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

If the frame rate is already above your refresh rate, that extra 23%/15% is getting tossed on the floor by G-sync, v-sync, frame limiters.

If the frame rate on Ultra isn't fast enough, switch to Very High, High, Medium, etc - you can tune it to increase FPS to meet your 144hz / 120hz / 60hz display. And, there are other switchable settings that can be disabled to increase frame rate without reducing quality.

You can try to make a point that Intel is 10%-20% faster in some things - not all things - AMD is faster in some things too!

As soon as you move to 1440p/2160p, the CPU bottleneck moves to a GPU bottleneck, Intel's advantage disappears. At least until GPU's get us into the 144hz range in 4k, and by then we can buy a new AMD CPU for our AM4 / TR4 motherboards.

There really isn't a compelling reason to recommend Intel, it's got a short lifespan. The middle of next year you'll want to dump your investment and get the new Intel z390. A little after that the 10nm parts will arrive and a whole new socket and motherboard chipset.

Intel better be faster now, because you're likely stuck with it for 2-3 years until you can justify blowing $$$$ on a whole new system build.

 

Ok, I love the comparison, but I must remind people we have a thread for that. A little comparison is fine, but this is turning into a versus comparison with people arguing both sides. I'd say keep what we have, but move further discussion to the other thread...

 

You're not tossing aside the extra frames, a lot of those games struggle to even get to 144fps, and don't forget that these charts are showing average fps, I'm thinking the minimum framerate is also consequently higher on the Intel systems (so you're not throwing away fps there either).

Dialing down graphics settings is largely only applicable to GPU performance, so you don't get huge increases in framerate by dialing back most graphics options in CPU limited scenarios, and who wants to compromise with downgrading settings when you don't need to.

Yes, I agree!

Yep

No way Jose! We know Intel is a better gaming CPU like we've been discussing, and the 8700K beats the 1700X in most or at least a lot of productivity tests too - again that was seen in the link I provided earlier.

Yeah, future upgradeability of Intel motherboards with new CPUs is not having a very good track record!

EDIT: only just read your post above this one ajc9988. Without comparisons you can't really judge performance, so I think it's ok to have AMD vs Intel performance comparisons in an AMD/Vega thread. I'm not gonna keep discussing with hmscott over many posts & pages, we're nearing the end I believe, at least for today!

 

Hey, I already pointed out that if you are pushing Intel or versus, that is the other thread, not the AMD Ryzen thread. Can you not understand that? Now, if you didn't string all those comments, I would have thought you didn't see my comment. But I'm pretty sure you did. So stop with the OT and move it to the proper thread!

Edit: Here is the proper thread link for those that do not have it:
http://forum.notebookreview.com/thr...edt-x299-vs-x399-xeon-vs-epyc.805695/page-179

 

Solder vs. Intels thermal paste is a difference. But not the power concumption. Expect with delidded 8700K and Liquid metal insted, you will see there aint a big difference. Watt = Heat.

 

Hey man, chill dude! ;-) Who made you God, apart from your ubiquitous, copious posting in this particular thread. Ha!

EDIT: also read my EDIT in my previous post where I mentioned your post.

 

Ok, read the edit. I'm just trying to get this converted more to troubleshooting and use/performance now. That was more what I was getting at.

And I do agree with comparisons to help, but when it starts going to people getting entrenched, that is why I made the other thread, so we could hash it out at length without issues of people complaining. That is also why I said leave it up to make sure people saw the good info on comparing. I've just seen/been involved in much of the back and forth, so wanted to point out a better place for the comparison.

Personally, Kaby refresh (covfefe lake) seems rushed and was meant as a response. But, depending when Zen+ (Ryzen refresh) comes out (I'm betting late Q1 or early Q2), we may see a bit better response, which then gets hit in the summer by true coffee lake, followed by zen2 in late 2018 to early 2019, with ice hitting in the first half of 2019 (most likely), then Zen 3 dropping late 2019 to early 2020, followed by Tiger in 2020. So, it is a tit-for-tat on releases and the real question is whether AMD will continue to close that gap with each subsequent release....

