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

Not that I'm a proponent of massive corporations, but regardless, Intel did quite a lot of damage with its 'tactics' which gave it both the market and mind share.
The aftermath of their dealings with OEM's damaged AMD's ability to get into the mobile space among other things, and OEM's intentionally crippled all AMD laptops and offering little to no AMD systems in general to consumers - the aftermath of which was felt all the way until Ryzen+.

As for Intel being run by different people... how different?
They've messed up grossly with 10nm what with continuous delays, have been over-charging for 4 cores instead of advancing the CPU's in core counts and dropping costs, provided minimal IPC increases while forcing people to change motherboards with almost every CPU revision.
To top it off, they demo-ed a 5GhZ Xeon while conveniently omitting the fact they needed an industrial chiller (which is a massive power hog and can't be legally bought anywhere) to pull it off, and that server part is nowhere to be seen (at least, I cannot certainly find it anywhere) and all their parts are susceptible to security vulnerabilities (which were seemingly patched up yes, but this in turn ended up dropping Intel's performance by a large amount - 5x more vs AMD... and AMD is not even susceptible to 95% of those security issues).

As for users supporting Intel with their wallets coupled with 'best' or 'brand names'... this is an issue of people not really educating themselves and falling for marketing tricks.

Why should I give Intel the 'benefit of the doubt' when they demonstrated time and again they would mess people over?

Now, I'm not deluded to think AMD cannot behave the same nor that it doesn't have flaws.
They certainly could behave like that and do have flaws... but their actions say something different.
AMD supports Open Source (which is a lot easier to use and benefits everyone - and its probably keeping costs down while at the same time doing new things), they provide long term chipset support so you can only replace the CPU and not the whole motherboard on top of that - keeping your overall costs down (though the laptop OEM's are certainly stingy and completely unfair for not releasing BIOS updates to support future CPU upgrades like they do for desktops - but this has little/nothing to do with AMD itself), they also look towards future by implementing new technologies and were the first to come out x64, not to mention to give commercial/mainstream users 8 cores and above at an affordable price no less - at the same time, they don't compromise on security (which, granted, may not be too much of an issues for mainstream users, but it certainly helps both us and the data centers).

On top of that AMD offers not just cheaper hardware, but one that's competitive in terms of performance while being similar or better at efficiency (especially when undervolted) so I can both game and do content creation.

So, yes, taking everything into account, I do think Intel needs a 'time out', and emotions play little to no part in this situation given the options available (but can you honestly say that you'd be OK with a corporation exploiting you and your ignorance and continue to get their products? Especially when there are better alternatives on the market?).

 

This is why I said AMD did something I didn't think they could do, make the 9900K pricing seem "reasonable," especially when it is not. They priced it such that that chip can still be sold in certain situations.

As to performance lead, they are single digits on ST performance. How is that a "lead"? Especially when at the price point stance, MT they don't have ANYTHING.

Go back and read my comment on their pricing right after announcing the 3950X (I may have said something with the 3900X also, but I made a clear point with E3).

Meanwhile, although you have a point on corps being ran by people with fiduciary duties to maximize shareholder value, you are WRONG on the competition part. Take, for example, AMDs pricing on the 5700X. They put a product out that is competitive with Nvidia's product. Prices are NOT coming down! That means competition only amounts to choice, not value pricing. That BS capitalist argument of "competition bringing down prices" has been disproved COUNTLESS times. Just like privatization bringing down costs has been disproved, which is why corporations LITERALLY DIRECTLY STATED that they could not compete with a public option, which is non-profit.

For tech tubers, I watch GN, Hardware Unboxed, Level1Techs, JayzTwoCentz, and TechYesCity for news and reviews. I watch RedGamingTech for the rumor mill, along with a little of Moore'slawisdead for speculation.

For websites, WCCFtech for speculation, along with a couple others. Anandtech, EETimes, Digitimes, Wikichip, Extremetech, semiengineering, and a couple others for my deep dives and industry news. I stopped reading Tom's Hardware a LONG time ago, although there are times where they are the only one's that covered certain things. PCWorld is OK, but not my cup of tea and I have other sites I do like better (you've seen where all I quote from).

Definitely the pricing is out of hand. It will take collective action by consumers to stop it. But, if there are enough people willing to pay the inflated prices so that the loss of sales to others is made up for with the higher margin, all while reducing costs because you don't have to produce as much volume, you will continue to see the trend in pricing.

 

The Asus Zephyrus GA502 uses the AMD 3750H Mobile Ryzen CPU step up from their 3700U series,for entry level gaming in mobile. The CPU is 4c/8t and the integrated GPU is 10 CU, with the discrete Nvidia 1660ti 6GB. It gets a better rating against the Intel competition, whereby Linus goes into where Intel screwed up allowing AMD to compete so effectively against Intel at this price point.

This AMD 3750H 4c/8t CPU isn't going to "win any drag races" - skip to 01:50 to avoid that silly "race" - with higher end CPU's, but it doesn't need to for good gaming results with it's 1660ti 6GB, and Linus recommends it over the Intel CPU's matched with 1660ti's costing more.

AMD Is Crushing Intel in Laptops Too - ASUS Zephyrus GA502 Review
Linus Tech Tips
Published on Jun 17, 2019


Here is the ASRock Taichi x570 board breakdown, with details on the build, power, cooling, and observations on the features. No price, running test - Ryzen 3 or otherwise - yet.

With only 1x 8-pin + 1x 4-pin motherboard extra power as compared to Intel motherboards with 2x 8-pin perhaps even the 16 core isn't going to need as much power as Intel CPU's?

There isn't a huge VRM setup for the chipset power, which makes sense in that it's only 11w(-15w?) maximum power draw - the chipset fan should run slow and quiet.

ASRock X570 Taichi Motherboard Analysis of VRM & PCB
Gamers Nexus
Published on Jun 17, 2019
Buildzoid looks at the ASRock X570 Taichi motherboard in our ongoing series of X570 motherboard round-ups. Cheaper boards are next! Find our Gigabyte X570 Xtreme comparison here: https://www.youtube.com/watch?v=qszxx...

