Personal computing discussed

Moderators: renee, mac_h8r1, Nelliesboo

  • 1
  • 2
  • 3
  • 4
  • 5
  • 7
 
maxxcool
Gerbil Elite
Posts: 855
Joined: Thu Sep 12, 2002 8:40 am
Location: %^&*%$$
Contact:

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 3:57 pm

Glorious wrote:
maxxcool wrote:
As in post #1,  part#1.  TDP per wiki is a heat rating.


No.

Wikipedia wrote:
The thermal design power (TDP), sometimes called thermal design point, is the maximum amount of heat generated by a computer chip or component (often the CPU or GPU) that the cooling system in a computer is designed to dissipate in typical operation


A "heat rating" would imply it's how much heat the device could sustain. Not that "heat rating" is any sort of commonly understood term, though.

maxxcool wrote:
After 2006 it gets murky.  But AMD stuck to it.


"Murky" as in you didn't read the footnote to that one line in the wikipedia article, nor do you understand the overall context?

maxxcool wrote:
Regardless point#1 is I ran,  as I have noted several times in the past, 5 Thuban systems that would throttle with the OEM heatsink immediately.  And would throttle with decent 'for the day' Cooler Master cooling towers under load here at the office.


Yes. Regardless indeed.

Not only is this not any sort of anything, it's also about a processor released by AMD over 6 years ago.

maxxcool wrote:
Is it scientific .. nope.  but it is reality I dealt with until I replace them all with DC silicon.  AMD throttled in a 72f office all day long with then 'better end' cooling solutions let alone the 'rated' cooler on hand in the box.


You're not even pretending to have measured or even considered the power consumption, are you?

maxxcool wrote:
Large scale hashing.  All six cores would be performing hashing on drive array's for days at a time on each device.  


....what? And you're telling me you were actually CPU-bound? Or that this scenario makes some sort of sense? What on earth where you hashing, why/what/huh?

Try again?


/Shrug/  I never said I had anything that was 'recent' AMD silicon, You never asked :)  All i offered was experience in usage and opinion.  Nothing more.  And even advised that the future for AMD may be different as we will have to wait and see.

edit : fixed 'any' to be 'may' .. typing too many things at once.
Cybert said: Capitlization and periods are hard for you, aren't they? I've given over $100 to techforums. I should have you banned for my money.
 
whm1974
Emperor Gerbilius I
Topic Author
Posts: 6361
Joined: Fri Dec 05, 2014 5:29 am

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 4:12 pm

Vhalidictes wrote:
Semi-Relevant Story Time: 

So, a year ago I doubled my Skyrim framerate with the same video card (a AMD 6870) and memory (2*Mushkin 8GB 1333MHz modules, I believe). I "upgraded" from a AMD Athlon X4 860K running at 4.2GHz to a Intel Celeron G550 (dual-core 2.6GHz).

A processor with almost no cache, running at almost half the speed, and without either half the cores or HT (depending on how you look at modules), doubled my performance in single-threaded gaming. I still can't believe that and I saw it myself.

So IPC is everything then? That doesn't surprise me at all.
 
maxxcool
Gerbil Elite
Posts: 855
Joined: Thu Sep 12, 2002 8:40 am
Location: %^&*%$$
Contact:

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 4:15 pm

whm1974 wrote:
Vhalidictes wrote:
Semi-Relevant Story Time: 

So, a year ago I doubled my Skyrim framerate with the same video card (a AMD 6870) and memory (2*Mushkin 8GB 1333MHz modules, I believe). I "upgraded" from a AMD Athlon X4 860K running at 4.2GHz to a Intel Celeron G550 (dual-core 2.6GHz).

A processor with almost no cache, running at almost half the speed, and without either half the cores or HT (depending on how you look at modules), doubled my performance in single-threaded gaming. I still can't believe that and I saw it myself.

So IPC is everything then? That doesn't surprise me at all.

Yeah but to play devils advocate the lack of large cache on the Intel side can lead to cache misses a severe short term IPC loss...
Cybert said: Capitlization and periods are hard for you, aren't they? I've given over $100 to techforums. I should have you banned for my money.
 
Glorious
Gerbilus Supremus
Posts: 12343
Joined: Tue Aug 27, 2002 6:35 pm

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 4:43 pm

maxxcool wrote:
/Shrug/ I never said I had anything that was 'recent' AMD silicon, You never asked :) All i offered was experience in usage and opinion. Nothing more. And even advised that the future for AMD may be different as we will have to wait and see.


Absolute nonsense!

