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

That would be very disappoint being as there are 1800x' systems well over their 20,000 in multi. Either links the number are horrid.

https://browser.primatelabs.com/v4/cpu/search?utf8=✓&q=1800x

 

Some are calling it fake. Some are saying it isn't recognizing the second die. Some are saying bios and scheduling aren't working right. Many of those seem plausible.

Sent from my SM-G900P using Tapatalk

 

One of the reasons AMD should have been on top of the benchmarks. The best way to dispute leaks is to put out the real ones and if there are issues to have them addressed. Anyone, including myself, going by these numbers will just stay far away from TR or even the thought of it.

 

I'll wait for confirmation, as I wouldn't buy less than 12 cores, so it is until August anyways. But, if confirmed...

Sent from my SM-G900P using Tapatalk

 

I am just saying, AMD had best hop on this one ASAP.

 

I agree with @TANWare, AMD needs to come out and state those are fake scores. Even though anyone with some kind of common sense/logic would see those are fake.

 

Not sure why they wouldn't, it could only result in more sales.

 

The problem is AMD is being so secretive. In a way, they are damned if they do, damned if they don't. If they do come out and publicly address the fake scores, people are going to demand more information on TR. If they don't address the fake scores, some are going to think the scores are legit and go with Intel.

 

Correct, they will rightfully so demand more info. They were promising more info at computex and failed miserably. way past the time to start giving it up.

 

Exactly why they should be pretty much hemorrhaging real information right now. They're going to shoot themselves in the foot if they don't.

 

Geez, what is up with you guys freakin out over not havng information "leaked" before a release?

Knowing that AMD should have let Ryzen bake a few more weeks / months before releasing due to the BIOS issues upon release.

Adding pressure to release information on Threadripper before it's ready is ludicrous, stop it.

Intel won't even have anything near 16 cores in i9 for a ***long*** time means AMD can take their time and get it right before releasing.

Relax, play a game, stop obsessing on ThreadRipper, you'll have an aneurysm and miss the release date completely.

 

Actually, one correction. Intel announced the 14, 16, and 18 core for October. Just saying, they moved things around which changes the meaning of long.

Sent from my SM-G900P using Tapatalk

 

The problem, from my opinion, is they were misleading the media about Comutex release information. The media then misled us too, and then almost nothing. FYI I do not want leaked info, I want the real deal. We end up taking leaked info as that is all we are getting.

 

Seems suspicious for Intel to make that massive of a schedule change, doesn't it? I can't think of a single time in Intel's history where they have shaved that much time off of a delivery schedule, and made the dates

 

All I can tell from this is that you are easily misled

I don't take any rumor's or off hand comments by people that aren't official as promises, unless they are followed up with official announcements, of which there were none.

So, like I said before we should all lighten up and stop scraping for slivers of what look like shiny information leaks, and enjoy our lives.

AMD and Intel will officially ship when they ship, not before

 

LOL, pot calling the kettle black.

1.) I can easily see when something is wrong so I am not asilly misled, on that same note I am not the one to say what is wrong without it in hand. I am also not the one leaving myself open to false information.

2.) right out when it is out, and that is when?

Edit and links;

 

All I can recommend for you is to stop reading rumors, and find something else to occupy your time, and then when you see it on the shelf, check it all out again.

You seem a bit too mercurial when jumping from rumor claim to rumor claim, and it's affecting your opinion in a wide range.

There's really nothing solid to go on right now, so it's all a waste of time.

 

Speculation is half the fun and Intel HAS released information on what they are launching in a week. AMD could have at least gave us information on the 10c/20t TR chip to counter Intel.

In my opinion, the HEDT market has always been about performance first, value second. There is nothing "budget" about spending 2K+ on a HEDT platform and AMD has to deliver on the first point. But @hmscott, you're right. Regardless of the amount of poking and shouting we do, AMD and Intel will ship their platforms when they are ready. The question I ask you, will the majority of HEDT consumers care when AMD does ship?