 

I completely missed this quote, thank you for the correction. It's at the very least, some improvement, where I previously had to tell people to outright ignore VCE because it wasn't worth the hassle.

It does, however, have quiteee the ways to go:


They only mentioned FPS, though, rather than compression level, so if compression level is unchanged and it simply is... more bandwidth, then it's not a useful change (most places cannot display, and oft reject, above 60fps, so recording at 150 is not useful). The above picture is purely for comparison (and GPU was in light use).

 

I just saw this myself, and thought I'd check if the Area 51m guys have tried it yet. Not many notebooks with 9900k's in them, I guess the Clevo guys might have tried this too?

Not sure how much help it is, not integrated in to the CPU / drivers like AMD CPU "auto-maximizer", reminds me of the original Asus tools for tuning desktop motherboards / CPU's from long ago.

Anyone tried it?
http://forum.notebookreview.com/thr...-owners-lounge.826831/page-1301#post-10924398

 

I have to disagree on AMD crushing laptop offerings. While they are nice it would also be nice to have too much CPU. While I have a 2700u I can not complain on it as a daily driver for net and office applications. On the same note I would not have complained If it were 8 instead of four cores. GPU is irrelevant to me.

 

You'd be surprised of how relevant the GPU is... even for everyday tasks... but, I would agree that 6 to 8 core version of an AMD low power APU would have been nice.

I wouldn't have been surprised if AMD did that with Zen 2 actually... do a chiplet version of 25W TDP 8c/16th APU with Navi iGP.
Even if they kept the core clocks same as Zen1, you'd think they'd be able to stuff 8 cores into the thing and get better performance in the CPU and gpu segments.
Oh well, perhaps with 7nm+ EUV.

 

amd not even close in mobile segment. linus is probably just trying to grab consumer's attention and get more subs/vewis.

 

Sadly I agree when it comes to the mobile segment. Until AMD supports undervolting, I'll be sticking with Intel for the mobile market...

 

He was being specific in his praise for the new AMD mobile CPU being priced better with adequate performance to drive a low end GPU for that particular market.

AMD is just starting out again to cover the entire range of laptops from U through H CPU's, with desktop CPU's at the top end.

I do hope AMD adds voltage control for the mobile CPU's, but I wouldn't count on it, or hold off on buying them until they do, it's irrelevant if the design and tuning of the laptop cools the CPU adequately.

Intel is the worst offender of overvolting CPU's in firmware out of the box and most laptops with Intel CPU's can not only benefit from undervolting but many need it to avoid thermal throttling without needing a repaste, I'm not sure that AMD mobile CPU's are in the same dire need as Intel CPU's for undervolting.

It's definitely desirable to have voltage controls for the CPU even with the low voltage / power CPU's, but if it works well then it's not a problem. And, I don't necessarily consider prime95 100% all cores as "necessary" function in a low end laptop.

 

1) The cooling will never be good enough. This has been a problem long time and won't be any better the thinner the notebook being designed. The trend is moving to Applebook size and design. The cooling will suffer.

2) The voltage control (undervolt) is needed if you try to get most out of overclock if this is possible.

 

3) Most people just want to turn on their laptop and go to work or play using it, and don't want to spend time tuning it for "maximum performance"

There's something to be said for wasting the amazing amount of time we do to get that last % bit of performance, and most people call it a "waste" of time.

Most people don't put their laptops through sustained 100% usage, and just want to game at worst and browse the internet, play media, and edit documents on those low end laptops, and won't stress things enough for it to matter.

The lower power CPU's can fair better with small form factors and cooling.

AMD's 7nm CPU's are showing a marked decrease in TDP rating - down to 65w for the 8 core, and the other 7nm parts should be lower power too.

I know that doesn't matter if your personal laptop is overheating from your use - mine usually do if I try to do my normal work load on thin laptops, which is why I don't buy them, but most people won't notice.

Which is why AMD may never adapt their software / low power low end CPU's for voltage control. That's why I said I wouldn't wait for it.

Actually we may still see it for 7nm Zen 2 / Ryzen 3 APU's, as they trail the new process releases - the CPU's we are seeing now are the leading edge, then Threadripper / EPYC, then the APU's using the new process tech will arrive, which would be the ones you are hoping for with new Navi GPU cores too.

 

I have followed the ThrottleStop Guide forum from the beginning. Many with low specs notebooks is very keen to get most out of it and post a lot questions (help for overclock, Lower the Power consumption, Lower the temp etc). Sometimes they are even in majority and post a lot vs vs. those with more powerful notebooks (the so-called enthusiasts). But I can see what you mean

Neither forget that it has become more sanitized / popular with overclocking of notebook hardware the last 2-3 years. And without the right tools, it will quickly becomes more difficult to get the results you want. The trend is here to stay I think. But that said, as you say we are all different

 

While undervolted mobile CPU chips from AMD would be desirable (the desktop variants in laptops can be undervolted), if the units have adequate cooling (and OEMs seem to be doing that increasingly for AMD laptops), it wouldn't be an issue.

Ryzen in general seems to have better operational temperatures than Intel, and upcoming Apus from AMD will have a thermal solder as opposed to the paste like the first generation (which definitely helps things).

I think we need to insist from OEMs they improve their cooling methodology on all laptops.

 

Sadly my experiences with my 2500u were not good in the A485, it just ran hot and the fan spun all the time. Looking at the voltages (guestimations in HWINFO), it was easy to see that AMD's stock voltage table was way too much. Yes cooling needs to happen from the manufacturer, but there is always room for improvement when it comes to temps and performance.

While @hmscott and you are echoing each other's sentiments, @Papusan and myself are in agreement in that we want total control and the ability to squeeze every ounce of performance out of our laptops. I won't be buying an AMD laptop until they either release voltage control, or they finally release 7nm. Since they won't be releasing 7nm for a while, I'll be on Intel until then. It may be viewed as irreverent to some, but it certainly is relevant to me.

 

Oh I'm aware of how the OEMs fudged the cooling for 2500u and 2700u... But the cooling seems to be fine if that apu is combined with a dedicated gpu.

If the laptop has an apu on it's own, then yes the cooling will be shoddy at best (depending on the OEM)... But this isn't AMD fault or something that could necessarily be fixed with an undervolt.