You literally said you had a "long history" of AMD CPUS in your "lab", here, let me help you:

maxxcool wrote:
Given my long history with AMD cpus in my lab


Which, as you belatedly admitted, was nothing more than your completely anecdotal experience with a singular CPU model from AMD that was released six years.

You implied rigor. You implied a pattern. You outright said the "ratings" were optimistic when, in fact, you never verified the rating at all and it's abundantly clear you not only fail to understand what TDP actually means but several underlying physical concepts as well.

Heat and temperature, while obviously related, are actually distinct concepts, and in this context that distinction is extremely important. But yet, in this entire discussion, you have continually treated them as the same despite my repeated attempts to communicate that this is not the case.

I'm sorry if you being wrong hurts your feelings, but your vaunted experience still does not uphold your statements, because you don't actually understand what that experience is telling you.

Just as you don't get what I'm trying to tell you: The fact that a CPU overheats is simply not evidence that its published TDP is in err. The temperature of the die, while related to the heat output of the processor, is not actually the same thing! You are falsely conflating the two, which was the basis of my rejection of your statement from the very outset.
 
whm1974
Emperor Gerbilius I
Topic Author
Posts: 6361
Joined: Fri Dec 05, 2014 5:29 am

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 4:49 pm

maxxcool wrote:
whm1974 wrote:
Vhalidictes wrote:
Semi-Relevant Story Time: 

So, a year ago I doubled my Skyrim framerate with the same video card (a AMD 6870) and memory (2*Mushkin 8GB 1333MHz modules, I believe). I "upgraded" from a AMD Athlon X4 860K running at 4.2GHz to a Intel Celeron G550 (dual-core 2.6GHz).

A processor with almost no cache, running at almost half the speed, and without either half the cores or HT (depending on how you look at modules), doubled my performance in single-threaded gaming. I still can't believe that and I saw it myself.

So IPC is everything then? That doesn't surprise me at all.

Yeah but to play devils advocate the lack of large cache on the Intel side can lead to cache misses a severe short term IPC loss...

Yeah CPU design sure does involve a number of tradeoffs.
 
just brew it!
Administrator
Posts: 54500
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 4:59 pm

whm1974 wrote:
Yeah CPU design sure does involve a number of tradeoffs.

If it was easy everyone would be doing it. :wink:
Nostalgia isn't what it used to be.
 
maxxcool
Gerbil Elite
Posts: 855
Joined: Thu Sep 12, 2002 8:40 am
Location: %^&*%$$
Contact:

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 5:13 pm

Glorious wrote:
maxxcool wrote:
/Shrug/  I never said I had anything that was 'recent' AMD silicon, You never asked :)  All i offered was experience in usage and opinion.  Nothing more.  And even advised that the future for AMD may be different as we will have to wait and see.


Absolute nonsense!

You literally said you had a "long history" of AMD CPUS in your "lab", here, let me help you:

maxxcool wrote:
Given my long history with AMD cpus in my lab


Which, as you belatedly admitted, was nothing more than your completely anecdotal experience with a singular CPU model from AMD that was released six years.

You implied rigor. You implied a pattern. You outright said the "ratings" were optimistic when, in fact, you never verified the rating at all and it's abundantly clear you not only fail to understand what TDP actually means but several underlying physical concepts as well.

Heat and temperature, while obviously related, are actually distinct concepts, and in this context that distinction is extremely important. But yet, in this entire discussion, you have continually treated them as the same despite my repeated attempts to communicate that this is not the case.

I'm sorry if you being wrong hurts your feelings, but your vaunted experience still does not uphold your statements, because you don't actually understand what that experience is telling you.

Just as you don't get what I'm trying to tell you: The fact that a CPU overheats is simply not evidence that its published TDP is in err. The temperature of the die, while related to the heat output of the processor, is not actually the same thing! You are falsely conflating the two, which was the basis of my rejection of your statement from the very outset.

:) Well my friend,  don't worry about my feelings.,  I have none :)  

As to Heat VS temperature while you may appreciate the minutia (as well as a few others on the TR 'boards) most all won't and I admit I am one of them.   I do not sit with my amp-meter measuring at the wall or at the socket,   I look at the specs on the box, buy it. plug it in and expect it to work.   If you think that's atrocious so be it..  It do not work and I had to reinvest in additional cooling which while irksome was not world ending.  On the next iteration of boxes I tried with a DC 4670 as a test, and satisfied with peaking 70c under full sustained loads converted the remaining 4 devices to DC silicon as well. 

Glorious wrote:
Glorious:3. If it is throttling with their stock cooler in a reasonably-sized/cooled case at room temperatures, than it is not only falsely-advertised but plainly defective. It would also be immediately evident in the reviews, which all discerning people should be waiting for anyway.



