AMD THREADRIPPER VS INTEL SKYLAKE X

[Rroff]

I know how it sounds but its all born out in the GPU section, etc. if anyone wants to go drag up old posts :s

 

[humbug]

I know how it sounds but its all born out in the GPU section, etc. if anyone wants to go drag up old posts :s

I don't think you do. not quit anyway.
Make good on your threats. Drag up your posts, see where that gets you. Take a spade.

 

[Rroff]

I'm not threatening you or set out to attack you I just would have thought after our posting history over the years you wouldn't be so quick to dismiss what I said out of hand.

 

[humbug]

I'm not threatening you or set out to attack you I just would have thought after our posting history over the years you wouldn't be so quick to dismiss what I said out of hand.

'Mutual' respect. one thing you should have learned from our posting history.

Its obvious we are not going to agree.
I've just come from an Ancient Samsung 830 to an NVMe drive with twice the IOPS and its made 0 difference to my boot time, it has made a small difference in multi application load-up times but we are talking 2X + IOPS speed difference for that effect.

Another member who has both systems to compare told you actually it makes no real difference, it may well be that in circumstance in your mind it can make a difference, yet what's in your mind be it right or wrong can't be that significant if people who have these differences in IOPS don't see it in practice.

I'm not saying you are fundamentally wrong, just that you are attributing far more significance to it than there actually is.

Lets move on shall we?

 

[FredFlint]

Oh and before I get ahead of myself - we don't know how Threadripper will perform in this respect yet but if it holds out it will be a much bigger issue on the kind of uses that platform would be put to compared to desktop use where it is less than ideal but not a game breaker.



It is amusing after all this time you still flat out deny anything I say that you don't like to hear and eventually come around to it later.

Is it SATA IOPS or m.2/pci-e IOPS? If its just SATA that would not be so bad.

 

[FredFlint]

Another member who has both systems to compare told you actually it makes no real difference, it may well be that in circumstance in your mind it can make a difference, yet what's in your mind be it right or wrong can't be that significant if people who have these differences in IOPS don't see it in practice.
I'm not saying you are fundamentally wrong, just that you are attributing far more significance to it than there actually is.

IOPS make a big difference to people developing/using IOPS intensive applications (the people that are the main target of high-end workstations, not gamers).
The main reason I have so many SSD's and hard drives is that it takes so long to run tests so I split the load over several drives.
It takes:
1 hour for a test run on mechanical drives
25 minutes on fast SATA SSDs
<5 minutes on PCI-E drives

 

[humbug]

IOPS make a big difference to people developing/using IOPS intensive applications (the people that are the main target of high-end workstations, not gamers).
The main reason I have so many SSD's and hard drives is that it takes so long to run tests so I split the load over several drives.
It takes:
1 hour for a test run on mechanical drives
25 minutes on fast SATA SSDs
<5 minutes on PCI-E drives

Fair enough, yet the platform this was tested on was a mainstream X370, the professional platform should, i would think, be very difference, for a start using the M.2 / PCEe controllers on the CPU, not the Chipset which isn't even AMD's own. The results were also not hugely different.

PS: just in case anyone is thinking "he just so happens to have a new drive, well that is convenient "

I really do and the old one really was old, its still going strong tho...

What a difference higher VAT makes, thank you Mr Osborn.

 

[Rroff]

Is it SATA IOPS or m.2/pci-e IOPS? If its just SATA that would not be so bad.

Tests show both though its more obvious on the higher end setups - one of the things that concerns me is:

As posted by another poster above on Ryzen:

Versus my slightly older drive of the same line up on Intel:



I've got the older version of the drive as well - the 850 is rated for max 97K/88K - be interesting to see what other people are getting - posts on reddit, etc. seem to bear out that those IOPS numbers are down for other people as well - if that is the case rather than the odd user having configuration issues that is concerning.

 

[Scougar]

.... nvm

 

[FredFlint]

AMD cannot afford to have a buggy launch like Ryzen, Theadripper needs to have solid memory compatibility (not overclocking speeds) and fast IO out of the gate or people will just go Intel.

 

[LoadsaMoney]

del

 

[gavinh87]

Here is my 850 evo on ryzen which ties in with your screenshot.
I have not found a single thing where this would impact me

If this was as big as you are making out there would be a lot more coverage on it.

 

[subbytna]

AMD cannot afford to have a buggy launch like Ryzen, Theadripper needs to have solid memory compatibility (not overclocking speeds) and fast IO out of the gate or people will just go Intel.

Needs quad channel memory too, not dual

 

[CAT-THE-FIFTH]

So basically you are using numbers from a consumer platform to determine numbers from an HEDT platform??? Why don't you actually wait until we see the platform in action.

Edit!!

Also if Threadripper is launching like 5 months later after the Ryzen 1800X,none of us know what tweaks have been made to actual silicon,whether they have different microcode revisions,etc or even if the chips used are a new stepping.

Considering that Ryzen tended to be better optimised out of the gate for non-gaming software,it is clear where AMD was putting more effort into before the launch,and Threadripper will be even more orientated towards the non-gaming markets.

Plus AMD wants to get more into servers,so it would make no sense if they launched an HEDT and commercial platform with IO issues.

