Under 'normal' circumstances what you're saying would be right. Most companies like to conserve pennies. What 8Pack is saying though is that for some the critical factor is time and then an overclocked system can make sense. I mean if a system has a 20% overclock for example, that could shave a full month off of a 5 month task. You get the idea.
-
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
- Thread starter faceman123
- Start date
Under 'normal' circumstances what you're saying would be right. Most companies like to conserve pennies. What 8Pack is saying though is that for some the critical factor is time and then an overclocked system can make sense. I mean if a system has a 20% overclock for example, that could shave a full month off of a 5 month task. You get the idea.
Caporegime
I suspect there's more to it than just the "time is key", ie, more than just the performance of the CPU. At the moment I would guess it's not easy for a company to take a gamble on the unproven Threadripper CPU's. They need proven CPU's here and now, with part availability and low chance of failure and high MTBF. THat's what I'm guessing anyway.
Besides, people and even companies stick to what they know - it's sometimes hard to break a habit.
It's clear from a few threads that if money is no object, for whatever reasons, still the recommendation is Intel.
Except apparently all these companies used overclocked CPUs,which made me LOL,especially looking at some the large clusters I have seen which are MASSIVE.
Basically one floor used to handle cooling,etc and the like. If anything I would argue looking at what some large companies have said they are looking to drop power,as the cooling alone was adding massively to the power consumption(!) since having refrigeration plants to keep the servers cool,and ways to expel all the heat added massively to cost and complexity which itself also impacts MTBF.
Then you have large companies like Baidu,who seem to contradict AMD won't be used by any companies:
https://www.youtube.com/watch?v=lIXTm63e974
Baidu has a market cap of nearly $80 billlion - they apparently feel Eypc is good enough for them to use.
Micrsoft is using Ryzen as part of its Azure platform and they have even a bigger market cap.
Its almost like Intel needs to try and bury it. Remember with even a billion dollars a year in bribes,even Dell decided to use the Athlon 64.
Look at all the major server companies they are releasing AMD power servers,which means they think its worth their while.
The problem is that overclocking also increase MTBF,so this is why lots of these large installations don't use overclocked CPUs,since you would need to validate it long enough to make sure it won't have statistically bigger chance of throwing errors after such a long time period.
Its no point if towards the end the system BSODs 4 months into the project so you need to start again. If would be cheaper to just add more nodes for the project.
This must be for task which are short in nature,or something like that,which are very peaky in nature and not that critical as they have sufficient redundancy already.
Remember this Intel has introduced high clockspeed SKUs even back in the Core i7 920 days running two cores at 4.4GHZ.
If Intel thinks there is a massive commercial market for high clockspeed CPUs they would release SKUs.
Looking at the Xeon Gold range,etc I would argue looking at the clockspeeds,that the market is very niche.
Last edited:
Caporegime
I suspect there's more to it than just the "time is key", ie, more than just the performance of the CPU. At the moment I would guess it's not easy for a company to take a gamble on the unproven Threadripper CPU's. They need proven CPU's here and now, with part availability and low chance of failure and high MTBF. THat's what I'm guessing anyway
Besides, people and even companies stick to what they know - it's sometimes hard to break a habit.
It's clear from a few threads that if money is no object, for whatever reasons, still the recommendation is Intel.
That's Intel marketing right there, are you just part of it or have you fallen for their obvious twaddle, so much so that you are repeating it?
If Intel had some of AMD's glue they would be a lot better, more cost effective, more power efficient.... they might not even be so much slower.
As for failure chance, its Intel's CPU that you would be taking a chance on, Intel's Core i9's are the ones frying VRMs and burning CPU pins, they should come with a fire extinguisher.
Look at this.....
That's ^^^^ down to insane amounts of power consumption.
I wouldn't want that in my house.
Last edited:
Obviously there are some case scenario's where the heightened risk of failure is outweighed by the need for quick results (and 8Pack stands by his systems). Also depending on the task/software in question a BSOD may not mean restarting in full but resuming from where the system left the task. As for whether it's one big task that lasts ages or a series of small tasks that are quick the point remains. A percentage reduction in time.