As you put in a previous post.  It did not work with the stock heatsink, at room temp...  But it was not defective so the only other sane conclusion as you noted was the manufacturer was in error of its rating, based on that experience I posted caution to WHM to not spend hard currency until we see real load based testing from TR and other reputable sources.

and as always.. it is my opinion and advised a wait and see strategy.  
Cybert said: Capitlization and periods are hard for you, aren't they? I've given over $100 to techforums. I should have you banned for my money.
 
Vhalidictes
Gerbil Jedi
Posts: 1835
Joined: Fri Jan 07, 2005 2:32 pm
Location: Paragon City, RI

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 5:21 pm

maxxcool wrote:
As you put in a previous post.  It did not work with the stock heatsink, at room temp...  But it was not defective so the only other sane conclusion as you noted was the manufacturer was in error of its rating, based on that experience I posted caution to WHM to not spend hard currency until we see real load based testing from TR and other reputable sources.


and as always.. it is my opinion and advised a wait and see strategy.  

I'm a bit lost here.

You're saying "it did not work". I'll bet it didn't.

PCs aren't designed to be run 100% for days at a time on the stock heatsink. That's not in the fevered imaginings of the people who develop the TDP ratings... or the SOHO hardware that you were clearly using to run the program.

If a rack mount server sitting in a data center overheated, sure I'd blame the manufacturer. If my home PC overheats/throttles when I run Prime95 all afternoon on every core using a mid-grade tower air cooler, that's on me.
 
synthtel2
Gerbil Elite
Posts: 956
Joined: Mon Nov 16, 2015 10:30 am

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 5:32 pm

Oh great, the flamewar squad arrived. :roll:

ptsant wrote:
I sort of expected AMD to restrict ECC support to the Naples or however the "server" edition is called. Are you guessing that ECC support will be included or do you have information to that effect?

Just a guess based on AMD's history of ECC support. I'll be very disappointed in them if it isn't available, but I put it on the list of requirements because it wouldn't be inconceivable for them to screw it up.

just brew it! wrote:
Unless you really need (or want...) *both* small form factor *and* 8 cores, why wouldn't you go for a lower TDP quad core CPU, or step up to a micro-ATX form factor?
Kretschmer wrote:
Why on earth would you want to use a 16-thread low-clock CPU in a SFF gaming system?

It's just as much a workstation as a gaming system, ease of transport is a big plus, and with my cooling setup the downsides are hardly going to exist. I can't speak for whm1974's situation, though.

Topinio wrote:
[I miss auxy.]

+1.

Vhalidictes wrote:
So, a year ago I doubled my Skyrim framerate with the same video card (a AMD 6870) and memory (2*Mushkin 8GB 1333MHz modules, I believe). I "upgraded" from a AMD Athlon X4 860K running at 4.2GHz to a Intel Celeron G550 (dual-core 2.6GHz).

A processor with almost no cache, running at almost half the speed, and without either half the cores or HT (depending on how you look at modules), doubled my performance in single-threaded gaming. I still can't believe that and I saw it myself.

Interestingly enough, when my brother upgraded from a G550 to an FX-6300 at four-point-something, it was a solid upgrade for Skyrim (I forget exactly how much, but definitely noticable). That was with 4GB of DDR3-1333 and either an HD 7770 or 260X (I forget which one he was using at the time). The 7770 has since made its way over to a friend's 6700K rig, and that one's Skyrim framerates are in an entirely different ballpark (and I'm sure they got another huge boost since that guy upgraded to an RX 480). One thing is for sure, Skyrim isn't the most consistent about this stuff, especially when mods get thrown into the equation.
 
Glorious
Gerbilus Supremus
Posts: 12343
Joined: Tue Aug 27, 2002 6:35 pm

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 5:41 pm

maxxcool wrote:
On the next iteration of boxes I tried with a DC 4670 as a test, and satisfied with peaking 70c under full sustained loads converted the remaining 4 devices to DC silicon as well.


I assume DC silicon is some sort of thermal grease then?

If changing that means it went from constantly throttling to OK operation.... like, I don't know, this is getting even crazier. There shouldn't be such a stark difference between properly applied paste.

maxxcool wrote:
It did not work with the stock heatsink, at room temp... But it was not defective so the only other sane conclusion as you noted was the manufacturer was in error of its rating,


Uhh??

My statement was that if it was throttling at room temperature in a reasonable sized/cooled case with the stock cooler, yes, it is defective. They supplied the cooler as part of the retail package, so if that combination does not perform appropriately in a reasonable enclosure, than it is unacceptable.

I don't know if this was the situation with those six cored thubans @ 125w, but if it was, my statement applies to them and I have no problem saying they were defective.

