• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD THREADRIPPER VS INTEL SKYLAKE X

^^ To be fair I don't think people who would be using 2133MHz DDR4, etc. would be making much use of Infinity Fabric anyhow and those that do have requirements that would take advantage of it would probably be looking to max out its capabilities and not pair it with slower RAM anyhow.
 
it's not even ryzen users beta testing, there's a few things overall

the imc has a weaker design compared to intels, it's ridiculously easy to crack past 4000mhz on 4 sticks with Intel, but for some people can't even get past 2666mhz with two sticks even with the newest aegsa update.

so they really need to address this, which isn't possible with threadripper as the chips as essentially r7s stitched together.

half of the launch issues with ryzen were amd gave them no time with the cpus to make more matured bios', motherboard makers were apparently pleading with amd not to do this with threadripper at computex (according to gamernexus and Jayz when they spoke to vendors.


AMD has had 4 months to work with DDR4 RAM support. Intel has had 3 years. So you acknowledge that with time all of these problems can fixed, good.
 
AMD has had 4 months to work with DDR4 RAM support. Intel has had 3 years. So you acknowledge that with time all of these problems can fixed, good.

some of these problems can be fixed, not all.

part of it simply is the IMC design, which is what it is, at this point its safe to say the average most people can hope for is 3000-3200mhz, if they're on a good motherboard with a good imc, but this is something amd can hopefully revise with zen+ next year.

^^ To be fair I don't think people who would be using 2133MHz DDR4, etc. would be making much use of Infinity Fabric anyhow and those that do have requirements that would take advantage of it would probably be looking to max out its capabilities and not pair it with slower RAM anyhow.

but isn't the infinity fabric used in cross ccx communication? isn't that why even something like cinebench sees fairly large improvements from ddr4 speed increases on ryzen?
 
but isn't the infinity fabric used in cross ccx communication? isn't that why even something like cinebench sees fairly large improvements from ddr4 speed increases on ryzen?

I was talking more in the context of applications or pairing up hardware for environments where there are significant gains to be had from using such a system.
 
LOL back to this of inventing a massive amount of things you think I said - that I did not say at all. How is SATA express and its success or otherwise relevant to anything I've been saying? (go back read my posts there is nothing specific to SATA or PCI-e/NVMe I'm talking about the storage IO capabilities as a whole which appears, though its a mixed and far from straight forward story, to be showing in all types and areas) sure you get a speed up going to PCI-e based storage but the articles still show areas of not so great performance compared to the on paper specs for storage and compared to what Intel platforms are managing while as you said in the real world people are often IO limited on these platforms so they will want to get the best out of their hardware and so this could be a significant point for a CPU that is designed for those environments. (This is obviously an assumption that what is seen with Ryzen so far translates to Threadripper which remains to be seen).

LOL,at inventing the scenario where all HEDT users are IO limited but don't use PCI-E based storage even though the user you quoted showed he had huge gains. SATA Express did not succeed even though forum enthusiasts were excited about it,since they are more worried about running benchmarks all day to see if their new SSD is 5.4% faster than the one they bought last year.

If the market was that worried they would have put their faith in SATA Express and would have been the end of SATA3.0 - apparently not.

I also do know people who work in places which have dozens(maybe more) HEDT workstations due to the nature of the work being done - none of them are overclocked. Most HEDT desktops sold are not even configured with SSDs as standard,and you can go and check that yourself - they are either running HDDs,and if IO was so big a deal as you make out to be - then all of the older workstations would have been configured with PCI-E based storage or the latest fast SATA drive of some sort,which from my experience they are not and that is going from what I hear from loads of people.

Not even Dell,HP,etc agree with you - standard configurations are with HDDs,so that means they know plenty of customs don't seem to require one it appears.

Plus not all the actual enterprise drives are that fast - since they are made for reliability not out and out speed.

Look on Dell,HP,etc and look at the proper workstations you will notice the standard drives are SATA HDDs.

