Problems? See this thread at archive.org.
The Intel HEDT is over priced and needs reduced. The 2950X seems to be a killer CPU for the money if you need all those cores and threads. I don't so I will stick with my 6/12 and soon 8/16 core CPUs. I don't really need an 8/16 but will grab it for the hell of it and donate the 8700K to the laptop.
-
Lots more to watch, but I thought this one stood out @ 00:19:48
Threadripper 2 2990WX in-depth review & benchmarks
Last edited: Aug 14, 2018ajc9988 likes this. -
That is what I meant by extreme multitasking. Those numbers of rendering on two different items at the same time is phenomenal. Seriously, never seen anything like it. But, as I also mentioned, it is hard to stretch this CPUs legs!
![[IMG]](images/storyImages/ryzen_threadripper_2990wx_simultaneous_blender_premiere-100767450-large.jpg)
PCWorld review: https://www.pcworld.com/article/329...amds-32-core-cpu-is-insanely-fast.html?page=2Last edited: Aug 13, 2018hmscott likes this. -
From the video, these charts stand out as the cross-over point for threads vs performance advantage - left side advantage 7980XE - right side advantage AMD 2990WX, note the cross over point is about 25T's.
ThreadRipper 2990WX vs 7980XE
Thread Load Scaling 1-64 Threads
ThreadRipper 2990WX vs 7980XE
Percent Performance 1-64 Threads
PC Worlds Dollars per Core chart with new ThreadRipper 2 CPUs added
Threadripper 2 2990WX in-depth review & benchmarksajc9988 likes this. -
A couple of new articles, Phoronix.
A Look At Linux Gaming Performance Scaling On The Threadripper 2950X
Written by Michael Larabel in Linux Gaming on 16 August 2018. Page 1 of 4. 10 Comments
https://www.phoronix.com/scan.php?page=article&item=threadripper-2950x-gaming&num=1
"There are a lot of real-world Linux workloads that can benefit from 16 cores / 32 threads on the Threadripper 2950X (or even 32 cores / 64 threads with the Threadripper 2990WX), but Linux gaming isn't close to being one of them."
A Quick Look At The Windows Server vs. Linux Performance On The Threadripper 2990WX
Written by Michael Larabel in Operating Systems on 16 August 2018. Page 1 of 4. 14 Comments
https://www.phoronix.com/scan.php?page=article&item=windows-server-2990wx&num=1
"Windows Server did end up being faster than Windows 10 at the time to run various Git revision control system commands, but that too was still behind the Linux performance."Last edited: Aug 16, 2018 -
I recall reading this Dropbox blog post a while back, he mentioned Epyc favorable, briefly, a year ago...:
"Specifically, if you go the Intel path, you are looking for at least Haswell/Broadwell and ideally Skylake CPUs. If you are going with AMD, EPYC has quite impressive performance."
That shows how long the cycle is for bringing in new datacenter hardware.
https://blogs.dropbox.com/tech/2017/09/optimizing-web-servers-for-high-throughput-and-low-latency/
Here's the recent blog post from AMD:
Dropbox Designs its Custom-Built Infrastructure with Single-Socket AMD EPYC Platform
Posted by scott.aylor![[IMG]](images/storyImages/roleBadge-4-1002-1431754362075.png)
in AMD Business on Aug 21, 2018 11:01:15 AM
https://community.amd.com/community...tructure-with-single-socket-amd-epyc-platform
"As vast as the datacenter market is, it’s a relatively short list of companies working together in the day-to-day business. I don’t typically have the pleasure of engaging closely with a company that literally has hundreds of millions of customers like Dropbox. With over 500 million users and 300,000 Dropbox Business customers accessing its global collaboration platform, Dropbox is the latest big name in cloud to deploy the AMD EPYC™ processor in their custom-built infrastructure.
“AMD EPYC is a compelling processor option for our compute technology, providing Dropbox with the technical specifications required to support the workloads that matter to teams and our individual users,” said Rami Aljamal, Head of Hardware Engineering and Supply Chain at Dropbox. “We are excited to deploy EPYC processors and look forward to working closely with AMD in the future.”
Dropbox will leverage AMD EPYC™ 7351P one-socket processor platforms to support future growth beyond its current capabilities and refresh its existing infrastructure for its most demanding compute workloads.