I had a MSI factory overclocked Geforce 460 that had a problem like this: Defective. Some high performance/gaming Laptops have this problem. I've denounced them before.

This is entirely independent of the accuracy of the TDP: if the stocker cooler can't keep up, well, it can't keep up.

I'm not taking your word for it, but I've seen enough dinky stock coolers to fully believe it's entirely possible that the one AMD shipped with your Thubans wasn't suitable.
 
gmskking
Gerbil
Posts: 48
Joined: Tue Oct 11, 2011 4:13 pm
Location: Gerbil Kingdom

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 5:42 pm

Now the question will be...Is it worth the extra $150 dollars for the 3.2GHz version of the 8-Core?
Main: Core i7 4790K @ 4.7GHz | Samsung 850 Evo | 16GB DDR3 | Antec P110 Luce | Windows 10
Other: Core i5 2500K | Toshiba TR200 240GB | 8GB DDR3 | Windows 10
Gaming: Xbox One X
 
DancinJack
Maximum Gerbil
Posts: 4494
Joined: Sat Nov 25, 2006 3:21 pm
Location: Kansas

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 5:47 pm

gmskking wrote:
Now the question will be...Is it worth the extra $150 dollars for the 3.2GHz version of the 8-Core?


No.
i7 6700K - Z170 - 16GiB DDR4 - GTX 1080 - 512GB SSD - 256GB SSD - 500GB SSD - 3TB HDD- 27" IPS G-sync - Win10 Pro x64 - Ubuntu/Mint x64 :: 2015 13" rMBP Sierra :: Canon EOS 80D/Sony RX100
 
just brew it!
Administrator
Posts: 54500
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 5:48 pm

synthtel2 wrote:
Vhalidictes wrote:
So, a year ago I doubled my Skyrim framerate with the same video card (a AMD 6870) and memory (2*Mushkin 8GB 1333MHz modules, I believe). I "upgraded" from a AMD Athlon X4 860K running at 4.2GHz to a Intel Celeron G550 (dual-core 2.6GHz).

A processor with almost no cache, running at almost half the speed, and without either half the cores or HT (depending on how you look at modules), doubled my performance in single-threaded gaming. I still can't believe that and I saw it myself.

Interestingly enough, when my brother upgraded from a G550 to an FX-6300 at four-point-something, it was a solid upgrade for Skyrim (I forget exactly how much, but definitely noticable). That was with 4GB of DDR3-1333 and either an HD 7770 or 260X (I forget which one he was using at the time). The 7770 has since made its way over to a friend's 6700K rig, and that one's Skyrim framerates are in an entirely different ballpark (and I'm sure they got another huge boost since that guy upgraded to an RX 480). One thing is for sure, Skyrim isn't the most consistent about this stuff, especially when mods get thrown into the equation.

Kind of makes me wonder if there was something else wrong with the 860K system. Something with the motherboard that was holding back CPU or GPU performance, perhaps...
Nostalgia isn't what it used to be.
 
gmskking
Gerbil
Posts: 48
Joined: Tue Oct 11, 2011 4:13 pm
Location: Gerbil Kingdom

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 5:51 pm

DancinJack wrote:
gmskking wrote:
Now the question will be...Is it worth the extra $150 dollars for the 3.2GHz version of the 8-Core?


No.

Kind of what I was thinking as well.
Main: Core i7 4790K @ 4.7GHz | Samsung 850 Evo | 16GB DDR3 | Antec P110 Luce | Windows 10
Other: Core i5 2500K | Toshiba TR200 240GB | 8GB DDR3 | Windows 10
Gaming: Xbox One X
 
DancinJack
Maximum Gerbil
Posts: 4494
Joined: Sat Nov 25, 2006 3:21 pm
Location: Kansas

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 5:53 pm

just brew it! wrote:
Kind of makes me wonder if there was something else wrong with the 860K system. Something with the motherboard that was holding back CPU or GPU performance, perhaps...


Me too. A G550 should in no way beat a 860K by that much (if at all). No way.
i7 6700K - Z170 - 16GiB DDR4 - GTX 1080 - 512GB SSD - 256GB SSD - 500GB SSD - 3TB HDD- 27" IPS G-sync - Win10 Pro x64 - Ubuntu/Mint x64 :: 2015 13" rMBP Sierra :: Canon EOS 80D/Sony RX100
 
Vhalidictes
Gerbil Jedi
Posts: 1835
Joined: Fri Jan 07, 2005 2:32 pm
Location: Paragon City, RI

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 6:45 pm

DancinJack wrote:
just brew it! wrote:
Kind of makes me wonder if there was something else wrong with the 860K system. Something with the motherboard that was holding back CPU or GPU performance, perhaps...


