing shows 4-9% per generation. But I digress.

In any case, this will have its rides and should be ridden longer term, so long as you believe AMD will meet its debt obligations, as you said. Buying in the dips can allow price mitigation if you bought at a higher price. Still, buying when they were at $2 would have been the best.

The IP is there. The counted blockbusters (Naples, Vega, and potentially ThreadRipper), along with a full quarter of Ryzen without a MB shortage will help. I need to read the 10-Q this weekend to examine what the information related to inventory is. If the spending was building up Ryzen inventory, ok. Vega and Naples, that's wonderful. Carizo= problem. 400 series that cannot be switched over would be a problem. So, without internet, I cannot competently speak on it other than the suspicion it was overblown and arbitrage is happening.

But, one reason to examine securities information is 8-K press releases! The other filings share how it is governed and you can glean some product development.

So, I agree in part and disagree in part, but am saying that the earnings and the Vega benchmark rumor is what is driving this, in combination. Those reporting the Vega rumor are presenting it as the flagship being only as powerful as a 1070. No discussion of the 7 Vega skus in recent drivers, that this used a generic VGA driver, not the AMD driver, or they might mention little Vega (Vega 11) at the end of the article, with others suggesting outright that this card is big Vega and their flagship (Vega 10, which a company exec recently started compared well to the Ti and Titan, which are way more powerful than the 1070).

But, to be fair, even Intel and Nvidia took some lumps after their earnings reports, as did other tech companies. So...

Sent from my SM-G900P using Tapatalk

 

It is a market correction because it rose so high because of the 'playing on passions' earlier.

Letting others report actual results is one thing. Saying that they have 'worlds best performance' and other such $%#@ is where I have a problem.

There is no double standard with the info I currently have; Intel states something and actually delivers it. Many reviewers have stated this over and over; '...but yeah; they hit (and sometimes surpass...) their clearly stated goals'. That is why there is real enthusiasm (from me) when Intel announces something; if they deliver; they'll deliver at least what was expected... if not more. (i.e. they let their products do the talking).

AMD needs to create some fanfare, I get it. But 'some' turns closer to a 'circus', seemingly every time with them.

All the scenarios you state that may be indicated by perusal of the financials won't mean a thing; their next P&L statement is the only thing that counts. The bottom line.

Giving a good talk only goes so far. What is delivered and actually measurable in $$$$$$$ (and not less $$$$$$) is what counts.

Is there reason to hope for AMD? Sure. But let it build on actual successes; not on what is supposed to be coming down the road.

 

Intel, back in the AMD being competitive days, was infamous for vaporware. Eventually the items may show up delivering as promised but sometimes before they got to show up they were outclassed by something else. So I do not go by the announced future releases from anyone as you are insane if you go by just them.

 

1) if you are reverting to being a fanboy, leave.

2) Intel regularly over promises and under delivers. They regularly state 15% improvement between generations, but in real life uses, it is 4-9%. That is fact, deal with it.

3) market correction happen when a catalyst (some form of information) causes the market value to change to come in line with another value. There are many other values that can be used. Book value is one. Private market value (which is very similar; the value of all assets if sold in private market sales) is another. You can do a present discount valuation relative to some of these figures, etc. Now, let's look at the information present. When the information in the release (10-Q which contains the earnings along with other required financial disclosures) was there, you saw a 7% drop. Considering Intel's, that made sense. But, 26% in a single day, that went beyond what every analyst valued the company at, with almost all having a hold rating, a few with buy, some with sell. This means the market is not compounding information rationally in a market correction (which, in 2014, even the Supreme Court adopted in Halliburton II that stocks may not trade on recurrent systems, as shown by experts in the field). So your theory on happening doesn't work. That means an examination of what information that has been made public between that point and now and how the market participants react, with the acknowledgement of over-corrections being possible by late buyers that got spooked by the behavior of others, which isn't rational or disciplined, it is an emotional, visceral response!

So, the information available is Ryzen and Polaris gave an 18% growth and reduced a normal first quarter loss. Second, many companies have first quarter losses, from writing down last year's inventory, investing in new inventory, often released second and third quarters, investing in promotional material, losing margin in this quarter by lowering costs to clear inventory channels to prevent compete write-offs, etc. These are normal things. So, to increase margin at all in the first quarter is good news. Analysts predicted it, AMD did it.