It seems with the zen+ Apus though, they are using a thermal solder as opposed to the paste, and Asus may have learned their lesson and beefed up the cooling, but again, those zen+ Mobile chips are paired with a dgpu, so they would have beefed up cooling anyway.

I'm thinking that things may change now with zen+ Apus out there and of course zen 2 coming out.

We need to wait and see... But I'll reiterate that any cooling issues are down to OEMs, not AMD. Besides, amd Apus already have relatively tight voltages, so I'm not sure how much help undervolting would be.

 

An undervolt may not have fixed it, but it would have made it better. The 2500u is a bare die, so I'm not sure why you mention solder vs paste. I even tried liquid metal on it. Zen 2 isn't going to hit the mobile platform for a while sadly. Do you have any sources for AMD having tight voltages on their APUs? I'm just curious because my 1700x and 2600x could undervolt very well (and overclock without additional voltage). My 2500u voltages were pretty high, but I have no idea if I could have brought them down.

 

It all really comes down to having realistic expectations for what you are purchasing. Know your use case and what kind of load that puts on the laptop, and look for a laptop that can handle that use case and load.

A thin and light laptop with clearly too much power usage and thermal output for the cooling design is going to overheat during heavy use and extended gaming. Undervolting or even re-pasting can't help enough to solve a laptop dropping performance due to inadequate cooling.

Time after time, and without exception it's been proven that there are no magical solutions available. If you want high performance in a laptop get a laptop sized with the physical dimensions to hold the cooling system required.

The physical requirements for cooling a high thermal load can't be ignored. It is impossible to extract and vent the thermal energy of a full sized laptop at 100% performance when that performance is stuffed into a thin and light laptop frame.

Don't buy a thin and light laptop if you need a laptop for heavy use and/or extended gaming.

Even a low power CPU / GPU design that puts out too much thermal energy to dissipate into that small physical frame will fail to run without thermal throttling once it reaches the physical limits of the cooling system.

For Undervolting (CPU voltage tuning) to be 100% effective in stopping thermal throttling it requires that the laptop have a cooling system designed to dissipate 100% of the thermal load under the highest performance.

Undervolting -100mV to drop the 100% load temp down 10c is greatly useful in those designs that stop thermal throttling when a temperature drop of 10c is enough. When that isn't enough then re-pasting is the last hope, but even that won't be enough if the laptop design is inadequate to dump the heat as fast or faster than it is generated.

If you buy a thin and light laptop you can't expect it to perform the same as a physically larger laptop that can fit adequately sized cooling hardware in it. If the power / thermal output of the CPU / GPU exceeds the physical limitations of the thin and light cooling system, undervolting and re-pasting can't solve the problem completely.

AMD knows all of this, and AMD can advise makers of the power and thermal properties of their CPU / GPU, but it's up to the laptop maker to put those components in a system that can handle full performance cooling.

If a laptop isn't adequate to your needs, don't buy it. If the general class of thin and light laptop's is inadequate to your needs, don't try to force it, accept it.

Expecting the big gaming laptop experience from a small thin and light laptop isn't realistic. Thinking that if undervolting control is added that will solve the thermal problems inherent in a thin and light laptop is also unrealistic. Undervolting and re-pasting won't magically fix inadequate cooling hardware.

Even if AMD adds user available voltage control in the 7nm CPU's / APU's, expect that laptop makers will continue to build inadequately sized cooling systems, and you will need to wait for independent reviewers that do undervolting as part of their testing - so that you know for sure that the laptop you are interested in purchasing can be tuned to handle your intended use.

 

no throttlestop or adjusting voltage are small sacrifices to make when you can get a 16 core cpu for only $750 while having superior IPC than intel's counter part.

now if that 16 cores can't go upwards 4.6 with a decent voltage then im probably out, frequency is important too.

assuming icelake is as good as intel claimed, 10nm++ is where its at. 12 cores 5.3 ghz yeehhh

 

All depends on the wattage. Wattage = Heat. Higher Power consumption = Higher operational temperatures. You mean AMD chips will run much cooler with exactly same wattage?

The ODMs will always go for design (thinner the better is what they think all will have) before functionality. Always.

See here.

100C is'nt a flaw regarding what Dell tech say. The chips is designed to run at this temp... Dell's engineers have improved how their notebooks is able to run so the chips easly can run at that High temperatures http://forum.notebookreview.com/thr...otebookcheck-net.808983/page-14#post-10873404


Btw.
Intel Ice Lake benchmarked - i7 performs better with lower TDP than Ryzen 7 3750H-APU
By Hilbert Hagedoorn on: 06/20/2019 08:34 AM

While the tide is turning in favor of AMD on the processor front you can rest assured that Intel is not let that happen quietly. We've already talked about the Ice Lake a couple of times and as it so happens, a new series of benchmarks have leaked.

Yeah, won't fix inadequate cooling but you could be able to run with slightly higher average clock speed. Aka halfway crippled

 

So, just realized on the WCCFtech rumor, if the 64-core TR is true, it would have to be 7nm Zen 2. Why? Because no 64-core Epyc until now. Why I didn't realize that before, who knows (brain fart). So, here's hoping that such a thing is planned for those that need it (even if I'm squarely in the 16-32 core camp, which I don't need 32 cores, that would be benching with all that extra core power).

 

I'm confused.
AMD already demo-ed EPYC ROME 64c/128th, did it not?
And they did it with an ES for the most part... which came a LOT later than 32c EPYC... so it stood to reason that ROME 64c will be a Zen 2 part on 7nm.
I mean, they already pushed the envelope with 32 cores on 12nm, and that seemed like the 'max' for that particular node (while not blowing the TDP to smithereens)... and since AMD usually sticks to a single socket for its data center CPU's, isn't it logical that 64c would be 7nm Zen 2?

That said, I wouldn't be surprised to see an increase to 76 cores on 7nm+ given that this revision would allow 20% higher transistor density... but then again, that depends on how AMD executes things and how different of an uArch Zen 3 is (which is slated for 7nm+).