 

Makes me wonder if they replaced fab time for coffee lake to get those done quicker (and who knows what else). If they used 14nm++, though, instead of 14nm+, which is what currently is used on SL-X, those HCC chips will be incredible and I'd have to revise everything. But, if they did that, there would be zero reason to go to CL-X next year. So, we'll find out soon enough.

Sent from my SM-G900P using Tapatalk

 

Yes, but how well Epyc scales, and information on ram support, achieved clocks, server motherboards, etc., comes in a week (or two for reviews to drop). This may change for information and what is said. So we'll know soon something, like whether they were silent for no reason.

Sent from my SM-G900P using Tapatalk

 

Given Intel's antics and people asking "what's up with Intel?" and the answer is "AMD!! is what's up with Intel", I think Intel has done a great job of raising market awareness for AMD's offerings.

If it weren't for the success of Ryzen it might take longer for Enterprise clients to wake up to AMD's presence.

Now that AMD's return is well known, AMD has to start the long road back into the evaluation and acceptance cycles for large clients.

Given how clients are always on the look out for technical advantages over competitors, especially when it comes to saving money, saving space, saving power and cooling, if AMD can show benefits in all those categories they have a good chance of coming back quickly, and big.

 

AMD Ryzen Threadripper Launches around 10th of August-Guru3d.com

"From what we have heard from sources close to KitGuru, AMD should launch Threadripper around the 10th of August, after moving the date from a late July launch. This is what we have heard through the channel and if true, would mean we will be seeing AMD and Intel square up once again, this time in the HEDT market."

"So that means Threadripper and Skylake-X i9-7920X (the 12 core part) would launch in the same month.

 

Yet another push, and it is no where near the end of July. Then they have to make it to market and get it in our hands. Since supposedly Intel will be later this or next year I could see AMD delaying yet again. Ugh..............

 

Links;
https://www.lowyat.net/2017/134180/rumour-amd-may-officially-launch-threadripper-cpus-august/
http://www.frostytech.com/permalink.cfm?NewsID=130416
I guess being a fake hasn't caught on;
http://segmentnext.com/2017/06/14/amd-ryzen-threadripper-1950x/

 

From what I'm hearing in my little world and like me, take it with some salt, AMD is holding back to ensure the availability of mobos. They do not want a repeat of Ryzen 7.

 

Those GeekBench numbers are back again, and not looking good.

 

I hope that turns out to be true. Maybe they did learn that 'first' means squat at the high end.

Even better; I hope they don't depend on 'optimizations' of their M/B's, O/S' or major programs that may or may not come in the next six months or more..

The release should be more than 90% of what TR is actually capable of (real world; not synthetic 'scores'). Not 70% and hoping that we'll see up to ~42% improvement over the next six plus months...

If TR releases at below ~85% in real world productivity of it's marketing so far? Pass...

Ryzen 7 went as well as I expected AMD to deliver - TR though is in a different league.

 

To me it's not about advance information, it's about correcting bad information. Sit on no information all day/week/month but bad information needs to be countered.

 

The company officially said summer. The late July number came from the AW information. Two weeks later isn't like Intel regularly pushing quarters recently. Now, they did push Vega a month for consumers, so that compounds any news of a delay, but that also was due to HBM2 availability. Meanwhile, look at Intel's info on releases, changing by the day, yet you don't attack that.

That would make sense with how that screwed them in March and MB manufacturers giving more to Intel releases over AMD.

Sent from my SM-G900P using Tapatalk

 

And again, a 1800x at 3.6 Ghz scores much higher in single threaded and multithreaded in Geekbench.
I can see the single threaded performance being right... but not Multithreading.
It's fishy to say the least.
How can a 8c 16th of the same architecture score better than 16c 32th in multithreaded tasks with about 200 mhz difference?

 

Wccftech took their leaked benchmark story from the other day down; http://wccftech.com/amd-ryzen-threadripper-1950x-specs-performance-leak/

Edit; I am not saying they are real, they are back again.

links (not related to leaked benchmark);
http://www.isportstimes.com/article...ng-this-august-faceoff-with-intel-skylake.htm

 

After updating AGESA and a few days of testing and playing all is good atleast with my memory at xmp 3200 14,14,14,34 and no errors in TM5. Pretty nice considering I could not even boot at those settings with the previous bios update.