Me too.  A G550 should in no way beat a 860K by that much (if at all).  No way.

I was just as surprised as everyone else. Originally, I decided to undervolt the 860K for power/heat reasons (and to see if I could). It worked like a charm, but Windows performance was completely terrible. After noticing that the CPU was running at 400MHz in Windows System Properties, I went through all the motherboard BIOS options and noticed that the system downclocked itself....aggressively when the "Cool N Quiet" option was enabled.

I ended up manually locking the processor to 4200MHz, turning off all of the various Sleep options, and the "slow GUI performance" problem went away. I think that's the last thing that I changed before switching systems due to the motherboard dying.

My original plan was to replace the board, but after seeing how well the G550 was working for older games, I just kept it for a year before finally getting a Sandy Bridge-E system from a friend.
The only performance problem I can see was that slow-*** 1333MHz CAS 11 memory (salvaged from a i7-920 build). 

But that doesn't seem like that would make that much difference, and as I said, Windows GUI performance was superb. The system didn't seem slow unless I was in a Gamebyro / Bethsoft game.

Using the same 6870, the average framerate went from the mid 30's to the high 50's. Change was very stark. I tend to run a lot of mods (around 150, depending on the game), but the motherboard/CPU combo was the only difference. Same drivers, same OS (Windows 7 Pro x64).
 
Chrispy_
Maximum Gerbil
Posts: 4670
Joined: Fri Apr 09, 2004 3:49 pm
Location: Europe, most frequently London.

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 8:42 pm

whm1974 wrote:
So IPC is everything then?

Really, absolutely, yes.

This is why Zen's IPC is so critically important; Bulldozer was such a complete turd that actually had significantly lower IPC than the older Phenoms. Clockspeed sort of half hid the IPC defecit, and over the last four years tweaks to make Piledriver/Steamroller/Excavator evolutions have probably pulled it back up to finally surpass Phenom IPC levels. Meanwhile, AMD has been haemorrhaging money trying to sell their hideous mistake for anything they could until Zen was baked. Honestly, Intel made this mistake with the Pentium 4 and it took them half a decade to recover. That's how long it takes to react and change direction for a CPU design to hit the store shelves.
Congratulations, you've noticed that this year's signature is based on outdated internet memes; CLICK HERE NOW to experience this unforgettable phenomenon. This sentence is just filler and as irrelevant as my signature.
 
maxxcool
Gerbil Elite
Posts: 855
Joined: Thu Sep 12, 2002 8:40 am
Location: %^&*%$$
Contact:

Re: SFF Zen 8c/16t system.

Tue Dec 06, 2016 9:16 pm

Glorious wrote:
maxxcool wrote:
On the next iteration of boxes I tried with a DC 4670 as a test, and satisfied with peaking 70c under full sustained loads converted the remaining 4 devices to DC silicon as well.


I assume DC silicon is some sort of thermal grease then?

If changing that means it went from constantly throttling to OK operation.... like, I don't know, this is getting even crazier. There shouldn't be such a stark difference between properly applied paste.

maxxcool wrote:
It did not work with the stock heatsink, at room temp... But it was not defective so the only other sane conclusion as you noted was the manufacturer was in error of its rating,


Uhh??

My statement was that if it was throttling at room temperature in a reasonable sized/cooled case with the stock cooler, yes, it is defective. They supplied the cooler as part of the retail package, so if that combination does not perform appropriately in a reasonable enclosure, than it is unacceptable.

I don't know if this was the situation with those six cored thubans @ 125w, but if it was, my statement applies to them and I have no problem saying they were defective.

I had a MSI factory overclocked Geforce 460 that had a problem like this: Defective. Some high performance/gaming Laptops have this problem. I've denounced them before.

This is entirely independent of the accuracy of the TDP: if the stocker cooler can't keep up, well, it can't keep up.

I'm not taking your word for it, but I've seen enough dinky stock coolers to fully believe it's entirely possible that the one AMD shipped with your Thubans wasn't suitable.


DC==DEVILSCANYON

And yes all 5 behaved the same... Out of box 100% sustained load decompressing and sha256 hashing hundreds of thousands of files == throttle..

PHONE POST BREVITY YADDA...
Cybert said: Capitlization and periods are hard for you, aren't they? I've given over $100 to techforums. I should have you banned for my money.
 
MOSFET
Gerbil XP
Posts: 373
Joined: Fri Aug 08, 2014 12:42 am

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 12:17 am

For those looking to challenge their own PC building skills in return for a superbly packaged machine that has uncompromised performance and is portable, the Sugo SG10 is the one to get.