It was previously stated that Zen 3 will produce 'moderate' performance improvements with higher emphasis on power efficiency... but we don't know how much of a bump up in IPC we can expect from Zen 3... except for an increase of 10% in raw performance.

 

I hate pointing out my ignorance, but I suggested moving like Epyc chips that can clock well enough to TR to clear inventory, which is a completely false and idiotic statement on my part.

The 64C Rome was shown last Oct/Nov at the server event. That is after the current lineup which tops at 32-cores.

This is talking Threadripper 64-core rumored part, which the only way to cram those in on the die is Zen 2, which is 7nm.

Also, AMD DOES NOT only or usually stick to a single socket. They have tons of dual socket designs. Single socket high I/O is great for financials, though, which is what they tried to do with the 7371 speed optimized chips.

Moreover, 76 cores makes NO sense. Not divisible by 8. That would require a complete redesign that I do not see happening. Instead, Digitimes said that 7nm+ yields are equal to or better than 7nm at TSMC. That means AMD may be banking on that and wanting to switch TR3 to that as soon as possible.

Moreover, instead of adding cores, with the 20% higher density allowed on FOUR layers (that's right, only 4 layers get EUV, not the entire stack, at least until 5nm), you might use it to try to go wide, like Intel did, to increase IPC again (AMD also went wider on Zen 2 in many respects, and took a lot of space savings to increase cache this round, meaning widening elsewhere is a potential candidate for Zen 3).

But you have to remember, AMD originally planned for Zen 2 just to be a node shrink. Instead, they changed up the uarch and made great IPC gains. This was because they thought they would have a speed regression, which TSMC solved. So they kind of got a large performance kick, more than planned. Whether they then want to take the next step in power savings or not, the iso-power/iso-performance curve says that OCing would then be improved.

Also, there is always a chance that the 64-core is Zen 2 to be released alongside of Cascade-X. Then, use either Zen 2 or Zen 3 with the rest of the lineup at CES, which end of January seems to be the late end of their range, with early Q4 being the early end of the range.

Finally, did you see that Intel is cutting desktop prices up to 10-15%? https://segmentnext.com/2019/06/21/intel-cpu-prices-drop-counter-ryzen/

For 7nm+ "[t]here’s also a 10% performance uplift or 15% power efficiency increase." https://financentra.com/2019/05/26/tsmcs-7nm-euv-is-in-production-improves-performance-by-10/

 

So, interesting news if people want to repurpose their old Ryzen/TR chips while upgrading (instead of just selling). I saw the leaked pricing on a RedGamingTech video this morning for the upcoming Epyc chips and went searching to find if any companies have released a more powerful 1P server board for Epyc with VRM that would support a decent OC (which still would need to find out how others have OCed the server chips, but that is a different point).

Evidently Asrock has put out a line of AM4 and sTR4 server boards for AMDs mainstream and HEDT chips. Here is the STH article on them.
https://www.servethehome.com/asrock-rack-x470d4u-amd-ryzen-server-motherboards/

Asrock went with the X470 chipset rather than X570 due to potentially needing active cooling and this being in a server case. The X399 board also looks interesting, especially if you use add-in cards!




Now, with the "P" series pricing (single socket), if there is a suitable board with good VRM, and where I could OC CPU and use regular or udimm ram, I'd definitely jump on that 16-core 7302P for around $1125. Other notable mentions are the 32-core 7502P for $3123 and the 64-core 7702P for just under $6K.

Still, with consumer facing 1P server boards, along with spreading it down to sTR4 and AM4 sockets, this is getting interesting (as it also means AMD's chips can start competing against Intel's Xeon E series straight up with server I/O optimizations).

Here is Tyan's offerings:
https://www.servethehome.com/tyan-tomcat-sx-s8020-amd-ryzen-threadripper-server-motherboard/

The AM4 platforms are able to give nearly the benefits of the Epyc embedded series, but with massively easily purchased hardware on the market.

So, to those thinking of getting a new Zen 2 and selling off their old chips, you may have another option, like getting one of these boards and building a router/firewall, like with pfsense, building your own NAS, or building a home server for whatever you would like to use it for.

Just wanted to mention it with the upcoming Zen 2 chips (plus, if these boards do well, then more will start producing them, like maybe a SuperO board, Asus Workstation and Server boards for these chips, and more variety). It's also nice if you hate LED blinding lights and want to return to a simpler look from decades ago.

Just wanted to share some more info!

 

Okay @ajc9988 someone made a video doing proper performance/visual comparisons and called out the maximum GPU load situation among certain other things with the AMD press conference example, check it here. I would suggest actually watching the entire video all the way down to the end, he even goes into streaming quality based optimizations for x264 presets that he uses, and why NVENC looks better in streams even if not exactly as sharp in many cases.

He also mentions toward the end that over 20+ threads starts negatively scaling at 1080p or below in streaming quality, which 100% backs the statement I made before about a 24-thread CPU being incapable of smooth streaming out the box. I thought it was 16-20 threads but I didn't know for sure.

One thing I need to nitpick at him though is he DOES show a streaming settings shot near the end and he has fallen into the noob trap of a keyframe of 2, when it should be 1 for the faster scenes, for a less sharp overall image but more consistent quality/less blockiness, so the visual clarity might improve significantly if he were to re-do those tasks, due to the way keyframes and b-frames work with bitrate distribution.

And finally, my trademark reflex reaction:

WCCFTech ignore

 

I noticed the additional shadow artifacts on the GN stream with slow and it really threw me off as it made medium and fast look a bit better in their chosen ROTTR example. But, you also know I used a 50" calibrated display to watch it on. I'm still going to finish it, but he is boring me with the obvious discussion of Chroma subsampling, etc. (I've taken a deeper dive into examining the HDR-10, HDR-10+, HLG, and Dolby Vision implementations over the past couple months). Just makes me wonder when AMD and Nvidia will choke it up and put an HDMI 2.1 compatible port, even if they will need to do an update to allow it for full HDMI 2.1, for consumers. Specifically, in regards to Nvidia, figuring out how to do VRR over HDMI, a feature that AMD has figured out (but since the VESA standard defines over DP1.x for VRR, it takes a separate implementation to do over HDMI, even though HDMI has their spec and I focus more on ITU and SMPTE standards, which are used throughout the movie and broadcast industries, over VESA's implementation, especially in regards to HDR content, which BT.2084, BT.2390, BT.2100 have more weight in defining color space and spectral highlights, as well as clipping or roll-off for tone mapping, etc., related to the switch to PQ EOTF gamma over the older power 2.2 calculations or the BT.1886 implementations; along with the move to the newer dEITP error formulas for color). But this leaves the HP Omen BFD as the main display for use with Nvidia if a person is trying to do like VRR (I'm wandering all over the field with this rant, with critiques of the state of graphics cards, not challenging the findings of him or GN, especially knowing the importance of shadow and gamma on perception).