Ryzen, which sold well, was sold only 1/3d of the first quarter. As such, you see some gains, but not the full impact. It also was used to offset clearing inventory of Carrizo. This was the company's stated reason to wait until March, stated either around CES or the December Horizon Event.

They have many new releases in the second quarter, including Ryzen 5 and the 500 Polaris series. This means the stories of inventory buildup cited by many analysts may be overblown. By knowing and tracking where the money is, new products that will sell versus old products that are marked down or written off altogether due to aging inventory, you have a better idea of how that effects future earnings. So it is extremely important. Once again, after I'm able to pour through some analyst reports and the 10-Q, I'll know more. But this was one point highlighted.

4) The other point highlighted on the negative is going from one to three articles improperly discussing upcoming graphics cards (already detailed, will not rehash).

So, when taken altogether, even if an over-correction, market irrationality, either by compounding improper information or by reacting emotionally to a sell off to go beyond valuation in a correction event, is present.

Further, if you are going to tout puffery by Intel, which regularly over-promises and under-delivers, but until recently had no competition, which means could do so with impunity, you should leave. As I said, rationally discussing all information, including nuance, is fine. The second you started saying Intel always delivers or over-delivers, you've shown you don't have as much understanding of their business our advertising as you say you do.

I concede Intel has, in recent history, delivered better products to AMD, but in some ways, BW-E sucked HW-E's nuts. Skylake didn't hit 15% promised. Kaby didn't hit 15% over skylake. This is all empirical proof of Intel over-promising and under-delivering. So, please, speak nuance, or stop the fanboy BS in an AMD page.

Sent from my SM-G900P using Tapatalk

 

Woah shots fired, anyone want some popcorn?

 

Thanks I'm ready

 

This isn't an 'AMD' page. Sigh... I thought we're discussing hardware here?

I'm not now and never have been a 'fanboy' of anything but actual performance.

For your point #2; where is your proof that what Intel promised, they didn't deliver? The reviews I've seen show the opposite, almost every single time.

The market correction doesn't have to line up to your expectations or anyone else's... it is a fact (i.e. it is now in history...). Whether is stays there or not is what is more pertinent. Not how far it fell.

Whether it is rational to you or not.

I'm did not state Intel always delivers or always over delivers. I stated that many reviewers state that Intel met their targets...

I'm not emotionally invested in this; you are.

I'm discussing things logically, I invite you to do the same.

Just because Intel does some things better than AMD (and I point it out) doesn't mean you need to go on the defense for them...

My main point once again is that a Market Correction doesn't depend on analysts to predict correctly - it is in the hands of the market, right... And the market has spoken. That is a fact.

 

Yeah, Intel delivered higher clocks from Skylake to Kaby. Nice improvement Same for Haswell mobile... 4930mx up to 4940mx = +1 bin but worse overclocker. Yeah very nice improvement by Intel But were was the IPC improvement? Maybe they forgot?

 

I think I've stated this before: I don't upgrade CPU's. I upgrade platforms.

Especially with support for Optane and especially on the desktop side (for now...); Kaby Lake is more than the sum of it's parts. IPC be damned.

With a 10x improvement (performance/watt) since gen 1,

See:
https://beebom.com/intel-kaby-lake-vs-skylake/


And more features added every gen... comparing older platforms with irrelevant IPC 'scores' is the only way older CPU's will show a 'win'.

The complete platform is what is required to fairly test where we were and how far we've come.

Not that any and all upgrades are 'required', btw. But when an upgrade for your current platform is needed; the latest platform is usually the best bet (long term).

And for the ones that state that Intel is charging too much? Yeah; I want it cheaper too - but it still costs a cool $10K minimum today just like it did back in 199x... the difference is that $10K today is worth much less than $10K from circa 20+ years ago... (like about half...).

See:
https://data.bls.gov/cgi-bin/cpicalc.pl?cost1=10000&year1=199001&year2=201703

 

That isn't how market theory works! Just because the market acts in a way doesn't make it rational or efficient. Because of it reacting quickly to the news, you could not argue against efficiency in court. You could argue with an event study as to which news or event and statements the market reacted to on this occasion (which is my entire point of showing articles relevant to margin rates, inventory, Vega leaks, etc., to narrow possible emotional reactions).