 

I hope they do. I'd buy Ryzen if I had a 60Hz monitor, because of the larger number of cores for general future proofing, and for the better value overall.

 

I prefer gaming in higher res where none of this matters, so Ryzen was my clear choice. Now it is just setting up my custom loop that is a headache! Have to order a new o-ring, which means since I'm doing that, might as well buy more fittings for changes I want to make that I learned about while assembling the loop, etc. I really underestimated the work in building a custom loop, but it will be worth it!

 

Further update: while waiting for the O-rings, I decided to take it apart and look. Turns out two torx screws ARE STRIPPED!!! So, now I have to deal with getting them out, then replace the screws as well! Glorious! The gift that keeps on giving!

 

Are you doing soft or hard tubing? I have a potential hard tubing build coming up for a client for when/if we can get the 8700K in hand.

 

ANNNNND my pump died. LMAO! I should have known it was going. I booted up my TR and saw the temps reading at 71C immediately. It crashed. I tested the pump and it won't come on. Off I go to find a new one.

 

I'm going soft, but almost wish I had done hard considering some of the bends. Since I have to wait on the O-rings (I got the screws out and sure enough, the big one had thinned in the corner where it leaked and higher in that same vertical stretch), I'm going to order more pieces, such as a Q-Plus rotary from bitspower so that it will go 90 degree to male to male, to female to female extender to the top side going up (fill/bleed port aside from the one on the radiator) while the other side goes compression to straight stretch to compression to 90 degree on graphics card. Lot's of potential failure points, but it is a tight bend on the tube anyway and a fill/bleed port between the upper two 480 radiators will be nice, if for no other reason to help drain the system (whether by blowing or canned air).

As to the pump dying, tough luck! I have a redundancy on mine. After I get it built I'll post some picks. I never thought I'd say this, but my core x9 is too small for my build. I can only do push/pull on the CPU side and in the bottom because the second set of fans would hit the ports on the GPU block.

Also, if you are willing to tear down a bit more, EK teased their monoblock today on facebook for the Zenith Extreme!

 

This is the pump that died on me personally in six years of building open loops. It was bound to happen and lucky for me, all I have to do is replace it. This does give me an excuse to go a bit crazy. I am glad nothing got damaged during the leak of your build. I am going to talk to eBay (I purchased the pump from a seller there) to see what my options are since it has only been a little over a month. In the meantime, I am pump shopping.

 

Well, it soaked the MB and I pulled out the CMOS battery after it had been soaked, but haven't tested the system yet. But it hadn't been turned on, or even plugged in, for days before the testing. Meanwhile, just letting the board dry thoroughly before any testing. Still, it chaps my hide that I wasn't just told. But, this was the only one for sale at the time, so it just would have been factoring and ordering O-rings from the start, which would be no big deal. It just chaps me!

Yeah, you can pick up swiftech mcp35x or mcp50x for around $70, which isn't too bad. But, the MCP50X is a custom DDC, not a liang DDC, so other tops won't work on it like they will for the other.

 

Yup, that's why the other thread was made @Robbo99999 - check it out

Ryzen vs i7 (Mainstream); Threadripper vs i9 (HEDT); X299 vs X399; Xeon vs Epyc
http://forum.notebookreview.com/thr...-vs-i9-hedt-x299-vs-x399-xeon-vs-epyc.805695/

Mod's tend to move "vs." posts over there...

 

ASRock Fatal1ty X399 Professional Gaming Review + Linux Test

 

ASUS ROG Strix Vega 64 Tear-Down, Cooler, & PCB

 

Fake?... + Various Rumors...

Alleged Intel Slide Shows Upcoming Mobility Processors With AMD “Vega” Graphics Inside
http://wccftech.com/intel-mobile-processors-amd-vega-inside/

"We have heard rumors that Intel and AMD might be working together to design a processor that utilizes RTG’s graphics technologies. While Intel had previously denied such reports, a slide has leaked out that says otherwise."
AMD Navi GPU Released Mid 2018 With Multi Chip Design | Vega Inside - AMD GPU in Intel CPU?