Then you have companies like QNAP launching NAS boxes with consumer Ryzen CPUs:

http://www.anandtech.com/show/11523/qnap-launches-ts1277-nas-with-amd-ryzen-cpus
http://techreport.com/news/32010/qnap-ts-x77-nas-units-virtualize-with-ryzen

As one might expect from machines with Ryzen processors, the TS-x77 boxes are capable of more than just serving up files over a network connection. The NAS machines can also host virtual machines or application containers, provide full-content search capabilities, and can be used in conjunction with IP cameras as a surveillance DVR. Centralized management features are available through the company's QRM+ and Q'center software suites.

These start at a few 1000 USD,so I would expect they are orientated towards business and commerical use.

 

[Chuk_Chuk]

here's a question, if am4 isn't being changed/updated till what's 3 gens? 2020/21? does that mean you won't get pcie 4.0 till then?

curious as newer cards like the titan Xp and 1080ti are close to the limits of pcie x16, sli shows us they've already exceed pcie x8 by a bit, so surely by 2019/2020 (before am5 or whatever it's called) the 2280ti or whatever will be exceeding pcie 3.0 bandwidth?

Do you have any link proving that the Xp and 1080ti are close to saturating pcie X16 Gen 3?

Yep, PCIE 4 won't be coming till around 2020/21. Unless they decide to do an AM4+ chipset or something.

IOPS make a big difference to people developing/using IOPS intensive applications (the people that are the main target of high-end workstations, not gamers).
The main reason I have so many SSD's and hard drives is that it takes so long to run tests so I split the load over several drives.
It takes:
1 hour for a test run on mechanical drives
25 minutes on fast SATA SSDs
<5 minutes on PCI-E drives

Out of curiosity, what are you doing that requires co much data and IOPS?

AMD cannot afford to have a buggy launch like Ryzen, Theadripper needs to have solid memory compatibility (not overclocking speeds) and fast IO out of the gate or people will just go Intel.

You realise that Ryzen users where Beta testing for AMD right? What do you think AMD did with all the feedback they got? There will be bugs at launch but not to the extent that we saw on the mainstream platform.

 

[Rroff]

Here is my 850 evo on ryzen which ties in with your screenshot.
I have not found a single thing where this would impact me

If this was as big as you are making out there would be a lot more coverage on it.

That is your screenshot I used I've got a few other screenshots but they are from mixed SSDs so harder to compare against - there are discussions on it here and there on reddit, etc. but a lot of people haven't really picked up on it - so its hard to work out how much is down to individual systems or if its seen wider across the platform - the articles like I linked however do show some areas where IOPS, etc. performs poorly on Ryzen when tentatively bears it out. You might be largely unconcerned about it but in my experience the IOPS and time it takes to complete the benchmark in Samsung Magician usually translate into something approximately indicative of OS boot and game load times, etc. with the whole performance degradation issue with the 840 EVO drive when my benchmarks were dropping down that low I was seeing around 30-40% longer times on waiting screens, etc.

So basically you are using numbers from a consumer platform to determine numbers from an HEDT platform??? Why don't you actually wait until we see the platform in action.

Edit!!

Also if Threadripper is launching like 5 months later after the Ryzen 1800X,none of us know what tweaks have been made to actual silicon,whether they have different microcode revisions,etc or even if the chips used are a new stepping.

I did acknowledge those aspects in my posts - but I don't think it unfair to be concerned about an area which is often quite critical to the kind of environments Threadripper is designed for that shows some less than ideal sides in what we've seen so far.

 

[CAT-THE-FIFTH]

I did acknowledge those aspects in my posts - but I don't think it unfair to be concerned about an area which is often quite critical to the kind of environments Threadripper is designed for that shows some less than ideal sides in what we've seen so far.

The point is until these are released we really won't know anyway,and certainly even "if" consumer Ryzen has limitations companies like QNAP think the platform is good enough to be put into some of their pricier products.

Plus TBH if you are that worried about I/O you wouldn't be using a SATA drive,but a PCI-E based system instead!!

Edit!!

I would also suspect you would also be looking at expensive SSDs too. The enterprise drives cost silly money and would probably dwarf the cost of the CPU TBH!!

 

[humbug]

No one has picked up on it or is remotely bothered by it because there is nothing in it.

 

[FredFlint]

Do you have any link proving that the Xp and 1080ti are close to saturating pcie X16 Gen 3?

Yep, PCIE 4 won't be coming till around 2020/21. Unless they decide to do an AM4+ chipset or something.



Out of curiosity, what are you doing that requires co much data and IOPS?


You realise that Ryzen users where Beta testing for AMD right? What do you think AMD did with all the feedback they got? There will be bugs at launch but not to the extent that we saw on the mainstream platform.

Point of sale processing software for near real-time sales stats.

 

[Rroff]

That point is until these are released we really won't know anyway,and certainly even "if" consumer Ryzen has limitations companies like QNAP think the platform is good enough to be put into some of their pricier products.

Sure we won't know for sure I acknowledged that.

Plus TBH if you are that worried about I/O you wouldn't be using a SATA drive,but a PCI-E based system instead!!

Irrelevant - If I did stuff where Threadripper was suited to my needs I'd be using a very different setup and probably more like dual CPU Xeons than a 4820K.

No one has picked up on it or is remotely bothered by it because there is nothing in it.

Yeah lets go down this path again.