The AMD EPYC™ 7000 series delivers compelling options for the Dropbox offering, meeting performance demands throughout evaluation, qualification and deployment. With 16 high-performance cores on the EPYC 7351P processor and leading-edge memory bandwidth, AMD continues to drive a strong balance of compute and connectivity while eliminating the need for a second socket." -
Dropbox was an early sampler, but did get on board with sampling after other partners did. So, them announcing deployment is just toward the first of the rain that is coming in Epyc adoption. In fact, this third quarter is only supposed to bring them to 1-1.5% market share in the server space, IIRC. It is Q4 that brings them to 4-5% market share in servers, then next year, Intel said holding AMD to 15-20% market share is their goal, meaning they expect AMD to grab, at a minimum, that much, and that is coming from the competitor and largest player in the space.
Intel's fear has caused them to announce this:
![[IMG]](images/storyImages/115d3d1a6a59e70a.png)
So, they are throwing their board partners under the bus to allow backward compatibility to Haswell because of fear of making them buy new platforms, which favors just switching to AMD. It throws out the two gen cadence and shows they artificially restricted sockets. But, this is trying to get the drop in sales rather than full replacement servers. But, if the boards are not wired for the specs on the new chips, that will limit what features of the new chips can be used (like 48 lanes cannot be used on boards wired for only 40, etc.).
This just seems really desperate on their part, honestly. -
-
Oh boy this is gonna be a long one.
I am talking about well before Zen even launched, when Polaris had JUST emerged and people wanted MXM RX 480s. I was stating (and correctly so) that Polaris is not efficient enough to compete with the 1060, and that you either have a tradeoff... a larger brick than people want for smaller laptops (78W TGP for a 1060N and 55W spike for an intel mobile CPU is fine on a physically tiny 150W brick, impossible on a RX 480), and a power hungry, hot and heavy unit with an entry-level class current-gen GPU. Alternately, you have severely reduced performance to fit the lower envelope... which means once again you could buy a much weaker card for similar performance, like a 1050Ti.
ASUS chose the latter. 65W RX 580, when the full RX 480 required 192W to play witcher 3 at 1440p at STOCK CLOCKS once you disabled power throttling. The RX 580 barely handles 980M performance, far less arriving near a 1060N. And if you want to say it can beat a 980M, then overclock the 980M; something you don't have the power envelope to do in that ASUS (or the cooling, apparently, considering how hot it gets while being bafflingly loud... it's about as loud as my P870, and almost as large too).
As for ASUS not releasing updates... they always. Without a shadow of a doubt. Prevented upgrades. Even when they used MXM they used a custom PCB design so you couldn't physically insert a later MXM card even from their own units into your previous gen unit... you ALWAYS need to buy-upgrade with ASUS. And further to that, I also doubt the VRM solution on that board supports an upgrade in the first place, like most B350 boards; 2700X chips burn em out easy. Though I would forgive you for not realizing this fact, but then the unit isn't sold with a 1700X or 1800X, and those have higher power limits, so... some signs were there.
To be honest, you've been on this forum long enough that having seen NONE of that talk is impossible. It's all over in threads I know you've seen, unless you've only checked maybe three or four threads since coming to this forum entirely. Which isn't possible with your post count.
Now onto the 1060N... the 1060N is worst-case scenario 15% slower than the desktop card, usually less (without any voltage frequency tuning that is; since it was heavily talked about for AMD's side it's fair game here). The OC vBIOS for the 1060 which is 90W TGP is much closer in performance to desktop (probably nearly matching with undervolts). The 65W RX 580 is UNDER 980M performance in quite a few games according to notebookcheck... ignore the synthetics, look at the games. Fortnite and Star Wars Battlefront are both over 15% slower than a 980M, which is generally 30% slower than a 980, with a 980 being what the 1060 ballparks. That's nowhere CLOSE to being the desktop card performance, at all.
As I said above, the 1060N and RX 580 don't even come close to having the same performance, but have about the same TGP (65W for the RX 580, 80W TBP), so... what's the comparison? And how is the RX 580 more efficient when the 1060N creams it pretty much everywhere? Even on the frostbite 3 engine with Star Wars Battlefront 2, WHICH IS AN AMD-FOCUSED TITLE on an engine that LOVES extra CPU power and scales to multiple cores like a squirrel scaling a tree trunk. Instead, the 1060N is a whopping 37% faster:
As for the lowered performance and clockspeeds... this is why you wait a short amount of time for reviews like from Notebookcheck, where you can read the raw data. And speaking of reviews, this brings us to my next point:![[IMG]](images/storyImages/unknown.png)
ASUS, like Razer, always review well. Because they send cherry picked units and in some cases fly out techs to outlets to make sure the unit is "working properly". And if you review the unit badly, guess who is unlikely to get more units in the future? You're never going to find a bad review of a notebook that doesn't completely break (like locking a CPU at 800MHz), because reviewers aren't in the business of giving you a proper picture.