His point mirrored mine on using affinity, but it shows that X264 is showing its age a bit. His comparison with Linus makes little sense, as LTT uses raw 4K and 8K footage to start with in-house, while having many other things that make the choice of a larger core processor more desirable, while he compares it to streaming. Also, the 64-core is NOT for streaming. It IS for different types of content creation, such as rendering in specific programs (i.e. blender), h.265 encoding, working with 4K and 8K footage, etc., not streaming with limited bandwidth when you will be working with over 100,000 bitrates for the media it is meant for, not the slower bitrates of game streaming, which makes his comment a little disingenuous in that instance (not suggesting LTT can use above the Youtube/floatplane/twitch cap, more that they are dealing with different specs, while the higher bitrates are for more of blu ray bitrates, which doesn't apply to LTT or streamers at all). He then mentions h.264's original purpose (and x264) in passing right after that.

Overall, I'm more interested in where things will be once they enable HDMI 2.1 compression so that you don't have to subsample to 4:2:2 or 4:2:0 chroma in order to get 4K@120/144 or 8K frames. That is more a QoL enhancement for the gamer, not the streamer. Also, when are more monitors going to do hardware 3D LUT implementations similar to how many companies producing TVs are implementing that in partnership with Display Portraits (CalMAN software)? (Side note - this became more prominent with the adoption of Dolby Vision calibration, especially after Vizio's implementation, leading to LG to use a DV 2.0 variant where you could upload the correction file, then later moving to autocal).

But, they have their points. Also, GN was able to replicate, roughly, the exact use case AMD showed, just followed up that it isn't worth it and instead they should have showed a dual stream at medium, one to online and one for local copy, as a more realistic means of showing the performance of the new chips, which is a fair critique.

Edit: Also, I didn't fully explain the benefits of compression or that the new HDMI 2.1 cables with 48Gbps bandwidth allows for 4K@120 without compression, in part by not discussing when the compression is used. Here is a table to better show the frequency, bit depth, and chroma on when the new standard would use the compression and when it wouldn't.

Edit 2: Here is a decent HDR primer from SMPTE for people looking to understand color space, gamma, and the new standards being implemented with the new generations of displays. We are undergoing a significant shift in display technology right now, whether people recognize it or not. As such, we will need graphics card manufacturers, working with standards committees like HDMI and VESA, along with display manufacturers working with ITU and SMPTE, to help bring together this changing forum. Adobe RGB and DCI P3 have been color spaces within the PC segment for a long while, called WCG (wide color gamut). With HDR, standards related to luminance for spectral highlights has entered the discussion. Previously, with WCG, displays were still to work within the accepted 80-120nits (up to 150nits) of SMPTE standards designed for REC.709. Now, we are going beyond this. As such, it will be interesting to see where, when, and how these changes are implemented.

https://www.smpte.org/sites/default/files/section-files/HDR.pdf

Also note, changes to video do not always match those of photography.

 

@ajc9988 some the preview of 3600 came out here https://elchapuzasinformatico.com/2019/06/amd-ryzen-5-3600-x470-review/

CB15/CB20 both shows IPC over intel at around 7-7.5%. pretty damn nice, just the clocking frequency is quite sad.

@Papusan clevo with AMD 2020 lets go

 

The 3600 rumor is for an engineering sample running downclocked, and it matches closely to 9700k single core scores.

No mention if the Intel CPU has Security Patches On or Off for any of the leaked benchmark results...

Mojo JoJo says...

 

intel patches has like nothing on CB ST test. the 7-7.5% is based off rough estimation. I have the 3600 at 4.2ghz while 9900k only at 4.7ghz instead of 5ghz. if its at 5ghz then difference will be even greater. that ST performance really matters a lot to me so its the same as frequency.

9900k at 5ghz is around 19% higher clocks than zen2 at 4.2ghz so any IPC is welcomed here.

 

I just ran the math, and if scaled from 5GHz, it showed IPC lower than that of the 2700X, which just shouldn't happen or effect CB ST even with mitigations. Also, his ram is questionable on setup and timings, along with many other factors.

So, doing it from 5GHz gave 14% IPC over Intel, doing it from 4.7GHz showed higher IPC for the 9900K over the 2700X, which is proper, but shows 7.5% IPC over the Intel chips. At 7.5%, you would need 4.675GHz to reach an Intel chip on performance with IPC and frequency in regards to single thread performance. Due to AMD having better SMT implementation, multi-threaded will likely need less than that.

But doing the math on the multithread or trying to do the scale to 8 cores is coming up with wonky results that make no sense. It suggests an 8-core at 4.2 could beat a 9900K at 5GHz, while only getting 2051 for a scaled score to 5GHz in multicore for CB15.

For these reasons, I'm going to wait for my normal outlets. It's only two weeks anyways. I'm betting this is ES and that there are other things happening (like a throttle event below 4.7GHz for all core).

His single core score also is lower than even PCWorld's single thread by like 20 points. It is closer to their 8700K score.

Because of things like that, I will wait (plus, outlets like Anandtech already plan on doing a clock down and compare IPC with same frequencies, which is a much better way to address it).
https://www.pcworld.com/article/3315057/intel-9th-gen-core-i9-9900k-review.html

 

Keep telling yourself that the Intel patches won't affect your CB scores as you watch your IO and network CPU load increase XX%, and everything else you do that passes through those patches slows down everything around you - with those patches stealing CPU cycles away from everything during interactive and multi-task use.