AMD’s Next Generation Navi GPU Will Be Launching in August 2018 at SIGGRAPH – Monolithic vs MCM Design Yields Explored
http://wccftech.com/amd-navi-gpu-launching-siggraph-2018-monolithic-mcm-die-yields-explored/

 

Nvidia fire up against AMD's latest GPUs, RX Vega 56 & Vega 64 NVIDIA GTX 1070 Ti Confirmed with Release of MSI Afterburner 4.4.0 Beta 19

 

The 1070ti is a big nothing compared to the RX56, which can OC to outperform the GTX 1080.

If the 1070ti slots between the 1070 and 1080 it's going to lose against the RX56 OC / AIB cards, and eat away at the 1080 market netting Nvidia less money.

The 1070ti is pointless.

A 1080si that slotted in between the 1080 and 1080ti might be fun

 

1070TI would be correct in laptops. Instead of the 1080Max-Q Scam. But would decrease nGreedia's earnings. Hope they loose the battle.

 

That's the crazy thing about Nvidia, they are in slo mo lose mode lately.

They are too late with the 1070ti idea, it should have been the solution used instead of the 1080 Max-Q, sheesh!!

The 1080si idea would be killer in a laptop or desktop, a whole step up between the 1080 and 1080ti, lots more room for a full step up between those two than between the 1070 and 1080!!

!!'s are rolling off the keyboard thinking about this!!

Nvidia you could have had 2 great products, a "1070ti" and "1080si" instead of one stinky Max-Q 1080!!

 

Just remember guys, there is a chance that between SIGGRAPH and Q4, we might see a consumer Navi. If we do, I will buy that up, as a 7nm over a 12nm Volta. But, without more, I'll just pass on Pascal...

 

From what I see i5-8400 6 Core (182$) actually attacks BIG on Ryzen 1600 6 Core (220$) IF you can buy one for suggested price. Intel is waaaay faster gaming-wise AND cheaper! Although Ryzen comes with Cooler Intel has iGPU! So what leaves to AMD? Charity for competition and arguable overclocking to 4GHz which still isn't enough? Maybe cheaper motherboards and arguable hope for future games too buuuuut... you have something to think about. Good that Zen+ is coming soon.

 

Awesome Hardware #0124-A
48:45 AMD RX Vega finally supported by Afterburner.

AMD RX Vega finally supported by Afterburner
MSI Afterburner Beta gets voltage control for RX Vega and GTX 1070 Ti
The latest Beta of Afterburner brings Vega overclocking support.
https://videocardz.com/newz/msi-afterburner-beta-gets-voltage-control-for-rx-vega-and-gtx-1070-ti

"The Beta 19 of Afterburner 4.4.0 (the last version before going into 4.5.0 branch) has new low-level access to SMC microcontroller, which gives you the full control over Vega voltage. Of course, it works both ways, you can also undervolt the GPU. Although there is no per P-state control, a simple slider is working as intended.

This means that alongside AMD’s Wattman, Sapphire’s TRIX and ASUS’ GPU Tweak you have a fourth option for Vega overclocking.

I took Vega for a quick spin with Afterburner and the clock speed is similar to what I observed with my own Wattman overclocking. Below you can see my settings applied, also notice new GPU power consumption graph:"

Source: Guru3D

 

No Intel CPUs with "Vega Inside" are coming soon
Vega is an Intel employee, not AMD's GPU architecture
https://www.overclock3d.net/news/cpu_mainboard/no_intel_cpus_with_vega_inside_are_coming_soon/1

" No Intel CPUs with "Vega Inside" are coming soon

Yesterday the following image leaked online, with Intel's corporate styling and the term "Vega Inside". This slide sent the internet on fire with new rumors that Intel would be licensing GPU tech from AMD, despite recent confirmations to the contrary.