A full RX 580 would require somewhere around 150W on its own for power. This instantly limits it to 230W+ power bricks if you want the advertised performance, as opposed to 150W or 180W for a 1060N unit. We've been over this before in the other thread I referenced at the start. Either you neuter performance, or you make a unit designed to house a higher end part from the other vendor, like a 125W 1070N or even a 150W 1080N, both of which outstrip a 580 so much it's laughable. As for the CPU, we never discussed this before, though i knew from a week after Zen's launch it would be horrible in any notebook. We don't have the RAM to feed it properly.
What you "overpaid" for, is the performance. You could have had a 1070N and a 7700HQ for $1800 USD with your existing storage/memory configuration easily around the time you bought it. Nvidia cards are indeed overpriced for notebooks. OEMs charge double what they're capable of selling them for to still make profit, despite what anybody else tells you. But what you bought, that cheap B350 chipset with that cheap 1700 and that completely neutered RX 580 for the price you paid and the poor cooling and loud noise when doing said cooling? That's beyond what any decent intel/Nvidia unit needs to run someone. Even the overpriced Aorus line from Gigabyte could be had for your price with 1070N cards and liquid metal repastes from GentechPC at the time.
No, just like what I said initially to you here, you STILL insist on an all-AMD notebook. Even if price is exorbitant for minimal/reduced performance. Knowing full well you will never achieve full performance out of AMD on notebooks because they won't use 3200MHz+ memory to properly support the CPU, along with the fact that notebooks generally tend to use single rank single sided memory too, so you don't even have the benefit of dual rank memory improvements unless you start with 32GB in four stick configurations or hunt certain kinds of memory sticks.
So once again, I make the statement: Your actions don't reflect what you said your thoughts are. When I told you RX 480s were ill-suited to mobile, you denied me. Look at the existing implementation now. The other unit that has it (with still-neutered performance) is a HP Omen with 120W TDP and unknown but certainly higher-than-120W TGP/TBP, far above even the high performance 1070N vBIOSes which are 125W TGP, which destroy it in performance, far less after undervolting is applied. Your above post JUST got through defending them doing this, saying it was ASUS' fault and not the tech's fault. You're still insisting Zen/Zen+ is the best case for a notebook, despite all known implementations lacking high speed memory support and any form of overclocking, with a 2700X unit still-to-come, when you could be using 4.8GHz 8700Ks for your production all this time with well over 3000MHz memory in a P870TM1 or even P750TM1 if the load was CPU-only and you bought it from the right vendor... it would not have been much more with a 1060N in it either in terms of price for your storage/memory setup, with better performance across the board.
But you still insist on all-AMD. What am I supposed to think? And then you didn't even know a 1080 and a 1080Ti were two different cards, AFTER Pascal's entire life cycle ended, meaning you barely even glanced at their product line.
Edit: Was pointed out that the HP Omen's 580 is ~120W board power and 85W TDP, not 120W TDP and higher board power. Still matches it to a 1070N which cleans its clock, though, but I am corrected on that.Last edited: Aug 24, 2018Cass-Olé, Papusan, yrekabakery and 3 others like this. -
This is misleading. Your RX 580's 68W is only the GPU core power, while a notebook 1060's 78W is total board power, which includes not only the GPU core but also GDDR5 memory and VRMs. If anything, the 1060N is more power efficient when you factor in its higher performance.
The title being compared was Star Wars: Battlefront II. This game uses EA/D.I.C.E.'s Frostbite 3.0 Engine, which is notorious for having a DX12 implementation that runs worse than DX11, on both Nvidia and AMD hardware.
Nvidia's Pascal GPUs also come highly overvolted from the factory. The voltage/frequency curve is overvolted by about 150mV more than necessary across the entire range. That's why on my GTX 1080, I can stably run 1848MHz @ 0.913V, whereas stock would have had me at about 1.050V at that clock speed.
Voltage and clock manipulation can be done using the curve editor in MSI Afterburner, it is not locked down:
![[IMG]](images/storyImages/uSNlhz9.png)
(Thinks back to 1080 Ti is overclocked 1080 comment).
Pot meet kettle, I guess. -
Book below
I pointed out a weak GPU strong CPU system to you for a similar price.
I made no errors I did not already correct with my last edit (which fair you did not see when quoting this).
