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

Vega 56 Hybrid at 400W: Bypassing AMD’s Artificial Limit (1742MHz)

Is AMD RX VEGA REALLY better for video encoding? We didn't expect this...

RX Vega 56 Flashed to Vega 64 Bios

AMD Launch 1900X With Benchmarks | Raja Koduri on Vega Price & Performance

 

@Rage Set

My Failed EK Supremacy EVO Threadripper Edition Review

 

This is without trying and minor tweaking of the memory. When I have time in a couple weeks, then I'll properly tune this bad boy! (Also, this is on my work OS, not my bench OS, which scored over 3300, barely, at 3.8GHz).

 

I wonder if AMD will manage/really want hold same prices as Nvidia now will demand higher prices due Memory shortage. Can AMD now finally start increasing the market share? Aka don't follow Nvidia regarding prices?

At least Nvidia is pushed to increase the prices.

 

Man, that guy sounds like he's having a hard time!

 

Hahaha! I watched the video yesterday and he looked like he wanted to cry/scream at the end of that video. I was using the same AIO cooler on my 1950X but I'm about to run my custom loop.

To be honest, I have had some of the custom water loops in the past (pump) crap out on my clients. Many things could affect the pump and i am sure none of my clients actually maintain their loops like I tell them. I have even created timed e-mail messages that are sent out in six month intervals (the first says check the water level and the second tells them to flush/change the fluid). No one listens. I personally never had a pump in either an AIO or custom loop die on me. *Knocks on wood*

That's great bro! What cooler are you using? What are your temps like?

 

I'm using a TT Water 3.0 Ultimate (360mm) AIO with 6x Noctua iPPC 3000rpm in push/pull with Thermal Grizzly Kryonaut applied as 1 big blog (not lines or dots, just one massive blob that literally spread to the edge of the coldplate on the AIO). I'm getting high 50s running CB15. My fans go to full at 55C on die. So it is keeping this baby cool until I get it under water properly.

 

I've had one AIO fail on me, but I'm not convinced the problem was with the pump itself, rather a short in the power cable.

 

On a Vega related topic:
https://www.pcgamesn.com/amd/amd-vega-gpu-specifications

It seems that IF was optimized for servers, but not games. And the article goes on to establish other features other than IF weren't properly optimized, or activated.

We will have to wait and see how this pans out with future driver releases and developer optimizations.

 

IF=??

 

Infinity Fabric

 

Seems so obvious now...

To me at first it was a big "IF", with a lifetime of alternative meanings.

Maybe write out infinity fabric from now on, "IF" is a bit confusing even in context. Even this wikichip article only uses "IF" in the title and 1 time in the text, otherwise infinity is spelled out 6 times in the entry, 9 times for fabric.

Infinity Fabric (IF) - AMD
https://en.wikichip.org/wiki/amd/infinity_fabric

 

Rage Red is officially under custom water (CPU). On idle with stock clocks, it's a cool 29C (down from 37C from the AIO). Full load, it never goes above 56C (down from 73C).

60mm Radiator (360mm) with three EK Vadar (1850RPM) in push. Not to overwhelm the thread with photos, I'll post them on imgur and post a link.

NOTE: My build is still a work in progress. I have no other case fans installed and my cable management is not done. I am currently on the leak/burn in stage.

 

Great temps, so it's a 3 fan 360mm radiator? Seems the 2 fan models aren't quite enough for either the ThreadRipper or RX64.

Are you going to add in the GPU(s)into that loop, or add another loop with another radiator?

 

Just fixed it, 360mm radiator. The GPU(s) will go in its own loop. I stopped creating single loop builds awhile ago due to an incident.

 

Yes, I've had an incident as well, was put off water for a while, was hoping someone would come out with 3m Novec or some other non-conductive coolant for consumer use:
http://www.3m.com/3M/en_US/novec-us/applications/immersion-cooling/
http://www.3m.com/3M/en_US/novec-us/

AIO coolers have had a good non-leaking track record for a while, but now the cost reductions for pump motor's seems to have caught up with failure modes, and those are problematic now.

Any word on non-conductive coolant that stays non-conductive for custom loops?

I thought this used to say non-conductive, now it says "low" conductive

PrimoChill ICE Low-Conductive Coolant
https://www.primochill.com/collections/ice
"WARNING: Use of this product with aluminum components (such as aluminum constructed radiators) is STRONGLY DISCOURAGED and will void the fluid warranty. PrimoChill will not support or replace fluids used with an aluminum radiator."
https://www.primochill.com/collections/triple-fan?sort_by=price-ascending

 

Other photos at the link below.

https://imgur.com/a/mb4dx

 

When I was a beginner, I've done a lot of research and all of the non-conductive coolants I read about become conductive over time. It is the nature of the beast. I check all of the fittings in my personal builds every two weeks; it has become a routine for me but I understand a lot of people don't have the time.