When you go to configure them(say a Dell one),half the time they don't even list the model as it is some OEM rebrand. Do people buying these system then start measurebating whether that drive is as fast as the next brand like people do on forums - nope and they have no clue whether it is faster or slower than XYZ drive because they don't care.

This is the reality - most HEDT systems don't run overclocked CPUs,they don't run fast RAM,they don't run fast storage,they don't have big PSUs,etc,and Intel will still sell more systems as they can ship simply more volume anyway. Plus for all your arguing about IO limitations,QNAP just launched a new range of systems,which if Ryzen were that bad would never be used in - I think its another case like with cameras people thinking 24MP is always better than 16MP,etc.

You are just trying to blow up a niche IO limited scenario when using SATA as the common scenario with HEDT workstations which looking at the market and the deployment of these workstations,the standard configurations,etc it shows a bigger picture that even though it is a consideration,it is not to the extent hardware enthusiasts on forums big up.

I am not going to have this discussion with you anymore TBH.
 
LOL,at inventing the scenario where all HEDT users are IO limited but don't use PCI-E based storage even though the user you quoted showed he had huge gains.

YOU invented this scenario - I categorically did NOT.

I am not going to have this discussion with you anymore TBH.

You are having a discussion with yourself - I haven't raised any of the points you are talking about!
 
@Rroff
I’m currently working on an application to compare table data stored in multiple databases to validate my sales importer. This should hammer the drives IOPS. I will test it when it’s completed on my Intel and AMD PC and find out how AMD’s IO is.
 
@Rroff
I’m currently working on an application to compare table data stored in multiple databases to validate my sales importer. This should hammer the drives IOPS. I will test it when it’s completed on my Intel and AMD PC and find out how AMD’s IO is.

Would be interesting to know - all the discussions I've seen on it have been mixed data that is impossible to compare directly and people not really understanding what the implications are.
 
Rroff said:
I'm surprised to see no one talking about the dismal storage IOPS (which can significantly slowdown OS booting and game load times, etc.) and poor random small read/write performance - going to be quite an issue if it bears out in enterprise environments where IOPS performance can be critical in systems utilising a lot of cores i.e. database management or backend services.



LOL,at inventing the scenario where all HEDT users are IO limited but don't use PCI-E based storage even though the user you quoted showed he had huge gains. SATA Express did not succeed even though forum enthusiasts were excited about it,since they are more worried about running benchmarks all day to see if their new SSD is 5.4% faster than the one they bought last year.

If the market was that worried they would have put their faith in SATA Express and would have been the end of SATA3.0 - apparently not.

I also do know people who work in places which have dozens(maybe more) HEDT workstations due to the nature of the work being done - none of them are overclocked. Most HEDT desktops sold are not even configured with SSDs as standard,and you can go and check that yourself - they are either running HDDs,and if IO was so big a deal as you make out to be - then all of the older workstations would have been configured with PCI-E based storage or the latest fast SATA drive of some sort,which from my experience they are not and that is going from what I hear from loads of people.

Not even Dell,HP,etc agree with you - standard configurations are with HDDs,so that means they know plenty of customs don't seem to require one it appears.

Plus not all the actual enterprise drives are that fast - since they are made for reliability not out and out speed.

Look on Dell,HP,etc and look at the proper workstations you will notice the standard drives are SATA HDDs.

When you go to configure them(say a Dell one),half the time they don't even list the model as it is some OEM rebrand. Do people buying these system then start measurebating whether that drive is as fast as the next brand like people do on forums - nope and they have no clue whether it is faster or slower than XYZ drive because they don't care.

This is the reality - most HEDT systems don't run overclocked CPUs,they don't run fast RAM,they don't run fast storage,they don't have big PSUs,etc,and Intel will still sell more systems as they can ship simply more volume anyway. Plus for all your arguing about IO limitations,QNAP just launched a new range of systems,which if Ryzen were that bad would never be used in - I think its another case like with cameras people thinking 24MP is always better than 16MP,etc.

You are just trying to blow up a niche IO limited scenario when using SATA as the common scenario with HEDT workstations which looking at the market and the deployment of these workstations,the standard configurations,etc it shows a bigger picture that even though it is a consideration,it is not to the extent hardware enthusiasts on forums big up.