Caporegime
Intel's core i9's don't even have ECC compatible Memory, so for people serious about servers the core i9 is almost out, talking about throwing errors Threadripper wich does have ECC compatible RAM is far more reliable.
And 44 PCIe lanes, content creators are going to look at that and think "Thats an IO bottleneck i don't want, Threadripper has 64"
You put 3 GPU's in that X299 system and one of them is bottlenecked, thats before you get to NVMe drives.
Realistically you couldn't run more than 2 GPU's with the 7900X. that's not good enough for anyone serious about it.
And 44 PCIe lanes, content creators are going to look at that and think "Thats an IO bottleneck i don't want, Threadripper has 64"
You put 3 GPU's in that X299 system and one of them is bottlenecked, thats before you get to NVMe drives.
Realistically you couldn't run more than 2 GPU's with the 7900X. that's not good enough for anyone serious about it.
Last edited:
if you dont like it then dont buy it simple but why rubbish other peoples opinions as its there right and money to do as they please,seems to me you say anyone that states anything about intel or x299 as they are literally buying into intels hype or fanboyism but from my perspective it seems your doing the same for amd both parties cpu,s have good and bad points and its down to an individuals or companies own preference which way they go just because it might not correspond with your choice doesnt make it the wrong choice for them
Caporegime
OK,I found a list of the latest Xeon Gold and Xeon Platinum ranges:
https://ark.intel.com/products/series/125191/Intel-Xeon-Scalable-Processors
These are Intels highest profit margin parts I suspect.
The highest clockspeed part runs at 3.5GHZ~4.2GHZ and has 8 cores.
The most expensive part is the 28 core Xeon Platinum 8180M which runs at 2.5GHZ~3.8GHZ and costs $13000.
If you look how far Eypc goes to,the equivalent Intel CPU to the top Eypc one is the 24 core Xeon Platinum 8160 running at 2.1GHZ~3.7GHZ. The AMD Epyc 7601 32 core CPU runs at between 2.2GH~3.2GHZ by comparison.
I like said they must be shorter more peakier tasks and looking at the Intel Xeon range,none of them break past 4.2GHZ,so even for Intel there seems no need for a 5GHZ SKU,which indicates its a niche market in the scheme of things.
Thats the thing anything which is measured in weeks,months,etc is no point trying to use overclocked CPUs,since downtime is still downtime in man hours,etc,and more importantly as I said look at large companies like Amazon,Google,etc with their data centres. The cooling itself was a major consideration for reductions in power consumption alone as they literally needed refrigeration plants for the farms.The fact is how long are going to test the overclock - days,weeks or months for such situations?
Honestly looking at that huge cluster at a university and having a whole floor dedicated to cooling made me go wow - also AMD being not being bought by companies is more down to the fact AMD suffered in performance/watt in the data centre,and with Bulldozer are you surprised?? Ryzen by its very nature is an entirely different beast,its like discounting the Core2 since the Pentium 4 was crap.
You really need to look at who buys most of these systems.
Also,if Intel is not offering overclocked servers,and massive multi-billlion dollar companies like Dell,HP,etc don't seem to offer them either,surely that indicates a very niche situation.
I mean not even the proper Dell workstations are overclocked - Alienware are their prosumer range.
Also apparently they don't support ECC memory so they can't be for situations which required maximum reliability if that is the case.
Wait,I thought they supported ECC memory??
Emm,LOL??
https://ark.intel.com/products/series/125191/Intel-Xeon-Scalable-Processors
These are Intels highest profit margin parts I suspect.
The highest clockspeed part runs at 3.5GHZ~4.2GHZ and has 8 cores.
The most expensive part is the 28 core Xeon Platinum 8180M which runs at 2.5GHZ~3.8GHZ and costs $13000.
If you look how far Eypc goes to,the equivalent Intel CPU to the top Eypc one is the 24 core Xeon Platinum 8160 running at 2.1GHZ~3.7GHZ. The AMD Epyc 7601 32 core CPU runs at between 2.2GH~3.2GHZ by comparison.