 

What multiplier are you using?

 

Yeah, I've gone air cooled Noctua since, and haven't had any issues, getting a really good AIB GPU with it's own cooling, and many quiet low RPM fans all around, and that's worked great.

The ThreadRipper Noctua air-cooling looks good, but nowhere near as good as a custom loop cooler, same of course for the RX 64 / 56.

It will be interesting to see how the AIB cards tame the heat with the RX 56 / 64, both air and water cooled.

 

36x. Right now I'm running 38.5x and I'm maxing out at 62.5C running AIDA64 stress test - CPU/FPU/cache.

Ambient temps: 21C. Later tonight when it drops in to mid 30's (degrees), I'm going to push this mofo!

 

What turned you off custom loops? I avoid buying the Chinese knock-offs on eBay and Amazon. I stick with EKWB and Bitspower. Even with those brands, I'm always checking and rechecking.

I think you should give it another go. Build a small custom loop, nothing fancy. Plus if you are going with RX Vega, you near damn need a waterblock, hahaha.

 

There's been no need for running cooler than the air-cooled solutions. I've been able to reach max OC and run long jobs without thermal throttling.

If I had needed better cooling to solve thermal issues I would have gotten back into water-cooling, but so far nothing has thermally challenged the continuing improvements in air-cooling.

Usually I run what clients have professionally, and run what family and friends run personally so I can support them with the expertise gained.

Professionally DC is air-cooling, and personally air is much easier for people to accept than water, I've tried both and there's less stress with air cooling.

Although I've had full case fan failures (bad batch) on air more often than pump failures on water, but that's likely due to the predominance of using air-cooling.

ThreadRipper and Vega + future high temp CPU / GPU's are going to insist on water-cooling, so I'm going to need to get back into it again at some point.

Not yet though, I'll watch you and the rest of the guys here implement theirs first and get a good idea as to what's working best.

 

I do not think LN2 is electrically conductive.

 

We definitely need some cutting edge development for heat transfer and exchange that is easy to use and cost effective.

Phase Change is loud, Peltier has condensation issues along with LN2 and other ultra low temp cooling, water shorts out components, even non-conductive coolant eventually picks up particles that make it "low-conductivity".

Maybe some new manufacturing techniques to seal off the components from water cooling? Find new lining materials for piping that won't shed particles?

Small electromagnetic bottles to pump heat into? Replace them to reclaim the heat and put in new "empty" bottles?

Something's gotta show up to save us from a growing desktop / laptop thermal gap.

 

As far as a medium for a liquid cooler has anyone tried DOT 3 Castrol brake fluid?

 

There is ZERO chance of closing that gap. What you want is hydrophobic research, which we've advanced greatly in recent years. But, once we shift to graphene (they found a way to pace the signal, which was the issue before), we are talking THz range, not GHz (within the decade, potentially). At that point, all of our convos about single threaded, etc., are moot. It's the only thing with processing on the level of quantum computing, but is built on traditional techniques.

I'm sure you guys have seen China and Japan's work on quantum satellites, which teleport data and have different encryption. So, we will have a leap soon in tech, same with photonic ram, light interposers, etc.

 

So, I currently have similar temps (62C) on AIDA64 stress test for 3.8. So nice to see I'm not far off of your 360 custom loop with an AIO with bad coverage. But, our builds are early yet, and 50MHz on this platform is nothing to sneeze at.

 

3 materials could have been used for that purpose:
1. Carbon nanotubes
2. Synthetic diamonds
3. Graphene

They are different variations of carbon yes, but some merging of synthetic diamonds and graphene in those states seemed to have produced very nice results in terms of cooling, waste heat recycling, etc. (for quantum computers).

There has been that little lung-like device that simulates 'breathing' and it was demonstrated years ago for laptop use.
This hasn't been implemented at all. Plus, we've seen those static oval shaped coolers for example without fans... but I've only seen very large versions for in-house use in bedrooms, etc. here in UK for sale - hadn't seen anything like that laptop sized (though the initial prototype if I'm not mistaken was that size).

So... we're looking at multi-factoral approach...
If the hardware itself moved from silicon to synthetic diamonds, CNT and graphene for basis of chip construction.... or at least created a hybrid solution where connection points sensitive to heat were replaced with those materials, we'd already see a massive drop in temperatures and probably improved efficiency.
A hybrid might need some kind of cooling, but a new hardware chip made out of those 3 materials in combination probably wouldn't need cooling to begin with.