Market value is divorced from the true value of the company. The IP portfolio alone in some companies is worth more than the market valuation. If the market correction results in not bringing the company into relation with the intrinsic value, it is acting irrational. Now, I've already said some analysts put the valuation at lower than it was pre-sell off. But, valuations were higher than post-sell off. This means over-correction and irrational behavior on some market participants. This isn't debate. This isn't the market speaking. This is empirical fact. Now, valuations can disagree. But the secondary market is not the best predictor of value. That is why you invest at all. If it was called correctly, you have little ability to capture anything on the correction.

Next, you said Intel puffs then Intel delivers. It was a blanket statement. I mentioned a couple ways it was false. Papusan did as well. Awhile back, one of the videos mentioned Kaby was like 4% over skylake. This is by reviewers, meaning you are wrong.

Finally, can you read the thread title?

Sent from my SM-G900P using Tapatalk

 

Sigh; it's called market 'theory' because that all it is. Can't predict every aspect that will/can/might affect the company...

The market itself is the final indicator of 'worth'. Regardless of how much intrinsic worth a company has 'on paper'.

I am telling you it wasn't a blanket statement; you keep telling me how to interpret what I've said. You're wrong here, okay?

I already responded to Papusan; please read it. A video that states KL is 4% over Skylake doesn't prove anything as that isn't what is important, in the end. Even a video by 'reviewers'.

Yeah; I can read. The thread title doesn't mean that you discuss the main topic in isolation. Other entities are at play here too.

 

I see that you are not very interested in IPC improvement. Aka pure performance improvement. But more interested in increased iGPU performance. Oh'well, You about that More and more high performing laptops skipping Optimus(increased battery performance) (not the underpowered BGABOOKS intended for email-web). No gains or any profit there either. And can we use and have big gain from OPTANE now for high performing laptops? And native support for Thunderbolt 3.0 what a big gain vs. Alpine Ridge Thunderbolt Controllers... I would call what Intel deliver... Minimum. Absolute minimum. Only MILKING. Same as NVIDIA with the new underpowered FRANKENSTEIN trash, All is about...


Hope AMD can push Intel one step further... Without AMD. Same as before. Absolutely MINIMUM from the BLUE Camp!!
@ajc9988 Clock for clock improvement between Kaby and Skylake is Zero. Infact Skylake score a bit higher in bench with same clock speed.

 

I'm getting more tempted by the day. Going to have to replace the mobo/CPU/RAM together if one fails at this point anyway, and that 1600 is looking nice.

 

Remember, check out reviews of all MBs on overclocking. If you want reference clock OC, there are very few to choose from. Getting the fastest ram possible (either by buying good OCing ram or highest stock settings) will help on performance! If you can wait for the AGESA update this month, which will open up ram overclocking a lot more, I would as that could affect board choice on which does best. Most support B-die Samsung best for overclocking, which a lot of G.Skill ram are that. So, if you can, and this build will have to last you awhile, wait a month. . . .

The data on which boards support what the best is mostly settled, but the unlocking of the ram with the AGESA update could change the field.

Sent from my SM-G900P using Tapatalk

 

http://www.tweaktown.com/news/57366/amd-radeon-rx-vega-specs-leaked-shy-monster/index.html

More on Vega:
http://www.tweaktown.com/news/57366/amd-radeon-rx-vega-specs-leaked-shy-monster/index.html

Sent from my SM-G900P using Tapatalk

 

You seem to think that IPC is the only improvement that can be gained; it's not. Hasn't been for a long time now.

But higher performance? Yeah; it's there in spades if you buy it with the right platform and let it run as it's intended. (Instead of artificially limiting it's clocks and saying it's 'slower', when it's not).

See:
http://www.pcworld.com/article/3152...one-big-change-makes-up-for-smaller-ones.html

Kaby Lake processors switch between states faster than anything previously; this will be huge when Optane is widely available.

Combine all the above into a single platform, and I really can't see the reason for the crying?

 

You continue the exhibition about Optane Ready, slightly more powerful graphics. + better battery performance due to iGPU. What more? Nothing. Oh I forgot, Yeah slightly higer overclock

Remember a lot of the latops with Kaby Lake haven't Optane support!! And you know it as well, Almost nothing out there if you want Optane drives for Mobile laptops. Aka all too early and too expencive for ordinary consumers. Maybe next year.