It's still too early to call if Intel is now on the other side of the benchmarking fence, firmly on the outside of the "winners circle".

 

if you understand CB then u'll know patches dont affect it. i dont need to tell myself, i tell you. besides who will patch those on consumer side anyway? i havent even updated my bios clevo to take 8 cores cause 1. it comes with the latest bios which ruins storage performance, 2. the heatsink mod is on hold and default heatsink can't really handle 8 cores anyway.

@ajc9988 we need Ian to test the 1:1 and 2:1 memory latency comparison. it just wouldnt make sense to me that AMD doing the 2:1 for only the sake of overclocking ram but actually have degression because of it, it is just too stupid. i'd imagine some workload does benefit from 2:1 because of faster bandwidth even if infinity fabric is at a lower speed.

also would love to see the efficiency as well as overclock, these are the 3 biggest thign im waiting for. ram/memory test, overclock ceiling on 7nm and power used by these chips. frequency ceiling will decide I buy AMD system this yr or wait for 7nm+, and power efficiency will dictate how many cores of CPU i'd buy, 16 24 or 32.

 

@ole!!! sometimes you have trouble reading what is written and only want to force it into how you want to read it.

The point is that CB scores don't matter. Even if the CB score isn't materially affected by the patches - so far - the whole OS and IO subsystem during interactive and multi-tasking usage are greatly affected - affecting everything you do, no matter what it is.

You aren't simply going to spend your day re-running CB over and over on an otherwise idle system to the exclusion of anything else - you are going to be doing lots of other things at the same time that will be affected, either directly or indirectly as being affected through the whole loss of OS performance.

One of the things that isn't being done, because it's difficult to replicate, is to run whole system tests to see the interference with applications and even benchmarks.

Running tests in isolation isn't normal behavior, and so results from such tests won't be indicative of real world usage. Running short duration small focused tests won't replicate normal rendering usage or results. Large disk IO, network IO - reading and writing data to and from disk / network and memory will be greatly affected and these slowdowns will negatively affect your end results.

That's all I was trying to say. I thought I said it clearly several times, but I guess I needed to explain the entire context for you to pick up on what I was trying to express. Sorry to have been too brief for full a explanation.

In addition, you will find that the new CPU rev's will have fixes that can't be turned off, that will use firmware changes that affect IO, and likely OS patches as well, and the hardware fixes are only for a limited subset of the known vulnerabilities and none of the unknown vulnerabilities.

You'll need to use down rev CPU hardware to be able to avoid the fixes. If you are comparing a Ryzen 3 upgrade to a potential Intel upgrade, you have to take this into account.

 

i mean you write a full wall of text lol why not just put down the important key points.

 

Not necessarily Ryzen, Vega, Epyc, etc., related, but this rumor may impact things down the road - https://wccftech.com/frank-azor-joins-amd/

If true, let's just hope he doesn't bring start a BGA to the motherboard for AMD components like CPU, GPU or whatever.

 

That already exists. AMD Epyc embedded 3000 series is a BGA selection. Most consoles are embedded solutions. Etc.

Hell, mobile APUs are BGA. They won't lock down desktop socket or server. Swapping chips is too important there.

Mobile, except for the few with desktop chips, how many AMD laptops offer other than BGA?

 

I should've been more specific in my comment about taking current LGA offerings and all of them going to BGA... It's really an inside joke to those following the XPS/ Alienware threads where this seems to be a common theme with the exception of one model, and that LGA model seems to be having its own problems.

In regards to the AMD LGA on a laptop, I believe there is one model, but keep hoping (against ALL ODDS) that there will be more.

 

Comments like that are why I end up ignoring you. I succinctly explain something and you pretend not to understand it. I further explain it and you misinterpret it. Then I try one last time by putting in the effort to fully explain something you are refusing to see, then you flippantly make some irrelevant remark while continuing to pretend like you don't see or understand.

Sticking your head in the sand and mumbling nonsense won't make the Intel security patches performance hit go away.

The Intel CPU security issues will continue to reduce performance even if a small tight benchmark dodges effect from the security patches, while the whole system is falling down around your ears, while your head is stuck firmly in the ground trying to ignore reality.

Ignoring it won't make the security mitigation performance degradation of Intel CPU's go away, and neither will Intel moving the firmware patches into the CPU die, the OS and firmware patches are still enabled and taking performance away from the system.

 

Yeah, it's a joke that just won't go away. The irrelevance of LGA / PGA vs BGA is the biggest joke here on NBR. Thermal throttling issues have nothing to do with what socket or soldered attachment is used, it has to do with laptop physical size and attention to details of thermal management.

Unlocking power limits and OC'ing beyond the physical capabilities of the system to expel thermal energy will throw out the window all of those careful designs once the demands outstrip the potential of the system to transfer and expel the heat generated.

Putting all of that on the irrelevant physical attachment of the CPU / GPU to the PCB is ludicrous.

 

you probably didn't know this, i read most of what you wrote. my reply with a short quote from you is an example to show when most disagree with you or dont like you, they'll read the first part and ignore the rest, usually stopping at the first point they come to disagree with.

the mitigation patch by turning off HT reduces performance by quite a bit but honestly only the business do that. I just dont see the need of doing it whos going to hack my computer to see all my pr0n collection? (hardware pr0n, yep).

also some did tests already, patches barely did anything to gaming and most software like gaming are still coded towards intel's processor. gaming at best loses 1-2% in fps.

 

I go through all the effort to explain it to you but you can't do the decent thing and reply intelligently. Of course I knew you read what I wrote, it was informative - how could you not enjoy reading it. You didn't like the inevitable conclusion, but the facts in evidence are well publicized - there's no getting around it.

Intel CPU's security vulnerabilities are a mess of gigantic and hopeless proportions and should be avoided whenever possible. Thank goodness that AMD has CPU's that are now up to the task of taking over for Intel until Intel can get new CPU architecture(s) and new processes online to recover from Intel's many sad messes. It will take Intel many years to be ready for a comeback.

You don't like what I wrote because you don't want to accept it, because it means you need to consider not buying Intel CPU's - the intelligent considered result of examining the facts in evidence.

