Amd Ryzen Threadripper & X399 MegaThread! FAQ & Resources

[jdwii]

^^^Man perfect scaling basically that's amazing to me.^^^^

 

[goldstone77]

^^^Man perfect scaling basically that's amazing to me.^^^^

Another feather to add to ThreadRipper's hat!

 

[juanrga]

Anand's Workstation CPU recommendations:

https://www.anandtech.com/show/11891/best-cpus-for-workstations-2017

I get that recommendations are subjective. But information he gives to justify his choice of ThreadRipper is plain wrong.

By "workstation" he seems to consider only people using Blender and Corona, so he gives "better performance-per-dollar ratio for almost all our pure throughput tests." Interesting that now workstation CPUs are only used for pure throughput. Main usage of a CPU latency seems gone.

His AMD "offers a full 60 PCIe lanes for coprocessors, compared to 44" for Intel is false. For X299 systems, 44 PCIe lanes are provided by the CPU and uses it's DMI 3.0 link to the PCH. 24 lanes are provided by the X299 PCH. So total number of lanes is 44+24=68. All of them on the Intel platform are PCIe 3.0 compliant. AMD has 64 lanes provided by the CPU, 4 of which are used for the connection to the X399 PCH. This leaves 60 CPU based PCIe 3.0 lanes which are usable. The X399 PCH has extra 8 PCIe lanes but they are only PCIe 2.0 compliant. Thus, AMD platform provides a total of 64-4+8=68 PCIe lanes. Not only Intel provides the same number of lanes than AMD, but all on Intel platforms are PCIe3, whereas AMD ThreadRipper gives a mixture of PCie 3 and Pcie 2. Moreover, how many lanes are usable on X399 depends on the motherboard's design and what's integrated into them. For instance, the GIGABYTE X399 Aorus Gaming 7 leaves you with 48 usable PCIe lanes via expansion slots. because M.2, USB 3.1, etc. consume the rest of available lanes.

Also Ian ignores anything specificaly related to workstation stuff, such as reliability and other aspects of the platfform. AMD's SATA performance is weaker and less flexible. Even with the latest AGESA code, AMD still lags behind Intel in SATA performance. It's array controls for SATA don't support the same amount of stripe sizes or RAID levels. Intel again has a more flexible RAID configuration than AMD does. Intel can handle up to 20 NVMe devices and multiple arrays. It is likely that applications like Video Toasters, ProTools, AutoDesk products, 3D Studio Max, and Lightwave which sometimes have proprietary hardware paired with them or not, will see certification of Intel's platform before AMD's.

So it is not so simple as Ian pretends, and I guess people that is purchasing Intel has done his homework and knows what is purchasing and why.

 

[jdwii]

"Also Ian ignores anything specificaly related to workstation stuff, such as reliability and other aspects of the platfform"

Probably over the fact that makes Amd a no brainer since threadripper supports ECC memory

 

[juanrga]

I have read some interesting remarks about ECC:

I think it's stupid that Intel doesn't offer ECC support, but again I think people on these forums overstate it's importance. ECC RAM is not a magic bullet that makes systems that much more reliable. I get why people think that reading about it on Wikipedia and shit, but in the real world, I've seen it make little difference in workstations. Again I've seen ECC equipped systems that crash all the time and regular boxes without it that don't crash for anything. If they need the absolute performance, aren't interested in overclocking and need a better and more flexible NVMe and storage subsystem, Intel's where it's at. Businesses can write off the cost of the CPU and 1k on a high end build is usually not a deal breaker unless they need to procure lots of them. Even then, you might still be able to sell them on it for other reasons.

The discussion continued with people asking if anyone has "actually seen working ECC on TR". No one replied, not even TR users. Also ECC enabled is contrary to running RAM overclocked, and most reviews tested TR with RAM overclocked. So if we add ECC support as bonus to the X399 platform, we have to ignore all the reviews that tested it with RAM oberclocked and if we consider the performance bonus that brings running TR with RAM overclocked then we have to ignore ECC support.

