Delete this thread plz

[CentroX]

Edit: since everyone is sensitive to negative rumors i request a mod to delete the thread.

Locked per OP request, we don't delete
Markfw
Anandtech Moderator

 

[whm1974]

I guess we have to wait until RyZen CPUs and boards are in the hands of reviewers to know more.

 

[CHADBOGA]

Hopefully this will be easily sorted out, yet at the same time I always am bemused by those who have to buy a new CPU/motherboard as soon as it comes out, instead of wait a week or two or four, to find out where the lay of the land really is.

 

[imported_jjj]

So now you are pushing the rumor that the CPU is defective, with no evidence at all.
This is how fake garbage starts and by tomorrow it's a fact on 400 sites.

 

[lolfail9001]

Well, i was not wrong, AMD could not let this boat sail smoothly.
Also, looks like initial PassMark latency results were not a fluke either.

So now you are pushing the rumor that the CPU is defective, with no evidence at all.

Evidence of memory latency issues was in both PassMark leaks. At least much i consider verified.

 

[lopri]

"Speed is great but latency not so much" <- What does this mean? I thought speed = latency.

"For example the 3200 Mhz DDR4 memory from the testkit doesn't work with 3200 Mhz." <- Since 3200 MHz is not officially supported, this sounds more like the reviewer does not know how to OC. 😉

"The launch seeems to be forced because every mainboard manufacturer is late regarding robust BIOS versions." <- Is this a direct translation or paraphrasing?

"It looks like OC results (and probably other) are improving with new BIOS versions which is annoying from the testers perspective because well, every day your old results might be invalid." <- This is a good new, not a bad news.

"In short if filling all 4 DIMM's set your speed to 2400MHz and work up from there.
If using 2 DIMM's put them in the A2/B2 slots and a max of 3200MHz should be possible." <- The more sticks you use the lower memory OC you get. This is true for every platform that has existed.

Every platform has teething issues in the beginning, especially for brand-new architecture like Zen. Maybe there is a major problem with RyZEN CPU or maybe not. But if there is, I do not see it in the quoted exchanges.

 

[imported_jjj]

Well, i was not wrong, AMD could not let this boat sail smoothly.
Also, looks like initial PassMark latency results were not a fluke either.

Evidence of memory latency issues was in both PassMark leaks. At least much i consider verified.

Software reporting anything is in no way evidence that the silicon is not ok, it's not even evidence that the there are any latency issues.

 

[lolfail9001]

Software reporting anything is in no way evidence that the silicon is not ok, it's not even evidence that the there are any latency issues.

When software that reports numbers in 20s-30s for about every Intel CPU with DDR4 reports 70s for Ryzen QS, something is up. You choose: is it latency, or is it time measurement (when it is clearly working fine in other cases).

 

[imported_jjj]

When software that reports numbers in 20s-30s for about every Intel CPU with DDR4 reports 70s for Ryzen QS, something is up. You choose: is it latency, or is it time measurement (when it is clearly working fine in other cases).

It's working fine with Intel doesn't mean that it will work fine with new hardware. Don't act like you have never seen a bug.
Everybody complains about sites like Wcc and then they go act the same way.

 

[majord]

When software that reports numbers in 20s-30s for about every Intel CPU with DDR4 reports 70s for Ryzen QS, something is up. You choose: is it latency, or is it time measurement (when it is clearly working fine in other cases).

For the record, it reports over 100ns on Bristol Ridge

 

[lolfail9001]

Don't act like you have never seen a bug.

Identify where it happens, then.

For the record, it reports over 100ns on Bristol Ridge

Well, Bristol Ridge performs fairly subpar in general, so i do not see contradictions here.

 

[itsmydamnation]

Well, Bristol Ridge performs fairly subpar in general, so i do not see contradictions here.

Except per clock Bristol ridge is the best performing CON core and piledriver reports latency of around 60ns with 1866 ddr3 memory in ADIA....

 

[tamz_msc]

The Stilt said that the BIOS for Ryzen will be much more complex than before, so these issues are not unexpected. Hopefully they can be sorted out as best as possible by March 2nd, with further improvements after that.

 

[lolfail9001]

Except per clock Bristol ridge is the best performing CON core

Oh, really? Last time i checked AIDA results for Bristol vs Steamroller, it was actually losing. Yes, pure synthetics, but nonetheless.

piledriver reports latency of around 60ns with 1866 ddr3 memory in ADIA....

PassMark and AIDA clearly measure different numbers, even if i think what AIDA reports is closer to reality , that said, however,

Top is Carrizo, bottom is Piledriver.