Your unpatched Intel CPU computer can get hacked in other ways once your passwords or data details are read, turning your PC into a bot, or it can be used to access your bank account, medical records, or other online accounts.

Your unpatched computer being a laptop can be active on public wifi and infect other wifi devices, and when you plug in to work or school the same thing can happen. Sharing files via USB device can transfer hacks to other computers too.

See how many things you ignore when you deflect things outright without due consideration?

You can stop responding to my posts and I'll stop replying to your's if you really can't get past this mental block. But, I am happy to help you through it too.

 

Ok, enough, stop the bickering!

 

I'm not worried about thermals, throttling, nor power delivery issues. LGA based laptops can suffer from same issues. I do agree that thinner and lighter is at the root of those issues.

But BGA is leading us to the "throw away" laptop, where manufacturers have everything disposable. A component goes bad, "we'll just swap out for a new system." What is the repair cost outside the warranty period? This is utter non-sense.

And now we're seeing more and more soldered wifi, ssd, and memory. I don't expect you to agree with that assessment, but in turn, don't expect me to gleefully accept this trend in laptops either.

 

What you are complaining about isn't BGA or soldered on components, it's the scammy business practices of companies refusing to do component level debugging and repair, and the now long tradition allowing them to get away with it.

Those wasteful scammy business practices are what need to change.

There are hundred's of surface mount parts used in most PC's, laptops, handheld's, and all of them can be replaced to repair malfunctioning boards, but for the most part the board's are replaced as a whole component, and the customer is charged for the whole board - instead of only charging for the repair and cost of components that have failed.

A single replaceable component of 1 LGA CPU would make only a fractional difference and in my experience the CPU rarely fails, so the whole motherboard is still replaced. The CPU can and will be pulled / desoldered and used in other builds - LGA or BGA.

Watch any number of self-repair video's on repairing consoles, handhelds, phones, laptops - and there are independent shops that also repair these devices doing component level repair / replacement of any parts, including BGA components.

They scavenge the parts from otherwise unrepairable devices to repair otherwise good devices, or they can order those replacement parts online from other shops that specialize in scavenging components from failed or rejected devices.

If little repair shops and motivated individuals can do the work themselves, then the large makers of laptops can do component level repairs too.

It's the business practice of scamming customers into believing that the only repair possible is swapping the entire motherboard, and charging the customer the entire cost of the whole board instead of the individual components debugging and repair costs.

Remember that the MXM GPU is a BGA GPU and VRAM are components soldered to the MXM board. The same component level repairs can be done on the GPU board as well. The same problem exists - the vendor wants to get away with charging for a whole MXM GPU board instead of doing component level repairs on it.

Recycling entire boards through component level repair in manufacturing saves lots of resources and costs for the world, and would do the same for the individuals paying for repairing them.

Component level debugging is still used in manufacturing, but after leaving manufacturing the scammy business practice that they can't or won't do component level repairs to repair customer purchased hardware - that is the real root cause problem, not BGA.

Soldered on BGA components are not impediments to repairing electronics.

AMD uses PGA instead of LGA, and there are several AMD PGA Ryzen 1 /2 CPU laptop models out there, with hopefully more laptops coming with Ryzen 3 desktop CPU's

AMD are using APU's to push into the laptop realm, and so far all of those are AM4 PGA socketed as well:
https://www.amd.com/en/products/ryzen-processors-laptop

AMD has embedded design Ryzen APU's with BGA sku's available:
https://www.amd.com/en/products/embedded-ryzen-v1000-series

 

Not a fan of rumors, but this seems to be a real review - premature based on the lack of adequate BIOS updates - so the results are of limited use. Let's hope the reviews after release day concur with and improve upon these early review results.

Review: AMD Ryzen 5 3600
BY IVÁN MARTÍNEZ, 06/25/2019
https://elchapuzasinformatico.com/2019/06/amd-ryzen-5-3600-x570-review/

Early Ryzen 5 3600 Review Shows Incredible Performance At Only $200
WccftechTV
Published on Jun 25, 2019
The first review of the AMD Ryzen 5 3600 processor has been published by Elchapuzasinformatico.


Juan Chemell 17 hours ago
"Those numbers without OC, RAM only @3200MHz, no BIOS update, old platform and the cheapest zen 2 processor... AMD smoked Intel."

modar aldihne 15 hours ago (edited)
"X470 motherboard wouldn't overclock zen 2 chip ?or no bios update yet"

WccftechTV 15 hours ago
"My understanding is that the bios wouldn't allow it. But, it's most likely because it hasn't received proper updating."

TheCooperman666 18 hours ago
""The 9900k is a 95w cpu" you mean the 9900k is 95w on standby and 150w when turboing...."

Kevin Francis 15 hours ago
"...Well, considering the Ryzen 2700x (105W TDP) draws 106W at full load (turbo) and the 2700 (65W TDP) draws 62W at full load - yes, the 3600 will end up drawing close to ~65-70W.

https://www.anandtech.com/show/12625/amd-second-generation-ryzen-7-2700x-2700-ryzen-5-2600x-2600/8 "

AMD Ryzen 5 3600 6 Core, 12 Thread CPU Tested on X470 Platform – Single-Core Performance On Par With The Core i9-9900K
By Hassan Mujtaba, Jun 24 2019
https://wccftech.com/amd-ryzen-5-3600-zen-2-7nm-cpu-review-published-online/

 

I disagree. By soldering on parts, it prevents those without the equipment or nearby shops from effectuating repairs. As components are additionally soldered on, it raises the board cost, when those other components could be pulled and put on a cheaper mb replacement if not soldered.

A better practice is leave the sockets, them allow companies to do a core charge for malfunctioning boards, similar to auto parts. By returning the bad board, while getting a new or refurbished board, then it allows OEMs and ODMs to do component level repair, while the quick switch allows the consumer to get up and running quicker. It also creates a better middle tier refurb board market once they prove reliability in repairing those boards.

That roughly accomplished your stated goals, while also allowing users lower costs.

The surface Mount movement is due to lower energy usage combined with the thin and light/lower thermal dissipation package designs.