Who said anything about finding you an 8c/16t? Your 8c/16t is smoked by the 8700K recommendation I made, which by your own admission was already out by the time of your purchase, as you can see here where the original post was on october 5th for its last edit, with a working link to the product page on HIDevolution. This is an invalid point, as I said you could get a weak GPU and strong CPU for a similar price above.
I mentioned specifically "from the right vendor" also in my 8700K statement, also invalid point.
Please don't tell me that laptop has a decent anything in it. We both know it doesn't. I see you admitting that they could have done far better but just because they're two rungs off the bottom where the rest of their notebooks are doesn't mean they're worth any defense. And before you say I am partial to any vendor of ANY kind, I am not. They do well, they get praised and recommended. They don't do well they get picked apart. Most don't do well. Someone asks me a question I pick the best thing for their use case that doesn't fail to perform in some way, whether it be overheating or artificial power throttling or just poor, poor quality control/post-sale service.
Would be nice to see if you could clarify this against the SWBF2 test I posted, then, if future drivers helped so much. That aside, Freesync and Gsync mobile are the same thing. Both are proprietary GPU vendor software-based tech that manipulate VRR in displayport screens, and BOTH ARE FREE. Gsync costs nothing, all Nvidia mobile cards come with it, and the end-user needs not pay for it. Straight from the horse's mouth Nvidia told me this much. Anybody charging for it is doing so to pocket more for themselves. There IS a screen whitelist, but that is all, really. And it doesn't have a cost trickling down. I know this because when I wrote the article bashing Max-Q, Nvidia contacted me to correct my statement about the "price" of mobile gsync being attached to the cards.. to tell me there was no price, so I could not make that statement, and that anybody selling "non-gsync" versions of gsync units are doing it themselves, aka putting in money into a vBIOS that avoids enabling gsync entirely.
I said TGP, aka TBP, which is full graphics draw (core + memory/VRMs/etc) is 78W on a 1060N, and 80W on that ASUS RX 580. It's not 68. It's 80. By that logic the 1060N is 60W and your point is MORE invalid. The performance is what it is. Trying to find AMD's absolute best case scenario (DirectX 12 in an AMD-focused title) doesn't make the card better than it is.
Mobile Nvidia has no voltage optimization either... you can undervolt them, though, by adjusting the voltage/frequency curve using MSI Afterburner. Get more out of the lower power limit. It is unlike Maxwell or Kepler or prior units where undervolting either was impossible or required a modded vBIOS. And in fact, mobile Nvidia cards are even WORSE at optimization than their desktop counterparts, because the voltage doesn't filter out at the level it needs... once you approach the top 5 boost bins (1860, 1873, 1886, 1898 and 1911 MHz bins) you automatically pump full voltage into them on notebook cards, where desktop cards will oft remain lower like 1.025v instead of pumping to 1.063 like notebook cards do. And the cards don't need that voltage AT ALL. Here, look at the voltage required for 4K firestrike if I adjust the voltage curve a little bit on my cards:
Pretty big drop from 1.063v flat ain't it? If I didn't adjust that curve look at how much throttle I'd be having due to power limits (not temperatures as you can CLEARLY see):![[IMG]](images/storyImages/3DMarkICFWorkload_2018_05_02_17_00_23_223.png)
![[IMG]](images/storyImages/3DMarkICFWorkload_2018_05_02_17_06_39_373.png)
So like I said... if you want to make the point about undervolting for performance, you need to apply it to Nvidia too... and oh boy does undervolting ever grant quite a bit of performance there on mobile. So, this point is pretty invalid.
Here it is.
Listen. I'm autistic. I literally only see in logic. Logically, there is no reason to do what you did unless you have a grudge against other options. You don't set out to get something knowing worse performance and from, as I believe it, a vendor you should know almost never delivers, for more money than it's worth ($60 B350 boards anyone? $220 RX 580 4GB? $275 1700? Near $2000 for laptop? Hello?) and just... no, I have nothing further to add. There is just no logical reason for you to make that decision.
Well now you know better about ASUS
Now that I'll be waiting to see.
Giving AMD a chance when products are comparable, is understandable. Giving AMD a chance when their products are inferior, or are inferior on the chosen platform (in this case in that notebook), when other superior products are available at a competitive price, I cannot understand at all. Which is the point I made above with P750TM1s.
I know Zen and Zen+ pretty well inside and out, I don't know about manufacturing processes that are limiting them.yrekabakery likes this. -
-
-
AMD's Ryzen CPUs (Ryzen/TR/Epyc) & Vega/Polaris/Navi GPUs
Discussion in 'Hardware Components and Aftermarket Upgrades' started by Rage Set, Dec 14, 2016.