That Sugo SG10 has fairly decent looks for what appears to be a pretty tiny mATX cubish case. I bet a 95W CPU would throttle under a maxxcool torture test, though :D
Be careful on inserting this (or any G34 chip) into the socket. Once you pull that restraining lever, it is either a good install or a piece of silicon jewelry.
 
whm1974
Emperor Gerbilius I
Topic Author
Posts: 6361
Joined: Fri Dec 05, 2014 5:29 am

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 12:50 am

synthtel2 wrote:
Oh great, the flamewar squad arrived. :roll:
just brew it!:Unless you really need (or want...) *both* small form factor *and* 8 cores, why wouldn't you go for a lower TDP quad core CPU, or step up to a micro-ATX form factor?
Kretschmer wrote:
Why on earth would you want to use a 16-thread low-clock CPU in a SFF gaming system?

It's just as much a workstation as a gaming system, ease of transport is a big plus, and with my cooling setup the downsides are hardly going to exist. I can't speak for whm1974's situation, though.

Thanks synthtel2, and yes this could be used as a portable workstation or even a portable game server for some folks.

Chrispy_ wrote:
whm1974 wrote:
So IPC is everything then?

Really, absolutely, yes.

This is why Zen's IPC is so critically important; Bulldozer was such a complete turd that actually had significantly lower IPC than the older Phenoms. Clockspeed sort of half hid the IPC defecit, and over the last four years tweaks to make Piledriver/Steamroller/Excavator evolutions have probably pulled it back up to finally surpass Phenom IPC levels. Meanwhile, AMD has been haemorrhaging money trying to sell their hideous mistake for anything they could until Zen was baked. Honestly, Intel made this mistake with the Pentium 4 and it took them half a decade to recover. That's how long it takes to react and change direction for a CPU design to hit the store shelves.

I'll keep this in mind if I ever start a CPU design firm or start designing CPUs.
 
just brew it!
Administrator
Posts: 54500
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 1:51 am

Chrispy_ wrote:
whm1974 wrote:
So IPC is everything then?

Really, absolutely, yes.

This is why Zen's IPC is so critically important; Bulldozer was such a complete turd that actually had significantly lower IPC than the older Phenoms. Clockspeed sort of half hid the IPC defecit, and over the last four years tweaks to make Piledriver/Steamroller/Excavator evolutions have probably pulled it back up to finally surpass Phenom IPC levels. Meanwhile, AMD has been haemorrhaging money trying to sell their hideous mistake for anything they could until Zen was baked. Honestly, Intel made this mistake with the Pentium 4 and it took them half a decade to recover. That's how long it takes to react and change direction for a CPU design to hit the store shelves.

Yes, I did a mental facepalm when I realized that AMD had "pulled a Pentium 4" with Bulldozer. As the saying goes, "Those who fail to learn from history are doomed to repeat it."
Nostalgia isn't what it used to be.
 
whm1974
Emperor Gerbilius I
Topic Author
Posts: 6361
Joined: Fri Dec 05, 2014 5:29 am

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 2:35 am

just brew it! wrote:
Chrispy_ wrote:
whm1974 wrote:
So IPC is everything then?

Really, absolutely, yes.

This is why Zen's IPC is so critically important; Bulldozer was such a complete turd that actually had significantly lower IPC than the older Phenoms. Clockspeed sort of half hid the IPC defecit, and over the last four years tweaks to make Piledriver/Steamroller/Excavator evolutions have probably pulled it back up to finally surpass Phenom IPC levels. Meanwhile, AMD has been haemorrhaging money trying to sell their hideous mistake for anything they could until Zen was baked. Honestly, Intel made this mistake with the Pentium 4 and it took them half a decade to recover. That's how long it takes to react and change direction for a CPU design to hit the store shelves.

Yes, I did a mental facepalm when I realized that AMD had "pulled a Pentium 4" with Bulldozer. As the saying goes, "Those who fail to learn from history are doomed to repeat it."

Hopefully AMD will be able to recover with Zen. Although that will take a while and in the meantime  Intel will not be standing still.
 
Airmantharp
Emperor Gerbilius I
Posts: 6192
Joined: Fri Oct 15, 2004 10:41 pm

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 7:13 am

whm1974 wrote:
Hopefully AMD will be able to recover with Zen. Although that will take a while and in the meantime  Intel will not be standing still.


Intel was still able to move product, by hook or by crook, where AMD currently isn't, and AMD wasn't objectively superior until they rolled the Athlon64 core with the first desktop IMC out. I'll remain surprised if AMD actually pulls out of this, though they do have the fact that their demise would trigger a round of trust-busting for Intel in their corner; Intel isn't really actively competing against them.
 