I like said they must be shorter more peakier tasks and looking at the Intel Xeon range,none of them break past 4.2GHZ,so even for Intel there seems no need for a 5GHZ SKU,which indicates its a niche market in the scheme of things.
Thats the thing anything which is measured in weeks,months,etc is no point trying to use overclocked CPUs,since downtime is still downtime in man hours,etc,and more importantly as I said look at large companies like Amazon,Google,etc with their data centres. The cooling itself was a major consideration for reductions in power consumption alone as they literally needed refrigeration plants for the farms.The fact is how long are going to test the overclock - days,weeks or months for such situations?
Honestly looking at that huge cluster at a university and having a whole floor dedicated to cooling made me go wow - also AMD being not being bought by companies is more down to the fact AMD suffered in performance/watt in the data centre,and with Bulldozer are you surprised?? Ryzen by its very nature is an entirely different beast,its like discounting the Core2 since the Pentium 4 was crap.
You really need to look at who buys most of these systems.
Also,if Intel is not offering overclocked servers,and massive multi-billlion dollar companies like Dell,HP,etc don't seem to offer them either,surely that indicates a very niche situation.
I mean not even the proper Dell workstations are overclocked - Alienware are their prosumer range.
Also apparently they don't support ECC memory so they can't be for situations which required maximum reliability if that is the case.
Wait,I thought they supported ECC memory??
Emm,LOL??
this..
Caporegime
Associate
This is what happens when Intel feels the heat. They rush the garbage out the door. Not going to work this time. It's not like when P4 was going up against Athlon/FX and the public was less edumacated and people were sill buying pentium 4's aka "The cr@pbox" like hotcakes.
I fell for this and bought a Pentium 4 @ 3.2Ghz in 2004. And it was PGA478. I was not educated and bought last year's kit in 2004 (PC came with an ATI 9800 Pro. X800 was already out. I didn't know.) I'm so glad I didn't start going on forums until I got a C2D in 2007 or I would have been so upset. Grr.
And my ignorance and idiocy didn't stop there! I built my uncle a Pentium D 805 based machine in 2005. Yes, I have sinned.
Caporegime
Wait,why has Intel lopped off ECC support for the Core i9 - the chip itself supports ECC memory.
My IB based Xeon E3 1230 V2 supports ECC RAM - why does Intel feel the need to segregate every feature like that??
Caporegime
Its cannibalising their mega money Xeon line, so its disabled.
Thats what you get when Intel think they own the roost, which they do, that needs to change.
Soldato
$$$$$
Reasons
Caporegime
See I don't understand the argument in this thread. No company would want Epyc because even if its cheaper, companies want moar performance and reliability and money is not an issue,but the Core i9 does not have ECC RAM.
If money is not a problem why don't they just get a Xeon Gold or a Xeon Platinum CPU instead? Some of these have 4.2GHZ Turbo too.
Soldato
- Joined
- 19 Oct 2008
- Posts
- 6,004
That's Intel marketing right there, are you just part of it or have you fallen for their obvious twaddle, so much so that you are repeating it?
If Intel had some of AMD's glue they would be a lot better, more cost effective, more power efficient.... they might not even be so much slower.
As for failure chance, its Intel's CPU that you would be taking a chance on, Intel's Core i9's are the ones frying VRMs and burning CPU pins, they should come with a fire extinguisher.
Look at this.....
That's ^^^^ down to insane amounts of power consumption.
I wouldn't want that in my house.
Not fallen for anything, I was trying to think how an organisation may think rather than us as enthusiasts and individuals. Heck, I've even picked up AMD stock recently hoping for a decent return, although a bit of a gamble that so far hasn't paid off as it's tanked compared to Intel (I sold quite a bit of Intel and moved it over to AMD - ha !).
Yep, i'm not disputing most of what you're saying here. I agree. Just that there are 'use case scenarios' where time is of most importance. As to how niche that is i couldn't say in all honesty.