Post Scriptum: Remember first Naples GB4 leak? Well, that one had terrible mem results too.

 

[A///]

New platform means new bugs. Happens with Intel, too.

 

[CentroX]

So now you are pushing the rumor that the CPU is defective, with no evidence at all.
This is how fake garbage starts and by tomorrow it's a fact on 400 sites.

Arent you a sensitive one. Maybe we only should report happy wccftech rumors and ignore to be objective? For the record i already prepaid 800 dollar for a 1800X and Asus crosshair IV Hero.

 

[coercitiv]

So to recap:

• Zen may have a hardware defect, may require re-manufacturing, but nobody can tell what that defect is or what the symptoms of the defect are.
  
• Zen may have "major" memory speed problems, except quoted problems refer to overclocking, which cannot be major by definition
  
• 6 days before launch without a final BIOS never happened before, except with Broadwell in recent history. Yes, the CPU that had issues with overclocking settings on some boards even months after launch.

I'm sorry OP, but this is textbook FUD coupled with some official sources only to add credibility.

 

[french toast]

The memory issue is clearly very real, the only issue seems to be whether its hardware or software and when it can be fixed.
Regardless reviews will be tainted by this bug.

 

[itsmydamnation]

Oh, really? Last time i checked AIDA results for Bristol vs Steamroller, it was actually losing. Yes, pure synthetics, but nonetheless.

I think your mis interpenetrating me, im saying in terms of performance Carrizo is the best performing CON core, but in synth memory tests its the worst.
http://www.anandtech.com/bench/product/1684?vs=675

 

[imported_jjj]

Identify where it happens, then.

Well, Bristol Ridge performs fairly subpar in general, so i do not see contradictions here.

Not even sure what benchmark you are talking about. If AIDA64 memory access latency, the 6950X does 60+ms with DDR4 3200 cas16 while Kaby Lake goes bellow 50ms.
For L3, haven't heard any specifics. We do know that GloFo is half a node behind but AMD packs the cache in some 15% less area than Intel so if the L3$ is slower than Intel's, it wouldn't be surprising.
The memory support with all slots populated has been listed by mobo makers for days and hopefully it gets better. If it doesn't, that is a problem.
The BIOS is immature and we already knew that. Look at how few mobos are in retail for now. AMD was in a rush to release it and that's not ideal.

But my point was that the OP spreads speculation that the chip is defective and there is nothing to suggest that at this point.
The X for XFR fake rumor was started just like this, someone had the idea and others took it as the obvious and it became a fact.
45W quad cores fake rumor was started by infantile math- 95W for 8 cores/2 rounded down.
In a few hours this will be on hundreds of sites from China to ... Bermuda and by tomorrow it becomes a well know fact, unless someone steps up to deny it.

 

[itsmydamnation]

We already Have L3 latency from CPC review......

 

[french toast]

We already Have L3 latency from CPC review......

IMC or bios is my thoughts.

 

[PotatoWithEarsOnSide]

So if it performs like an animal despite having bugs, doesn't that imply that it'll perform like a beast once bugs are ironed out?

 

[imported_jjj]

Arent you a sensitive one. Maybe we only should report happy wccftech rumors and ignore to be objective? For the record i already prepaid 800 dollar for a 1800X and Asus crosshair IV Hero.

Objective? I am not being objective? You spread baseless rumors that the chip is defective with no evidence at all, you don't even link to some random Intel bot on a random forum making the claim.
You are at the same level as conspiracy theory folks and you dare to tell me that i am not objective?

 

[lolfail9001]

I think your mis interpenetrating me, im saying in terms of performance Carrizo is the best performing CON core, but in synth memory tests its the worst.

Fair enough.

Not even sure what benchmark you are talking about.

Both Passmark and GB4 (though i do not have archive of Naples leak, and it was first revision ES that is itself a quad MCM of first revision ESs anyways).

If AIDA64 memory access latency, the 6950X does 60+ms with DDR4 3200 cas16 while Kaby Lake goes bellow 50ms.

Entirely reasonable difference between quad and dual channel.

AMD was in a rush to release it and that's not ideal.

I'd imagine they'd be lynched if they delayed the release any further, admittedly.

In a few hours this will be on hundreds of sites from China to ... Bermuda and by tomorrow it becomes a well know fact, unless someone steps up to deny it.

It probably is already all over Chinese forums. We'll see soon, anyways.

So if it performs like an animal despite having bugs, doesn't that imply that it'll perform like a beast once bugs are ironed out?

Yes? Ultimately that's the only reason i want this to be true: for Zen to actually perform way above what i have expected.