I am not going to have this discussion with you anymore TBH.

YOU invented this scenario - I categorically did NOT.

Nope you said it was an issue in the realworld and yet when it was pointed out by people here that if you are IO limited you would use PCI-E based systems anyway,you started backtracking.

You are one going on and on about the SATA bottleneck and ,how it is a big deal and the moment it is shown that most HEDT systems deployed in the realworld have HDDs,most of the ones for sale are still configured with HDDs,that the SSDs even given as options are OEM rebrands,etc you are suddenly quiet. Then when QNAP actually released systems based on Ryzen which most likely will be use in IO intensive situations,then ignore that too:

https://www.qnap.com/en/news/2017/c...iew-qnap-unveils-worlds-first-ryzen-based-nas

The TS-x77 series will be available in 6, 8, and 12-bay models with AMD Ryzen 7 (8-cores/16-threads) and AMD Ryzen 5 (6-cores/12-threads and 4-core/8-thread) processors that support AES-NI encryption acceleration and up to 64GB DDR4 RAM. Every model in the series provides three PCIe Gen.3 slots for incredible expansion potential for supporting 10GbE/40GbE NICs, PCIe NVMe SSD, graphics cards, and USB 3.1 expansion cards. Two M.2 SATA 6 GB/s SSD slots are provided for cache acceleration or high-performance storage pools. Incorporating incredible performance, scalability and reliability, the TS-x77 series provides an exceptional business-ready storage solution for running a wide range of business tasks (including cross-platform file sharing, backup, disaster recovery, and iSCSI & virtualization tasks) without breaking a sweat.

The TS-x77 series features an excellent combination of hardware and software that delivers high expandability, high reliability, and all the tools necessary to satisfy IOPS-demanding workloads and multitasking. Download the TS-x77 datasheet for more information.

Most of the HEDT systems out there are not limited by the motherboard/CPU is the speed of the storage they already have - you are just making a mountain out of a mole hill.

Why don't you go onto the Dell website and look at the configuration tool. HDDs as standard.

Don't bother answering me - all I can see is you just inventing some new scenario.

Edit!!

Whats even weirder is Dell is going to sell Threadripper based HEDT workstations,so if IO was so poor like you said it might be,then seems like they are going to loose a lot of money then! ;)
 
Nope you said it was an issue in the realworld and yet when it was pointed out by people here that if you are IO limited you would use PCI-E based systems anyway,you started backtracking.

When FredFlint asked "Is it SATA IOPS or m.2/pci-e IOPS? If its just SATA that would not be so bad." I said that tests are showing less than ideal performance in some areas on both which is backed up in the articles you can find on the subject including the ones I linked to.

My own IO needs (HINT I'm using a 4820K rather than like a 6+ core CPU) are entirely irrelevant to Threadripper so don't confuse that in with what I'm saying. I used screenshots from my own setup as it was the easiest to compare with a known Ryzen result which loosely corresponds to what people have been posting elsewhere but with too mixed drives and systems to produce much of a definite result from.

The whole thing about SATA is your own imagination reading things in my posts that are NOT there. Yet again you are reading my posts and going off on your own little journey and can't recognise it.

EDIT: PS one of the pages from the article I linked: http://www.tweaktown.com/articles/8073/amd-ryzen-ssd-storage-performance-preview/index3.html using the 960 EVO NVMe drive.

Conclusion: Intel offers significantly better PCIe storage performance than AMD. However, it's not light years ahead as it has been in the past. Ryzen is very promising, and the difference between it and Intel isn't enough to cause us to be unwilling to recommend Ryzen as a storage platform.

As per what I was saying before its not ideal when you aren't getting the best out of your hardware but not game breaking in this case but it does have bigger implications if that holds out onto the Threadripper platform.
 
Last edited:
If you read those reviews it is only slightly worse and could be many different things in play.

Also using 1GB test files for drives which read and write at 2GB/s is hilarious.