So the whole situation is much more complex than that Ian pretends with his recommendations, and as stated before he is giving incorrect information about lanes and so.

 

[-Fran-]

Anand's Workstation CPU recommendations:

https://www.anandtech.com/show/11891/best-cpus-for-workstations-2017

I get that recommendations are subjective. But information he gives to justify his choice of ThreadRipper is plain wrong.

By "workstation" he seems to consider only people using Blender and Corona, so he gives "better performance-per-dollar ratio for almost all our pure throughput tests." Interesting that now workstation CPUs are only used for pure throughput. Main usage of a CPU latency seems gone.

His AMD "offers a full 60 PCIe lanes for coprocessors, compared to 44" for Intel is false. For X299 systems, 44 PCIe lanes are provided by the CPU and uses it's DMI 3.0 link to the PCH. 24 lanes are provided by the X299 PCH. So total number of lanes is 44+24=68. All of them on the Intel platform are PCIe 3.0 compliant. AMD has 64 lanes provided by the CPU, 4 of which are used for the connection to the X399 PCH. This leaves 60 CPU based PCIe 3.0 lanes which are usable. The X399 PCH has extra 8 PCIe lanes but they are only PCIe 2.0 compliant. Thus, AMD platform provides a total of 64-4+8=68 PCIe lanes. Not only Intel provides the same number of lanes than AMD, but all on Intel platforms are PCIe3, whereas AMD ThreadRipper gives a mixture of PCie 3 and Pcie 2. Moreover, how many lanes are usable on X399 depends on the motherboard's design and what's integrated into them. For instance, the GIGABYTE X399 Aorus Gaming 7 leaves you with 48 usable PCIe lanes via expansion slots. because M.2, USB 3.1, etc. consume the rest of available lanes.

Also Ian ignores anything specificaly related to workstation stuff, such as reliability and other aspects of the platfform. AMD's SATA performance is weaker and less flexible. Even with the latest AGESA code, AMD still lags behind Intel in SATA performance. It's array controls for SATA don't support the same amount of stripe sizes or RAID levels. Intel again has a more flexible RAID configuration than AMD does. Intel can handle up to 20 NVMe devices and multiple arrays. It is likely that applications like Video Toasters, ProTools, AutoDesk products, 3D Studio Max, and Lightwave which sometimes have proprietary hardware paired with them or not, will see certification of Intel's platform before AMD's.

So it is not so simple as Ian pretends, and I guess people that is purchasing Intel has done his homework and knows what is purchasing and why.

Why are you ranting at me for it? lol

Plus, this tasty bit you wrote: "Also Ian ignores anything specifically related to workstation stuff, such as reliability and other aspects of the platform". Like jdwii points out, Intel doesn't offer ECC for anything other than Xeons. That is a bazillion less points for any serious workstation work "on the cheap". His paragraph explaining deserves a friggin' Gold medal:

When memory has an error rate of 1 error per GB per year, using 512GB ensures almost two bit errors per day: something that a 60-day simulation would find catastrophic.

And TR supports, how much RAM? Not cheap and probably overkill for HEDT enthusiasts, but anyway, how much?

Everything else you say sounds really bitter. Hell, I have never ever had an AMD platform fail on me before, but my i7 has given me headaches with RAID from time to time and they even had a reliability issue with a chipset revision involving sATA ports. Don't you forget about P67rev1! Irony!

And yes, it is as simple as Ian puts it there. For anyone looking to do SERIOUS work, they know *exactly* what they're looking for and most of the time won't bother with "normal" site reviews and go directly to big name benchmark sites for internal reference numbers. What Ian (my take) is doing, is just addressing the new question a lot of people might be having now thanks to AMD: "oh, what is this workstation-on-the-cheap you're talking about? how can I use this processing power?".