The usual rumor mills alleged that Intel was creating a mobile CPU with a Vega GPU to a future mobile CPU, given the use of the term "mobile performance" after the term "Vega Inside", though the truth behind the image is something that is interesting, but has absolutely nothing to do with any of these recent rumors. "

"This image is from an employee appreciation campaign from Intel, which highlights the achievements of Intel's staff by using the standard Intel "X Inside, Y Outside" marketing formula. See an example of this below.

So what about Vega? Well, in this case, the image is about an Intel employee called Vega, who seemingly works as part of Intel's mobile team. This means that this has nothing to do with AMD's Vega architecture, which means that we should not expect Intel CPUs with Gaming-oriented levels of GPU performance anytime soon.

From the start, it was pretty obvious that these rumours were false, as even if Intel licensed AMD's technology they would not market it as such, using their own brand name rather than give their competitor a huge PR boost. We are also yet to see Vega-based APUs from AMD yet, which makes it far too soon for Intel to be preparing their own, again giving these rumors a red flag. "

Is Intel Trolling AMD? - Confirmed

 

Interesting developments in the 8700k reviews with reviewers caught accidentally OC'ing the CPU at stock because the motherboard BIOS defaults enabled Multi-Core OC + All Core Sync, running all the cores at 4.7ghz instead of stock boost.

Dropping the boost clocks down to stock changes the results dramatically, and Ryzen 7 1700x jumps ahead of the 8700K, with Ryzen 5 1600x much closer than original benchmarked.

This appears to be done to pump up the "stock" scores of the 8700k because the "real stock" 8700k scores are below the Ryzen 7 1700x without the OC boost.

AdoredTV video explaining the issue
http://forum.notebookreview.com/thr...coffee-lake-z370.809268/page-17#post-10614718

Jayz2cents explaining the mistake, and retesting the 8700k at "real stock" boost speeds.
http://forum.notebookreview.com/thr...coffee-lake-z370.809268/page-17#post-10614749

 

I was reading about this briefly over on Guru3D. I find it hard to believe that reviewers wouldn't realise this prior to their testing & when they were testing, so to me it's kind of crazy that reviewers missed this! I'm not sure I believe the conspiracy theories though.

 

The hidden motherboard "auto" OC was one problem, and 4 or more of the reviewers got caught out not watching frequencies, power, and other monitoring while running the benchmarks - they completely missed the 8700k OC to 4.7ghz on all 4 cores... amazing.

Jayztwocents got further caught out by messing up the memory OC when doing the Ryzen OC CB and other tests. It's easy to do, the motherboard resetting the memory OC to stock once a failure occurs in the CPU OC and changes are made.

So Jay's results were skewed in both the stock test in Intel's favor, and in the OC test in Intel's favor.
http://forum.notebookreview.com/thr...coffee-lake-z370.809268/page-17#post-10614749

Jay's rerun of the benchmarks came up with the opposite result, this time Ryzen 7 1700x in stock and OC tests scored higher than the Intel 8700k:

 

VRM & PCB Analysis of ASUS ROG Strix Vega 64

 

Amacing to see a 8 core processor can beat a 6 core in benchmarks. Same as Coffee should beat Kaby The way it has to be

 

Yup, after collecting information from a dozen reviews, video's and printed, errors were found, and are still to be found, that skewed the results in Intel's favor.

After catching those mistakes and retesting with actual "stock settings" and actual "OC settings", the results finally make sense.

The Intel 8700k is slower than the AMD 1700x in Cinebench.

Spending time watching / reading a lot of reviews from a wide range of testers, is required to really know what is going on. The top 4 scores were from some of the most trusted reviewers, and they were the most wrong of all the reviewers this time.

It is disappointing to realize what I already knew, you need to collect a lot real measured data - not just one or two reviewers - to get a real world view of what is going on, or going wrong

 

As long we know the IPC, core count and clocks vs. the predecessor... Scores was where they should be. This isn’t magic. Or is it?