Optimus has become less used (forget about the bragged increased battery performance in Gaming lapbook series - aka the most powerful models).

Yes you can overclock higher with Kaby and this is very GOOD... But this is all. And nothing more.


No Battery gain due Optimus is almost not in use anymore for the most powerful laptops with Nvidia graphics.


The conclusion about 45W TDP Kaby BGA chips vs. former Skylake. Where is the 20% improvement in performance for 45wTDP Kaby Mobile i7? The article pcWorld referred to is ancient8from last year). slightly higher performance was with increased clock speed. This is quite logical



Only advantage I can see is overclocking!!! Not the extra features. At least not for laptops. And Yeah, I can't see the big improvement from Intel. Almost none. In other words... Almost none laptops with Kaby have Optane support, Almost none with battery improvement due no Optimus. And almost no increased performance!!! Yeees, Intel deliver as always. Only AMD can change it. Aka help Intel on track again

 

OK, time to put this off topic to an end! Next step is too strip this thread and go further if need be.

 

New CPU-Z Upgrade Lowers Ryzen performance - Guru3d.com

"Two weeks ago CPUID released CPU-Z 1.79 (download here). If you already grabbed the new version you will have noticed it contains an updated and revised benchmark. That new benchmark shows lower Ryzen scores, actually lowers AMD scores overall. It's a bit of weird move to make for any developer. In the information below the break you can see an explanation as to why an how CPUID deems this as a qualified move to make."

 

Beat me too it, quick draw @Papusan

 

See:
http://www.tomshardware.com/news/amd-stock-slump-ryzen-cpu,34318.html


Some possible 'logical' reasons why share price dropped and also why AMD toned down it's Q2 expectations (even if it's just 1%...).

 

I try to balance between not waiting too long and not jumping the gun on these sort of things. Doesn't always work, but in this case I'm waiting for a bit.

 

This lines up with my general observations of all the new models I've seen in action here, around 10% overall increase (with big margin for error) and GPU ultimately more important in determining system performance.

 

One is logical, one isn't, and overall, neither is the reason for the change in guidance, most likely.

First, the iGPU. The APUs are waiting for HBM2 wide availability, so they are releasing the mobile and APU lines in the second half. This has been known forever and anyone applying such a sophomoric analysis and basing the second quarter margin revision on it is ignorant of all factors involved. Simply, I've heard it a couple times now, it ignores what is coming and other factors, showing ignorance. The majority of chips AMD targeted from Intel, the Intel users use discrete graphics cards. So it is moot.

The argument of per core cost and cannibalizing sales, recommending going to be like Intel is a mixed bag. If you need the best OC, the top offering still does better. But, the 1700 is the best value as most can hit 3.9-4GHz.

But, that isn't the likely reason for the margin revision. Nvidia boxed AMD in on pricing by moving the 1080 to $500 and the Ti to $700. This means they must price the two competing cards at equal or less, unless it way outperforms.

Further, you have the upcoming server fight. This means they are going to use every tool in their bag. How much is sold in second versus third quarter AMD release push up because Intel pushed release up to June puts a lot in flux, where an August release would have been in third quarter sales. So, there could be nuance there in the battle.

Finally, you want to address the lower margin Ryzen 5 and 500 series graphics products released at the beginning of this quarter. The 1600X could take some of the 8-core sales, but that gets back to the one I acknowledged already as legitimate.

So, taken together, it makes sense. It is not unreasonable. And there are factors unknown. What matters most is getting the most market share in the server market as possible! Second is Vega doing well. If margin goes down, but market share goes up considerably for graphics and/or server market share, it is made up for in full, and then some. Hence, the market is not logical, rather acting on emotion.

Edit: also, skylake on average hitting 4.8 and Kaby hitting 5.0, color me unimpressed.

Sent from my SM-G900P using Tapatalk

 

The point is Intel's official statements say 15%. Your experience falls 33% short of that. That is not delivering on promised gains.

Sent from my SM-G900P using Tapatalk

 

The way it sounds is that people didn't consider it until just recently because of all the top end performance hype. Seems atypical of investors, but it's a short term issue as you said, perhaps a rebound will happen when HMB2 drops in price, or at least when there's a definite deadline for getting it into mass market chips/systems.