Caporegime
We all know Intel has rushed a response out to Ryzen - these would probably have been clocked lower and been released a bit later if Ryzen was not released this year or was poorer in non-gaming situations. If anything I would argue that the AMD SMT scaling being far better and more robust(outside gaming) than originally thought is probably what threw Intel a bit,as everybody before Ryzen launched expected it to have WORST scaling than what Intel had.
Rumours also hinted at much lower clockspeeds too.
Look at Coffeelake - its clear Intel has pushed it forward,when they have literally rushed out the Z370 which is going to be replaced by the proper chipset,the Z390 in a few months.
Give it a few more months,then I expect things will be better.
Intel basically does not want to be outperformed in performance,so it will rather rush something out just to claim the crown.
Most people buying their CPUs won't be buying the top ones and probably won't overclock,but will do so because Intel is the "fastest".
AMD itself probably has released Threadripper more as PR exercise too,in some ways to show they have a very fast desktop chip.
I expect consumer Ryzen and Eypc will be more important to them as a company.
Sure,I won't disagree with you but I just think for most cases outside gaming,AMD knows very well they are fighting stock clocked Intel CPUs,which are not running at anything close to say 5GHZ.
The sad thing for them is since gaming performance whilst being solid,is not quite upto some of the Intel CPUs,and it can't really overclock well,its kind of not helped with its appreciation from enthusiasts or enthusiast focused companies like OcUK,and these are obviously quite a vocal set of PC people!
I mean look at the Hardware.fr review of the 1950X:
http://www.hardware.fr/articles/967-8/compression-7-zip-winrar.html
Hardware.fr has never been very pro AMD and they are quite well known,but look at the application performance.
Last edited:
Caporegime
We all know Intel has rushed a response out to Ryzen - these would probably have been clocked lower and been released a bit later if Ryzen was not released this year or was poorer in non-gaming situations. If anything I would argue that the AMD SMT scaling being far better and more robust(outside gaming) than originally thought is probably what threw Intel a bit,as everybody before Ryzen launched expected it to have WORST scaling than what Intel had.
Rumours also hinted at much lower clockspeeds too.
Look at Coffeelake - its clear Intel has pushed it forward,when they have literally rushed out the Z370 which is going to be replaced by the proper chipset,the Z390 in a few months.
Give it a few more months,then I expect things will be better.
Intel basically does not want to be outperformed in performance,so it will rather rush something out just to claim the crown.
Most people buying their CPUs won't be buying the top ones and probably won't overclock,but will do so because Intel is the "fastest".
AMD itself probably has released Threadripper more as PR exercise too,in some ways to show they have a very fast desktop chip.
I expect consumer Ryzen and Eypc will be more important to them as a company.
Sure,I won't disagree with you but I just think for most cases outside gaming,AMD knows very well they are fighting stock clocked Intel CPUs,which are not running at anything close to say 5GHZ.
The sad thing for them is since gaming performance whilst being solid,is not quite upto some of the Intel CPUs,and it can't really overclock well,its kind of not helped with its appreciation from enthusiasts or enthusiast focused companies like OcUK,and these are obviously quite a vocal set of PC people!
I mean look at the Hardware.fr review of the 1950X:
http://www.hardware.fr/articles/967-8/compression-7-zip-winrar.html
Hardware.fr has never been very pro AMD and they are quite well known,but look at the application performance.
Well, what they are doing is putting all the Intel stuff at the top irregardless of the performance, if you scroll down to where all the AMD stuff is you can see with a lot stuff the 1950X is faster, other than WinRar and a couple of others. like using an old obsolete compiler because its an Intel back-end.
Also, they testing 7Zip 'Compression only' decompression is what you do mostly with Zip packers and in that Threadripper destroys the i9, so when you test 7Zip and test 'Compression only' what are doing is bias.
"Hardware.fr has never been very pro AMD"
That is very clear. ^^^^^
Last edited:
Caporegime
Look at the actual numbers and the final application scores - the 1950x is 14% faster than a 7900X. If a website that is not that pro-AMD is showing that it shows you how fast the 1950x is.