 

That's good, you've got tight timings there too, are you CR1 or 2, and what's your RAM voltage?

 

Which could easily suggest the numbers that 'returned' in the new link are just regurgitated from say another source.

Also, something is amiss with that Intel slide.
A 6c/12th CPU has a base clock of 3.5Ghz but boosts to 4 Ghz... while apparently 8c and above seem to boost to 4.5Ghz.

Are these numbers correct? Have they been verified?
The Xeons for example aren't boosting that high if I'm not mistaken and would be lucky if a single core boosts to above 4Ghz, and (again, if I'm not mistaken) the i9's would be based on those CPU's.
Intel would be lucky to get them above certain clocks...

Taking from Intel's own product webpage... this particular xeon is clocked relatively low:
https://www.intel.co.uk/content/www/uk/en/products/processors/xeon/e5-processors/e5-2695-v4.html

and it boost to 3.3 ghz.
Granted, the TDP is 120W (as opposed to the i9's 165W), so that would give intel wiggling room to boost the clocks... but would 45W be enough 'wiggling room' to reach 4.5 Ghz boost?
We're talking about roughly 36% increase in turbo boost clock speeds for just above 37% increase in TDP?

Even intel suffers from massive thermal inefficiency when clocking higher and higher... I'm just wondering if they were able to surpass this limitation somehow to keep the 4.5Ghz at 165W range (which seems... extremely improbable).

Edit: it seems that same slide suggest 14nm+ process.
Could a refined process grant so much performance and efficiency gains?
I can see 14nm+ giving an increase in clock speeds (by several hundred MhZ - maybe 500 Mhz) while retaining TDP at same levels, or raising them just a bit. But I doubt even Intel can keep all their cores (16 of them no less) running at 4.5 Ghz on 14nm+ at 165W.
That still sounds improbable - unless the 14nm+ process offers a near increase that would happen transitioning from one manuf. process to another for example.

In which case, Ryzen at 14nm+ might actually be in the same ballpark, if not better.

 

We will see how good Epyc is soon enough. AMD's push is real, at least for enterprise. MiT's are boring most of the time, though.

 

Giving vendors the time they need is important, to make sure they have time to get their hardware testing and ready for release.

As for the CPU itself, although more testing time is good, it's inevitable that users will run things in ways you couldn't imagine when making test plans and procedures.

Intel is adding Errata long after release as odd cases are discovered. Which is one of the reasons to keep installing BIOS and Windows chipset updates even if the vendor doesn't list a good set of "reasons", that's how the errata fixes get released.

One should always expect the unexpected with new hardware releases, this is all new stuff from AMD so there are bound to be issues.

Getting all freaky about these inevitable new product issues is an unprofessional response, usually only seen in the consumer realm.

In professional circles you note the issues, roll out the fixes, and keep looking forward. If you are in an evaluation and acceptance period, you might extend it based on findings soon after release.

AMD knows all this and isn't going to hold back a release unless there are found and unfixed issues, baking longer - past planned testing - is wasting product life cycle time and reducing revenue.

 

And TR is a .................

 

A consumer product from a price point perhaps, but not really in the budget range of MOST consumers.
A 1700 may be the upper budget limit of most consumers (heck, 1600 or 1600x might be a better estimate for that).

TR would be for rendering machines and massive workloads... granted, it would be GREAT to have for longevity sake and rendering indeed, but I think an 8c/16th would do the trick for now at least.

 

The professionals wouldn't get freaky about their Consumer level CPU's either; they use those on their office desktops too.

My comment wasn't based on product market, it was a comment in general that these outlandish swings of disappointment and glee aren't professional, in general.

 

Not just a price point. x399 and x299 are being targeted to consumers. This is shown as with the Alienware systems. You just may not realize the consumer market is expanding. Software may not have caught up to the trend but as mentioned before this was true of the market shift from 2 to four core with 8 threads.

And yes these will take place of quite a few Xeon systems purchased for real workloads but that is always the case, just not to the extent here.