 

Totally agree. I didn't mean to sound like I was arguing, just confirming.

 

So, two flavors are out on HBM2 currently because of yields, the second having lower speeds. But, wide availability is set for the second half. This makes sense on why Vega would hit toward the end of the second quarter, eating up early chips and getting the market first. You do the cheaper APUs when wider availability occurs in the second half. Because AMD helped create it, I know they had first access on HBM1. I don't know if HBM2 also had that. But, this is why Nvidia waited.

Now, Nvidia will either adopt HBM2 for Volta, or it may stick with GDDR6 if available early enough in the year. The reason I speculate that is a slide showing a benefit of GDDR6 over HBM on future energy use at an Nvidia event.

Sent from my SM-G900P using Tapatalk

 

Sorry, I didn't mean the sharper tone for you. I was using your experience as an exemplar of common experience with Kaby falling far short for tiller. Was not meaning to come off as arguing with you at all.

Sent from my SM-G900P using Tapatalk

 

What I've read so far tentatively points to NVidia sticking with GDDR6 at least for one gen going forward, but nothing solid.

 

For those interested, it's SK Hynix saying they will mass produce it for a high end GPU in 2018, so probably Volta.

 

Last warning, and I do mean the last one!

 

Thank you!

Sent from my SM-G900P using Tapatalk

 

People discuss non Ryzen issues else where. Do not get me wrong. I am the first to be glad to see it is time to challenge the varied manufacturers and their products and claims, it just belongs in other threads. Especially before the escalation that was about to happen.

 

TANWare, thank you for clarifying what the issue(s) were with the 'off topic' posts. Noted.

Be certain though that there was no 'escalation' about to happen though (at least not from my side).

Take care.

 

New AMD Ryzen Microcode Update Will Expand Memory Support - wccftech.com

 

They listed my 4133 Trident Z on that list!!! Plus, that puts a huge speed increase on infinity fabric! Now, just need to release threadripper and I'll be in business!!!

BTW, internet is back at the house, meaning I'm back baby!!! Yet @Papusan still ninjas me! But I got to see guardians of the galaxy 2 this afternoon, so all is right with the world!

Sent from my SM-G900P using Tapatalk

 

I think the XMP speeds are still of issue on sure fast rates. I think stable results so far are in the 3600MHz range. Maybe with the new microcode we will be able to get higher.

 

That's what this AGESA update is about. It was originally supposed to be version 1.0.0.5, but this article says it is 1.0.0.6 and adds support for 20 models of ram, providing a list. Now, it is important to note, until the motherboard manufacturers put out an updated BIOS with the new AGESA code updating the microcode that will add this support, these will not be supported. Further, as the article noted, the SoC voltage may need increased to run the speeds marked, or the ram voltages increased to as much as 1.5V to run their marked speeds with this update. What isn't said is the variability of supporting speeds by motherboard. Even with the update, that is still an unknown factor we want to wait to find out.

But you are absolutely correct for today. Yet, this is the update I've posted about for months finally coming home!

Sent from my SM-G900P using Tapatalk

 

This some of what I have been waiting for, I would love to see a 16c,32t running 4200MHz. I am sure it will set me back a few grand but I almost get a Woody thinking about it. You know from ToyStory, Where else do you think?

Edit: 16c,32t

 

LMFAO!!! Sounds like we are looking at similar rigs! I have 4X8GB 4133 G.Skill B-Die I've already ran dual channel with my 6700K at 4000 with [email protected], so if it can do anything near that in quad channel with a CL16, I'll be doing just fine. My board could not run 4133, so I couldn't test what I could do with rated speed and dropping the CL from 19. I run 3733 at CL14, so I'm sure this kit will be fine. My kit is also the older GTZKW instead of the newer listed, but I don't think that will be much of an issue (some looser timings). So, it won't be long before the real fun begins and I setup a house work horse and just get developer boards able to run steam and video players while offloading all processing onto that beast (including games using steam's ability to have another computer do the work, while just hooking up a fast home network) for around the house (I have 4 4k TVs in 3 rooms, but one is used as a monitor while the other is a TV in one room).

 

I do not have a job for it, I just want it!