Glorious
Gerbilus Supremus
Posts: 12343
Joined: Tue Aug 27, 2002 6:35 pm

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 7:28 am

maxxcool wrote:
DC==DEVILSCANYON


Ah Ok, thanks. I honestly was flailing in the dark trying to figure out what you were referring to.

maxxcool wrote:
And yes all 5 behaved the same... Out of box 100% sustained load decompressing and sha256 hashing hundreds of thousands of files == throttle..


I just can't believe this, sorry. Here, let me show you why:

$ openssl speed sha256                                                                                                     
Doing sha256 for 3s on 16 size blocks: 8253329 sha256's in 3.00s                                                                           
Doing sha256 for 3s on 64 size blocks: 4512008 sha256's in 3.00s                                                                           
Doing sha256 for 3s on 256 size blocks: 1931638 sha256's in 3.00s                                                                           
Doing sha256 for 3s on 1024 size blocks: 590103 sha256's in 3.00s                                                                           
Doing sha256 for 3s on 8192 size blocks: 78856 sha256's in 3.00s                                                                           
OpenSSL 1.0.2g  1 Mar 2016                                                     


$ cat /proc/cpuinfo                                                                                                        
processor       : 0                                                                                                                         
vendor_id       : AuthenticAMD                                                                                                             
cpu family      : 16                                                                                                                       
model           : 5                                                                                                                         
model name      : AMD Phenom(tm) II X4 840 Processor     


1024 x 590103 / 3 = ~200 MB/s

And the ssl speed test uses but a single core of my deneb, and you had -six- identical cores with your thuban, how on earth do you expect me to believe, for even the slightest moment, that you were CPU-bound?

No dude, that's absurdly ridiculous.

Try again, part deux?
 
maxxcool
Gerbil Elite
Posts: 855
Joined: Thu Sep 12, 2002 8:40 am
Location: %^&*%$$
Contact:

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 10:47 am

Glorious wrote:
maxxcool wrote:
DC==DEVILSCANYON


Ah Ok, thanks. I honestly was flailing in the dark trying to figure out what you were referring to.

maxxcool wrote:
And yes all 5 behaved the same... Out of box 100% sustained load decompressing and sha256 hashing hundreds of thousands of files == throttle..


I just can't believe this, sorry. Here, let me show you why:

$ openssl speed sha256                                                                                                     
Doing sha256 for 3s on 16 size blocks: 8253329 sha256's in 3.00s                                                                            
Doing sha256 for 3s on 64 size blocks: 4512008 sha256's in 3.00s                                                                            
Doing sha256 for 3s on 256 size blocks: 1931638 sha256's in 3.00s                                                                          
Doing sha256 for 3s on 1024 size blocks: 590103 sha256's in 3.00s                                                                          
Doing sha256 for 3s on 8192 size blocks: 78856 sha256's in 3.00s                                                                            
OpenSSL 1.0.2g  1 Mar 2016                                                    


$ cat /proc/cpuinfo                                                                                                        
processor       : 0                                                                                                                        
vendor_id       : AuthenticAMD                                                                                                              
cpu family      : 16                                                                                                                        
model           : 5                                                                                                                        
model name      : AMD Phenom(tm) II X4 840 Processor      


1024 x 590103 / 3 = ~200 MB/s

And the ssl speed test uses but a single core of my deneb, and you had -six- identical cores with your thuban, how on earth do you expect me to believe, for even the slightest moment, that you were CPU-bound?

No dude, that's absurdly ridiculous.

Try again, part deux?

Not sure what to tell you, your still being combative. 

Back then it was part of my job to take TB drives and hash their file contents.  Some files were mime containers, some were not, some were zipped, some were not, some we uploaded to the array, some we executed on drive that was submitted.  Some we virus scanned 1st then indexed,  Some we did not,  Some we did in vmware on the same devices,  some we did on host.  Some we re-zipped with 7z,  some we did not.   It just depended on the schedule. 
  
I can recall some archive jobs took up to a week to complete with 7z.. and I can recall the extraction and fingerprint tool taking days on 'recovered' email stores because decomposing mime containers takes a dogs age, then have to fingerprint the results of those individual files..  

Some jobs were simultaneous some were not.  it was what ever the client demanded.. jobs in parallel really strained the Thubans and the oem heat sink was not enough.  

That's all I can say .. belief or not. 

edit: word typo :P  too early and between meetings...
Cybert said: Capitlization and periods are hard for you, aren't they? I've given over $100 to techforums. I should have you banned for my money.
 