Point is, passive cooling with high-performing hardware in a slim laptop for example is more than doable... problem is, the system doesn't allow it because its not deemed 'cost effective'.
Technologically and resource-wise, it can be done sustainably in abundance - but sustainability, longevity and modularity are not exactly friends of the current economic structure.
Laptop manufacturers make it as hard as possible these days to upgrade even the simple stuff like HDD/SSD and RAM... nevermind the CPU or the GPU, because they can make access to components needlessly complex.

But if we are looking for the near interim... something that OEM's could integrate now... that lung like device and oval fanless cooler might do the trick - especially if they are placed in different parts of the laptop.
Combine it with graphene laced copper pipes (not sure if the two tried to be combined yet) and contact points for dies, and you could potentially improve heat transfer and reduce the cooling footprint.

Maybe a small-sized liquid cooling for the laptops?
I think they already suggested it for alienware laptops before... but this never came to fruition, and they mentioned that OEM's would charge a large amount for that implementation.

We know new technology is not necessarily cost prohibitive to mass produce... it's cost effective for OEM's, but they LOVE to slap a premium price on it just because it's 'new technology' (no other reason).

 

Here's some news about Lenovo unveiling its new line of all AMD laptops (Ryzen based) :
http://www.itproportal.com/news/lenovo-unveils-all-new-amd-powered-business-laptops/

This looks like Ryzen APU's will be used along with Vega as an iGP.
Wonder what they will cost though.

 

AMD's doing well! https://www.notebookcheck.net/AMD-p...st-online-PC-component-retailer.246204.0.html

 

I thought his results were a bit off, an engineering sample explains it.

 

Ive always liked amd. No hate for Intel, just always used amd. Glad to see them coming out with good products that aren't way behind in performance anymore

 

Intel had set themselves up for this. If Mores law continued since the early iCore days the compute speeds and IPC would be so high AMD would have stood little chance to catch back up. Since Intel got lazy and started to just soak the public for little to no advancement from generation to the next now they will have to pay.

Now it comes to another point. That is, other than gaming, how much more of a CPU is needed? If a 6 core Ryzen 4.0 GHz satisfies 95% of a home PC's needs who needs to spend 2x that for an Intel system just because it is faster? Who cares if for a few hundred more the page loads in 1/2 a blink of an eye compared to taking the entire blink?

 

Have we reached a point in our daily computing tasks that the last few generation of CPU's match/exceed our needs? Yes. At this point, if you setup two computers, one with a Ryzen CPU and one with an Intel proc from the past three generations, you wouldn't complain any of the the computers are slow (unless you have an HDD as the OS drive). However, there will always be people that will spend, no matter the cost, for the fastest components available. I have one half of myself in that group in regards to GPU's.

With my TR purchase, I can see it moving from one use case to another in the future. It is my personal computer and it has officially taken over work duties from my main X99 workstation.

 

That is the point, yes there will always be people but they are not the majority nor do they overall spend the most. AMD can afford to leave the niche market for Intel to rule.

 

Links;
https://www.hardocp.com/article/2017/09/03/amd_threadripper_gooseberry_hedt_world_record

 

Firestrike: 21,089
https://www.3dmark.com/fs/13534126

My goal is 22k+ with a single 1080. SLI is currently not working properly on this board. I only got 27k, should be around 30k.

 

What is your Geekbench 3 score?

 

https://browser.geekbench.com/geekbench3/8440041

 

Nice score. In line with quite a few others. Mine is [email protected] (got higher last night, but on beta fall creators edition -61600).

 

http://browser.geekbench.com/geekbench3/compare/8439627?baseline=8438403

 

So yeah, something is definitely off with SLI on the TR4 platform. Here's a comparison between the top 2 TR 1950X results and my 6850K rig.

https://www.3dmark.com/compare/fs/13535777/fs/13539033/fs/13051330

I am not sure if Asus (and others) need to work on the BIOS more or it is the benchmark itself. I should be well into 30K+ with this build.

 

Hahaha, bro you've earned a rival. I'll catch you! Amazing score!

 

It's both. You are one of many complaining about sli currently. Asus is aware and trying to figure it out...

 

Love it! Helping people with scores in the anandtech forum right now, too. But remember, I have a dedicated bench OS without anything except benching software and utilities, which helps keep overhead to a minimal! Got CB15 of [email protected], 3333mhz ram cl14 1T, with 3 word documents open (one is a couple hundred page book), an excel spreadsheet, an ebook reader with textbook open, both chrome and Firefox open, steam and onedrive in the tray. ;-)

 

I hope so. I have Asus in all of my personal builds and I even push the brand to clients, friends and family (if they have the money). I have been down this road many times with Asus but selling a "premium" board at a premium price, SLI should have been one of the first things working (better) out of the box. Looking over the TR4 results, it isn't just Asus that has an SLI problem. It appears that all of the TR4 mobos don't like SLI currently.

I may be looking at different mobo brands when it comes time to build my i9 rig.