 

I suppose on the plus side, most users buying 8700K with such motherboards will automatically see increased performance, so for them it would be 'stock' performance so to speak. Not stock performance for the Intel CPU, but stock performance for their system as a whole. If all motherboards do this, then I suppose there could be an argument that the 'automatic overclocking' is the stock configuration for 8700K. I don't know if motherboards do any kind of 'auto overclocking' on Ryzen, if they do I suppose you could consider that stock performance.

 

Except it's not supposed to be on by default, it's a mistake, per Asus (and?).

The BIOS is supposed to have a default stock setting out of the box with all OC disabled.

Per Asus (and?) that expected the MCE setting and All Core Sync to be disabled by default, but found out it wasn't.

Jay made a point of resetting the BIOS settings using "Optimized Settings", and MCE and All Core Sync was enabled.

It should be fixed with a BIOS update

 

Well, if it's a mistake, and if all motherboards don't do the 'auto overclock' then we certainly don't have an argument for it being called stock performance.

 

The puzzling problem left is why are 5 of the remaining 8700k reviewers scores in the "zone" expected, around 1425cb, and the others are almost 200cb lower?

Is there another "OC setting to be found? Perhaps a misconfiguration on the part of one group of scores or the other?

 

Maybe software related, maybe Windows 10 was doing something in the background, or something else eating up CPU cycles. RAM speed differences, although shouldn't account for a whole 200 points. cache frequency differences - another auto overclocking? Different designs of motherboards sometimes show SMALL differences in CPU performance, but not 200 points.

 

The thing to remember is these are experienced reviewers, that know how to get repeatable reliable scores from their benchmark runs, or mention otherwise, and none did.

Check out the same reviewers 7700k scores, they are all consistent around the same average. The exact same reviewers were able to get self consistent and overall consistent scores doing the same test on the 7700k.

Small differences from such random interference averaged over a few runs is expected, but not almost 200cb difference between top and bottom score, with several supporting declining points in between, stand out as problematic.

If you didn't watch the AdoredTV video yet, he's got a few solid ideas as to where the problem is, the silicon.

The Great Coffee Lake Con Job

 

Yeah, it's mysterious. I'll take a look at that later, off to work!

 

Gamers Nexus also made the mistake of leaving MCE on, and followed up with this video explaining the mistake and why it might be "accidentally" left enabled (Auto = ON!!) by a motherboard vendor

Explaining Coffee Lake Turbo Boost Variance & Multi-Core Enhancement


Explaining How Coffee Lake Turbo Works (8700K, 8600K)
By Steve Burke Published October 06, 2017 at 3:05 pm
https://www.gamersnexus.net/guides/3077-explaining-coffee-lake-turbo-8700k-8600k

 

From the link in the quoted reply:

Agreed 100%.

Effectively; this is why I don't overclock any more either (stable - initially - or not).

 

It was also mentioned that a couple of applications, Adobe video processing jobs and Blender in particular weren't stable on the 4x 4.7ghz as given by the "default" MCE option in the motherboard BIOS, additional tuning was required to get stable operation.

It's not a good idea to put any OC into the BIOS as a default given the variance of CPU's and the vagaries of application demands. Certainly not a 4x 400mhz OC without going through the process the BIOS auto tuning offers - that's the confusing part.

If Asus were truly trying to offer a benefit to owners by OC'ing as a default, then they would have initiated their long time available optional performance auto-tuning for higher stable performance, instead of sneaking in the OC through defaulting MCE Enabled - Auto is as good as Enabled.

It makes me wonder if some engineer did some benchmarking, found the Ryzen 7 CPU's were just beating the 8700k, and decided to find out how much fudge was required to turn the tables and make the 8700k benchmark just past the Ryzen 7 CPU's.

 

New Review of AMD Ryzen Threadripper 1950X Planet 3DNow! (de)

"Overclocking finally gave us a little surprise for us. Lower values than set 0.9875 volts did not start our system. And this voltage already suffices for 3.4 GHz, so that an underclocking of our 1950X makes no sense for efficiency reasons. But also when overclocking we had to realize that there is no real sweetspot. Any increase in the clock rate requires more extra voltage than the previous clock step. The necessary additional voltages are still relatively small up to 3.6 GHz, but the curve increases significantly more steeply. However, compared to the standard power consumption, a value of about 3,750 MHz could be considered a sweetspot. For about this clock rate would consume as much as without OC with activated turbo mode"

 

Agree with all but your last paragraph.