So, consumers with expertise like sockets to swap components to allow avoidance of component level repair, while systems manufacturers like the other for higher cost components and sometimes better margins. But, the manufacturers also are feeding the disposable issue and consumerism, mainly to drive profits at additional costs, including to the environment.

 

All of this hullabaloo over 1 socketed part, the CPU. The good news is AMD has all socketed parts for desktop and laptop - embedded CPU / APU's are BGA

An MXM GPU isn't socketed in any laptop, it's soldered on to the GPU board which is plugged in to the motherboard.

Everything else is soldered, even in the LGA / PGA CPU models.

There are many parts that are just as difficult to unsolder and replace without special desoldering equipment and training as the CPU / GPU chip.

The real problem is that the vendors are replacing the entire motherboard or GPU board with a new - or refurbished (previously component level debugged and fixed) - board to accomplish repairs, and charging for the whole board value instead of the component replacement value.

That's really the problem that needs to be resolved. Through legislation or through industry support for component level repair.

Without a major commitment to education, training, and an expensive investment in diagnostic and specialized desoldering equipment no customer can possibly be expected to accomplish 100% self-repair of their computers, laptops, and other electronics by debugging at the component level.

Plenty of people on NBR post about their disappointment after buying their socketed LGA CPU and swappable MXM GPU laptop only to find out down the road when they are ready to swap in new parts that it's a huge hassle to do such upgrades, and in most cases because of cost or lack of vendor support it's not worth it, it's more cost effective to sell their "upgradeable" laptop and buy a new one.

It often takes a real experimenters effort to get upgrades to work, and even then there are problems, caveat's, limitation on end results, and expected features that won't work. Not something the average person is going to want to commit time and effort to accomplish for a partially successful upgrade as compared to buying a new laptop with the new features.

If all the noise and drama over LGA vs BGA were converted into a push to get vendors to support motherboard / MXM board swaps and charge small fee's for repairs - component level debugging and replacement costs - that effort could provide real cost savings benefits to everyone.

 

BGA CPUs are just one aspect of this shift in laptop design. Yes, using "BGA" for this is not technically accurate. See http://forum.notebookreview.com/thr...-owners-lounge.826831/page-1148#post-10913911 But to many here, BGA has become the metaphorical acronym for non-interchangeable, soldered on components within a laptop. And who knows how long before it seeps its tendrils into desktop space?

Yes, it can be done at repair shops, but I'm not talking about that. While not having the proper equipment nor fine motor skills to handle delicate soldering, repairs may soon be out of my reach. And extending the life of a laptop or desktop is something that I may no longer be able to perform because of this. Before the "soldered evolution", I was perfectly capable of swapping out a bad DIMM, SO-DIMM, HDD, SSD, network card, and if you want to go old-school, a SoundBlaster Pro, etc. But alas this is becoming more of a thing of the past.

I think now we're getting away from point of my original post - in what capacity will Mr. Frank serve AMD and what impact may he have with their customers, in particular ODMs? We don't have positive confirmation of anything just yet, so who knows if any of this will come to pass? But if true, as you mentioned earlier AMD has a lot of different businesses. My hope is he ends up in gaming console land. That would be a great landing place for him, and he won't cause much damage to the laptop / desktop landscape.

 

This gets to my point of a "core" charge. Swap the boards for consumers, but once the old board is in custody of the repair service, there should be a refund of sorts due to that board now being used for refurb, whether it is a donor board or a recipient board moving forward. By doing such, it gets rid of the extra cost of a new board being higher, since the return of the old board acts to offset the value. In that way, we all would benefit.

Also, GPUs being soldered only became more prevalent on high end due to the aging MXM 3.0b standard and the lack of wanting to update the slot. But, except for losing out on a killer chip, if the motherboard core charge was implemented, the soldered components (except memory and storage should not be soldered, save gpu mem) are less of a big deal because the cost is lowered.

This then can work with the private component level repair shops as local means to service the core boards to then go back to authorized repair service vendors which can switch out the boards, as well as selling it online, while still being forced to offer the core charge refund on return for those that effectuate their own repairs.

This sort of model has been implemented in other industries without issue. Time to bring it here.


On another note, HWBot contains TWO benchmarks for the 3950X, both with ES chips, one at 5.275 and one at 4.3. These are not for points and reflect the world record, which had to be replicated for HWBot on GB4 in their hardware lab.

https://hwbot.org/submission/4176848_blueleader_geekbench4___multi_core_ryzen_9_3950x_61072_points


https://hwbot.org/submission/418300...nch4___multi_core_ryzen_9_3950x_64953_points/
-165C for this world record


So, specifically those two scores are confirmed and public facing. Running 4.3GHz on all cores for the 3950X on an AIO cooler means I could, with my cooling, likely outstrip that performance by 100MHz (although the voltage already being at 1.425V is oof, don't know LLC)!

Wonder if AMD plans on giving degradation guidance like they did with Zen 1?

Either way, def want to see what comes.

 

For out of warranty items, I like your idea about only paying for "what's busted" on a board swap, but TBH, I would be completely *stunned* if consumers see that put into practice.

That is an interesting concept too. Back in the day (1960s/70s/80s), you'd see TV repair shops almost everywhere. Even in small towns of 5000 people, there'd be at least one TV repair business. It might work, but at this point in time in our society, I have doubts. I'm seeing the local computer shops drying up too. For me locally, 20-25 yrs ago there were 10-12 different local computer shops. It's where I first learned about Sager/Clevo. But now with online and even Wal-Mart having offerings, that number has dwindled to just 3 in a 30 mile radius.

The way everything is now being pushed as 'disposable', I just don't know if this will be viable. Even if it did work, for me, I trust my own repairs vs. someone else. Go over to the Dell forums and read about the nightmare stories regarding their on-site repair folks. I guess if push comes to shove, and the industry moves this way completely, I might consider investing in a video magnification device, a good soldering station, and fine-tune my soldering skills just to keep my hobby alive. But that still doesn't mean I have to like where this shift is heading.

Finally, apologies to everyone for getting us OT from the AMD, Ryzen, Epyc, Vega, etc., news. I merely wanted to bring up the topic of the hiring as that may impact what's discussed regarding AMD on this thread. I did NOT want it to derail us into this other tangent.