Edit; When I look at this info I do so as a consumer. I am going to do this purchase as a consumer not as in purchasing systems for an office.

 

You shouldn't get upset over AMD having lower clocks over Intel.
You should be questioning the validity of those Intel slides in the first place and the numbers they propose.

In regards to 14nm+
Correct me if I'm wrong, but that's the same process Intel uses for 7700k (it also explains why its able to clock higher than Ryzen)
Great... we have a baseline for the i9's.

And again, I doubt that i9's will be clocked as high as those slides suggest.
Unless those numbers are verified, I'd be VERY skeptical.
Look at how much power 7700k sucks up at 4c/8th with 4.5 GhZ turbo.
It's rated at 90W TDP.

Can they actually offer 165W 16c/32th CPU at 165W while boosting to 4.5 Ghz?
Is it realistic?
Right now (and unless that slide was verified), given what we know of their current Xeon line-up and what they did with 14nm+, I'd have to say it seems unlikely at best.

Again, I could easily be wrong as my info on this is incomplete.

 

I think the 12 core up all the numbers are questionable. I saw somewhere that the 18 core variant may be looking at turbo 3 of 4.0 or 4.1 GHz only. In other words a declining clock as the cores increase to try and tame the TDP beast. Again we could be wrong here as well but it would explain why they want 18 cored to kill the TR as the increased clock speed may not be able to be counted upon.

 

I looked at it a bit further. It seems Geekbench itself is horrid for scaling though multiple cores, as the count rises, especially under W64, the scores go down. just look at the link below and this CPU is well known for running above the base speed.

https://browser.primatelabs.com/v4/cpu/2149371

 

Looking at Xeon's scaling, this is exactly what we see.
As the core count increases, clock speeds go down to keep TDP in check.
It would likely be the same with AMD - though how much their high density library design comes into play for example along with infinity fabric (which is different compared to what Intel is doing) - we don't know how this will affect clock rates and final TDP.

Now, we don't even have official clock rates for TR.
Could 16c/32th boost to say 3.8 or 3.9 GhZ?
What about TDP?
Do we have official figures from AMD on TR at all or, are numbers released thus far speculative at best?

Given what we know of Ryzen and Xeons...
i9's might be clocked slightly higher in comparison to TR, but the difference might not be large overall.
At 14nm, AMD might be at a bit of disadvantage clock-wise... however, with 14nm+, it will likely be able to clock as high, or possibly higher than i9's.

 

On the 1998x they had some pretty high claims. This may be the reason for the 1950x as the 1998x claims were not being met, at least these chips.

 

Please note my speculation and reaction are per my consumer side and are completely separate from any and all professional relationships I may or may not have. I don't comment on topics that may cause issues professionally. Nonetheless, I am waiting on AMD's TR with an open mind, something I couldn't say a couple of years back.

 

1 and 1.35 Volts

 

Now, how high can you overclock that ram?

Sent from my SM-G900P using Tapatalk

 

That's impressive to get such tight timings at 1.35V. Your Trfc is relatively loose though at 560, and I've read & experienced that having Trfc too loose can negatively impact your RAM performance - a loose Trfc will enable you to get tighter timings on the other variables, but at some point it will start to negatively impact performance; might be worth playing about with Trfc to see performance impact. I have Trfc at 240, and I remember seeing Mr Fox having a really loose Trfc of about 1000 and I remembered thinking/seeing he was getting poor RAM performance in spite of tight other variables - he then lowered Trfc and it brought it more in line with what I was expecting. If I remember rightly, testing with Cinebench 15 as well as the Physics tests in 3DMark was a good way to see diffrences.

 

Links;
http://segmentnext.com/2017/06/15/amd-threadripper-memory-overclocking/
http://www.isportstimes.com/article...aming-desktop-with-amd-ryzen-threadripper.htm

The leak being pushed.
https://segmentnext.com/2017/06/14/amd-ryzen-threadripper-1950x/
https://www.overclock3d.net/news/cp..._threadripper_cpu_has_appeared_on_geekbench/1