These benchmark programs simply weren't designed for nvme SSDs.
 
If you read those reviews it is only slightly worse and could be many different things in play.

Also using 1GB test files for drives which read and write at 2GB/s is hilarious.

These benchmark programs simply weren't designed for nvme SSDs.

There are a few other benchmarks on other sites with a range of testing - the "realworld" test from that article showed a 21% difference for a typical mainstream application but some of the tests in areas which could/would be more significant to higher end setups were down around 50% - obviously again we don't know how it will look on Threadripper but at the end of the day it is building off the same platform so should be something kept in consideration.

Reading and writing at 2+GB/s is most relevant to long sequential read/writes where Ryzen does between ok and excellent - the areas that are potentially of more concern are smaller random read/writes where speeds can be vastly lower.
 
I'm surprised to see no one talking about the dismal storage IOPS (which can significantly slowdown OS booting and game load times, etc.) and poor random small read/write performance - going to be quite an issue if it bears out in enterprise environments where IOPS performance can be critical in systems utilising a lot of cores i.e. database management or backend services.

When FredFlint asked "Is it SATA IOPS or m.2/pci-e IOPS? If its just SATA that would not be so bad." I said that tests are showing less than ideal performance in some areas on both which is backed up in the articles you can find on the subject including the ones I linked to.

My own IO needs are entirely irrelevant to Threadripper so don't confuse that in with what I'm saying. I used screenshots from my own setup as it was the easiest to compare with a known Ryzen result which loosely corresponds to what people have been posting elsewhere but with too mixed drives and systems to produce much of a definite result from.

The whole thing about SATA is your own imagination reading things in my posts that are NOT there. Yet again you are reading my posts and going off on your own little journey and can't recognise it.

You are the one going on about poor SATA performance and going off on your little journey,and imagining this massive bottleneck for HEDT workstations if you used Ryzen/Threadripper and in the end its kind of weird when QNAP actually is introducing consumer Ryzen based systems for:

The TS-x77 series features an excellent combination of hardware and software that delivers high expandability, high reliability, and all the tools necessary to satisfy IOPS-demanding workloads and multitasking.

If it were as "dismal" as you say it was then seems an EPIC own goal? That is an enterprise usage scenario itself. Like I said you are thinking of probably more niche scenarios where other options also exist.

Plus you might wish it were different but again look on the Dell website - do you honestly think the SATA controller or an HDD is going to be the major limitation on those Dell systems?? Even the Dell branded SSDs offered,how fast are they compared to the competitors?? Just look at how massive a variance there can be for consumer SSDs for example. You can use your own imagination and wish three times this was not the case,but it is - nobody who is buying those systems is actually going to care what speed the SSD is.

If they actually cared they would deploy their own in-house systems or go to a boutique PC builder.

Its like people here debating how much overclock you can get on Threadripper or SKL-X which is again very good for hardware forum enthusiasts but in reality means nothing. The same with how high you can run the RAM. Great for tweakers but most of the systems will be shipped by large OEMs at stock using whatever parts which are reasonably reliable and cheap.

This is why I mentioned the whole MP thing with cameras - people just like quoting numbers,but its like with lots of things it is an interesting discussion but its really not going to make as big a difference as hardware enthusiasts on forums think it will.

Like I said you seem quite adamant to make a mountain out of a molehill here,and if the platform had such issues with IO,companies like Dell and QNAP wouldn't be building relatively expensive systems around them.
 
Is Roff still trying make issues where none exist?

some of these problems can be fixed, not all.

part of it simply is the IMC design, which is what it is, at this point its safe to say the average most people can hope for is 3000-3200mhz, if they're on a good motherboard with a good imc, but this is something amd can hopefully revise with zen+ next year.



but isn't the infinity fabric used in cross ccx communication? isn't that why even something like cinebench sees fairly large improvements from ddr4 speed increases on ryzen?


The IMC (Integrated Memory Controller) is on the CPU, not the Motherboard.