So, in short, sounds like you're desperately trying to make excuses here. I won't humor you beyond this point.

 

[juanrga]

Anand's Workstation CPU recommendations:

https://www.anandtech.com/show/11891/best-cpus-for-workstations-2017

I get that recommendations are subjective. But information he gives to justify his choice of ThreadRipper is plain wrong.

By "workstation" he seems to consider only people using Blender and Corona, so he gives "better performance-per-dollar ratio for almost all our pure throughput tests." Interesting that now workstation CPUs are only used for pure throughput. Main usage of a CPU latency seems gone.

His AMD "offers a full 60 PCIe lanes for coprocessors, compared to 44" for Intel is false. For X299 systems, 44 PCIe lanes are provided by the CPU and uses it's DMI 3.0 link to the PCH. 24 lanes are provided by the X299 PCH. So total number of lanes is 44+24=68. All of them on the Intel platform are PCIe 3.0 compliant. AMD has 64 lanes provided by the CPU, 4 of which are used for the connection to the X399 PCH. This leaves 60 CPU based PCIe 3.0 lanes which are usable. The X399 PCH has extra 8 PCIe lanes but they are only PCIe 2.0 compliant. Thus, AMD platform provides a total of 64-4+8=68 PCIe lanes. Not only Intel provides the same number of lanes than AMD, but all on Intel platforms are PCIe3, whereas AMD ThreadRipper gives a mixture of PCie 3 and Pcie 2. Moreover, how many lanes are usable on X399 depends on the motherboard's design and what's integrated into them. For instance, the GIGABYTE X399 Aorus Gaming 7 leaves you with 48 usable PCIe lanes via expansion slots. because M.2, USB 3.1, etc. consume the rest of available lanes.

Also Ian ignores anything specificaly related to workstation stuff, such as reliability and other aspects of the platfform. AMD's SATA performance is weaker and less flexible. Even with the latest AGESA code, AMD still lags behind Intel in SATA performance. It's array controls for SATA don't support the same amount of stripe sizes or RAID levels. Intel again has a more flexible RAID configuration than AMD does. Intel can handle up to 20 NVMe devices and multiple arrays. It is likely that applications like Video Toasters, ProTools, AutoDesk products, 3D Studio Max, and Lightwave which sometimes have proprietary hardware paired with them or not, will see certification of Intel's platform before AMD's.

So it is not so simple as Ian pretends, and I guess people that is purchasing Intel has done his homework and knows what is purchasing and why.

Why are you ranting at me for it? lol

Plus, this tasty bit you wrote: "Also Ian ignores anything specifically related to workstation stuff, such as reliability and other aspects of the platform". Like jdwii points out, Intel doesn't offer ECC for anything other than Xeons. That is a bazillion less points for any serious workstation work "on the cheap". His paragraph explaining deserves a friggin' Gold medal:

When memory has an error rate of 1 error per GB per year, using 512GB ensures almost two bit errors per day: something that a 60-day simulation would find catastrophic.

And TR supports, how much RAM? Not cheap and probably overkill for HEDT enthusiasts, but anyway, how much?

Everything else you say sounds really bitter. Hell, I have never ever had an AMD platform fail on me before, but my i7 has given me headaches with RAID from time to time and they even had a reliability issue with a chipset revision involving sATA ports. Don't you forget about P67rev1! Irony!

And yes, it is as simple as Ian puts it there. For anyone looking to do SERIOUS work, they know *exactly* what they're looking for and most of the time won't bother with "normal" site reviews and go directly to big name benchmark sites for internal reference numbers. What Ian (my take) is doing, is just addressing the new question a lot of people might be having now thanks to AMD: "oh, what is this workstation-on-the-cheap you're talking about? how can I use this processing power?".

So, in short, sounds like you're desperately trying to make excuses here. I won't humor you beyond this point.