Glorious
Gerbilus Supremus
Posts: 12343
Joined: Tue Aug 27, 2002 6:35 pm

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 11:07 am

maxxcool wrote:
Not sure what to tell you, your still being combative.


Because what you said was nonsense.

maxxcool wrote:
Back then it was part of my job to take TB drives and hash their file contents.


But there is simply no way you were CPU-bound doing that with SHA256 on a thuban.

I'm sorry, what you are saying just cannot be true.

maxxcool wrote:
Some files were mime containers, some were not, some were zipped, some were not, some we uploaded to the array, some we executed on drive that was submitted. Some we virus scanned 1st then indexed, Some we did not, Some we did in vmware on the same devices, some we did on host.


This is just further evidence of the absurdity: you were actually touching the filesystem.

Cause, maybe, like with 6 drives dd streamed into sha256 you'd have something that would sort of put a load on it, kinda, maybe?

What you are saying? No way, IO-bound, sorry.

maxxcool wrote:
Some we re-zipped with 7z, some we did not. It just depended on the schedule.


Well, LZMA compression certainly would put you into the CPU-bound column, but I don't think 7zip supported LZMA2 (and thus multithreading) in the thuban era.

maxxcool wrote:
Some jobs were simultaneous some were not. it was what ever the client demanded.. jobs in parallel really strained the Thubans and the oem heat sink was not enough.
That's all I can say .. belief or not.


What service, exactly, where you providing?
 
maxxcool
Gerbil Elite
Posts: 855
Joined: Thu Sep 12, 2002 8:40 am
Location: %^&*%$$
Contact:

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 11:09 am

Glorious wrote:
maxxcool wrote:
Not sure what to tell you, your still being combative.


Because what you said was nonsense.

maxxcool wrote:
Back then it was part of my job to take TB drives and hash their file contents.


But there is simply no way you were CPU-bound doing that with SHA256 on a thuban.

I'm sorry, what you are saying just cannot be true.

maxxcool wrote:
Some files were mime containers, some were not, some were zipped, some were not, some we uploaded to the array, some we executed on drive that was submitted.  Some we virus scanned 1st then indexed,  Some we did not,  Some we did in vmware on the same devices,  some we did on host.


This is just further evidence of the absurdity: you were actually touching the filesystem.

Cause, maybe, like with 6 drives dd streamed into sha256 you'd have something that would sort of put a load on it, kinda, maybe?

What you are saying? No way, IO-bound, sorry.

maxxcool wrote:
Some we re-zipped with 7z,  some we did not.   It just depended on the schedule.


Well, LZMA compression certainly would put you into the CPU-bound column, but I don't think 7zip supported LZMA2 (and thus multithreading) in the thuban era.

maxxcool wrote:
Some jobs were simultaneous some were not.  it was what ever the client demanded.. jobs in parallel really strained the Thubans and the oem heat sink was not enough.  
That's all I can say .. belief or not.


What service, exactly, where you providing?

recovering "samples", that's as close as I can word it ... yes we touched the files in many cases.  we unzipped what we could, brute forced what we could, decomposed mime to smaller elements.. we had to touch the files to do that, even if only to copy to the array.  

edit for more details
Cybert said: Capitlization and periods are hard for you, aren't they? I've given over $100 to techforums. I should have you banned for my money.
 
DancinJack
Maximum Gerbil
Posts: 4494
Joined: Sat Nov 25, 2006 3:21 pm
Location: Kansas

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 11:18 am

Why on Earth would you not be using Xeons for that type of work? Consumer level hardware for that kind of stuff is just foolish.
i7 6700K - Z170 - 16GiB DDR4 - GTX 1080 - 512GB SSD - 256GB SSD - 500GB SSD - 3TB HDD- 27" IPS G-sync - Win10 Pro x64 - Ubuntu/Mint x64 :: 2015 13" rMBP Sierra :: Canon EOS 80D/Sony RX100
 
just brew it!
Administrator
Posts: 54500
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 11:22 am

Sometimes in the real world there are cost constraints...
Nostalgia isn't what it used to be.
 
whm1974
Emperor Gerbilius I
Topic Author
Posts: 6361
Joined: Fri Dec 05, 2014 5:29 am

Re: SFF Zen 8c/16t system.

Wed Dec 07, 2016 11:23 am

DancinJack wrote:
Why on Earth would you not be using Xeons for that type of work?  Consumer level hardware for that kind of stuff is just foolish.

I was going to say that does sound like something you would use Xeons for.
  • 1
  • 2
  • 3
  • 4
  • 5
  • 7

Who is online

Users browsing this forum: No registered users and 1 guest
GZIP: On