3000Mhz RAM is enough to keep up with Intel, so its not a problem, You're inventing problems to fix that don't exist in the first place, Next to Ryzen Intel have far far greater challenges to overcome, why don't you talk about some of them? like with AMD's ingenious way of dealing with the cost and yields of large CPU's, the very thing you are trying hard to be critical of, Infinity Fabric, How are Intel going to deal with it? Because if they don't AMD are going be a very serious and very constant problem to them, for as long as Intel cannot do their own Infinity Fabric AMD are going to have a huge advantage over them.
 
Last edited:
Is Roff still trying make issues where none exist?

Next to Ryzen Intel have far far greater challenges to overcome, why don't you talk about some of them? like with AMD's ingenious way of dealing with the cost and yields of large CPU's, the very thing you are trying hard to be critical of, Infinity Fabric, How are Intel going to deal with it? Because if they don't AMD are going be a very serious and very constant problem to them, for as long as Intel cannot do their own Infinity Fabric AMD are going to have a huge advantage over them.

https://www.intel.com/content/www/us/en/foundry/emib.html
 
Is Roff still trying make issues where none exist?




The IMC (Integrated Memory Controller) is on the CPU, not the Motherboard.

3000Mhz RAM is enough to keep up with Intel, so its not a problem, You're inventing problems to fix that don't exist in the first place, Next to Ryzen Intel have far far greater challenges to overcome, why don't you talk about some of them? like with AMD's ingenious way of dealing with the cost and yields of large CPU's, the very thing you are trying hard to be critical of, Infinity Fabric, How are Intel going to deal with it? Because if they don't AMD are going be a very serious and very constant problem to them, for as long as Intel cannot do their own Infinity Fabric AMD are going to have a huge advantage over them.

I'm aware the imc isn't on the morherboard, the whole quote was if you're on a good motherboard with a good imc (as in you're lucky your cpu has a good imc) you may be able to get 3200mhz, it's basically you need a silicone lottery type win on the imc on the cpu itself (as it seems to wildly vary) and you need one of the higher end motherboards that seem to be more stable with these ram speeds

I was merely saying that would be a nice thing for amd to focus on with zen+, alongside if they can somehow improve how infinity fabric is dependant on ram speed so much.

amd needed infinity fabric because they don't have the resources to use multiple dies, Intel does have these resources and thus it isn't an issue as such, Intel can and will charge higher prices, but as they are seen as (and are) the more *premium product*

although it remains to be see how long x86 has left, hell even cpus in general, at the moment Intel can just keep with what they've been doing, there's 10nm and 7nm to go, which gives them enough cpus till 2023 at least (that's with a 7nm+ and ++ design) who knows by then? maybe quantum computing will start taking off, maybe Intel will just use their own kind of infinity fabric (iirc they tested something similar a few years back but decided not to follow through with it)

maybe we'll see a new instruction set and it'll be replaced? it stands to reason that could happen, I know Intel said that I think it was tiger lake in 2019 they're changing up the core architecture, and stripping away mist of the (old) instruction sets they have on the chip as many of them are unused and waste efficiency/performance, apparently tiger lake is going to be heavily reduced on that side of things as well as moving away from the core architecture?

who knows tbh.

I'm curious how optane/3d xpoint is going to turn out, if it does what micron/Intel have said it will do (in coming years) it could be replacing ram entirely, which would be a fairly cool concept, not sure how that might effect future performance?

there's so many things that are always "coming soon" like new types of storage, or types of cpus, architectures, quantum computing (ok ok not soon but still coming) it's so hard to see what the future will bring.

hell, back in the 90s, with 16mb? ram, a 500mb hdd, a cpu that iirc was a p5? 50mhz iirc , I think I would have called you a crack addict if you told me a pc would have specs like today, and I wouldn't have to tell my parents not to use the phone was I was using the dial up 28k (ahhh **** that noise)

anyway, my point was who knows where we'll be in 10 years. (hopefully by then we're all living on the moon trump will have finally been removed from the white house, with may in chains lol)
 
Last edited:
Hey guys hows your Thread Ripper vs Skylake X testing going??

You all got chips and boards I take it???
 
Back
Top Bottom