If otherwise competent over clockers missed this - the Asus engineers could too.

 

On a side-note.
My Asus ROG GL702ZC will be arriving on Monday.
Ryzen 1700, RX 580 (4GB DDR5), 16GB RAM, 256GB SSD, 1TB HDD.
Will be interesting to play around with the machine when I get it and see if I can undervolt the RX 580 from the stock baseline and achieve better thermals.
I will also try to see if I can push the 1700 lower in terms of voltages too (though Ryzen is already pretty efficient as is - but the manuf. process and default voltages aren't that good for GPU's).

As for 8700K and an Asus engineer intentionally overclocking the CPU to beat Ryzen... well, it's a possibility, but I think it's also likely to be a simple mistake.
Correct me if I'm wrong, but only that one Asus Motherboard type was affected, and apparently, when Jay addressed this issue with Asus, they insisted that the setting was shipped as 'off' (indicating it could be a simple mistake or are lying intentionally to cover themselves - again, possible, but I doubt they would be doing this intentionally and risk overheating and throttling).
Asus might be releasing a patch in that case to address the setting and make sure its switched off... or update the BIOS settings on unreleased mobo's to turn it off.

Either way, most people getting the CPU in question on that motherboard (if they get the overclocked setting turned on) might end up frying their CPU's or experience thermal issues as a result because they are being run on too high voltages and of course produce much more heat than intended.

Personally, I would try to undervolt and overclock the CPU instead... or just undervolt it. It already runs fast enough and probably won't be an issue even on stock.

 

"It makes me wonder if some engineer did some benchmarking, found the Ryzen 7 CPU's were just beating the 8700k, and decided to find out how much fudge was required to turn the tables and make the 8700k benchmark just past the Ryzen 7 CPU's. "

Maybe it wasn't the 8700k, but at some point many years ago, some bright engineer dreamt up getting Asus a free motherboard performance boost to outperform the competition... at least until they copied the idea.

It may just be that everyone (4? reviewers) "forgot" that Asus defaults MCE to "Auto=Enable", in every(?) motherboard BIOS release??

The first time I noticed it was with my early Rampage motherboards.

A couple of examples:

2012 - Multicore Enhancement: The Debate about Free MHz
https://www.anandtech.com/show/6214/multicore-enhancement-the-debate-about-free-mhz

2016 - asus multicore enhancement
https://rog.asus.com/forum/showthread.php?83251-asus-multicore-enhancement

 

Wow, how exciting!!

Where did you end up ordering from? Good price? A little more or less than you expected / MSRP?

Very much looking forward to seeing how it runs for you, overclocking, temps, fun factor

IDK how Asus is going to handle it since MCE has "always" been there, and we know to turn it off when testing stock results, and when you OC it's disabled, it's only for a "stock boost".

The other vendors have something like it too, so you need to check any and every motherboard BIOS settings carefully before publishing benchmark results at stock.

See my other reply to @tilleroftheearth

Gigabyte motherboards seem to get bit by this frequently, and it's likely related to the "stock boost" mentality, checking all the BIOS settings is something I always do, boring as it might be until you learn to enjoy it, as there are often little settings "gotchas".

Yup, it's funny to remember it was long ago that the "method" was to use the highest recommended voltage of Intel for the CPU / motherboard vendor, and overclock from there.

It took a while for a few of us to independently discover "undervolting", simply known back then as "backing off from the maximum voltage".

I can't believe the motherboard vendors still don't automate all of this tuning as every CPU is different, and they can't expect the owner to do this time consuming task. Some of us enjoy it, but most just want it to work at peak performance so they can get "work" / "play" done.

 

We will see a lot fun. From your own post