I didn't mention you. I am analyzing the recommendations given by the Anandtech editor. And explaining the flaws in his arguments. I just answered jdwii post about ECC support. And the rest of my arguments aren't excuses, but facts. Ian claims that AMD has an advantage in PCIe lanes but as shown in a former post it is just the opposite: Intel has the advantage here. Intel offers 44+24 PCIe3 lanes, whereas AMD offers 60 PCIe3 lanes plus 8 PCIe2 lanes. How can be reporting those facts considered excuses? Ian is not counting the lanes given by the PCH on each platform, that is an fact. That Intel offers more options/performance on RAID/SATA/NVMe is another fact.

 

[jdwii]

Anand's Workstation CPU recommendations:

https://www.anandtech.com/show/11891/best-cpus-for-workstations-2017

I get that recommendations are subjective. But information he gives to justify his choice of ThreadRipper is plain wrong.

By "workstation" he seems to consider only people using Blender and Corona, so he gives "better performance-per-dollar ratio for almost all our pure throughput tests." Interesting that now workstation CPUs are only used for pure throughput. Main usage of a CPU latency seems gone.

His AMD "offers a full 60 PCIe lanes for coprocessors, compared to 44" for Intel is false. For X299 systems, 44 PCIe lanes are provided by the CPU and uses it's DMI 3.0 link to the PCH. 24 lanes are provided by the X299 PCH. So total number of lanes is 44+24=68. All of them on the Intel platform are PCIe 3.0 compliant. AMD has 64 lanes provided by the CPU, 4 of which are used for the connection to the X399 PCH. This leaves 60 CPU based PCIe 3.0 lanes which are usable. The X399 PCH has extra 8 PCIe lanes but they are only PCIe 2.0 compliant. Thus, AMD platform provides a total of 64-4+8=68 PCIe lanes. Not only Intel provides the same number of lanes than AMD, but all on Intel platforms are PCIe3, whereas AMD ThreadRipper gives a mixture of PCie 3 and Pcie 2. Moreover, how many lanes are usable on X399 depends on the motherboard's design and what's integrated into them. For instance, the GIGABYTE X399 Aorus Gaming 7 leaves you with 48 usable PCIe lanes via expansion slots. because M.2, USB 3.1, etc. consume the rest of available lanes.

Also Ian ignores anything specificaly related to workstation stuff, such as reliability and other aspects of the platfform. AMD's SATA performance is weaker and less flexible. Even with the latest AGESA code, AMD still lags behind Intel in SATA performance. It's array controls for SATA don't support the same amount of stripe sizes or RAID levels. Intel again has a more flexible RAID configuration than AMD does. Intel can handle up to 20 NVMe devices and multiple arrays. It is likely that applications like Video Toasters, ProTools, AutoDesk products, 3D Studio Max, and Lightwave which sometimes have proprietary hardware paired with them or not, will see certification of Intel's platform before AMD's.

So it is not so simple as Ian pretends, and I guess people that is purchasing Intel has done his homework and knows what is purchasing and why.

Why are you ranting at me for it? lol

Plus, this tasty bit you wrote: "Also Ian ignores anything specifically related to workstation stuff, such as reliability and other aspects of the platform". Like jdwii points out, Intel doesn't offer ECC for anything other than Xeons. That is a bazillion less points for any serious workstation work "on the cheap". His paragraph explaining deserves a friggin' Gold medal:

When memory has an error rate of 1 error per GB per year, using 512GB ensures almost two bit errors per day: something that a 60-day simulation would find catastrophic.

And TR supports, how much RAM? Not cheap and probably overkill for HEDT enthusiasts, but anyway, how much?

Everything else you say sounds really bitter. Hell, I have never ever had an AMD platform fail on me before, but my i7 has given me headaches with RAID from time to time and they even had a reliability issue with a chipset revision involving sATA ports. Don't you forget about P67rev1! Irony!

And yes, it is as simple as Ian puts it there. For anyone looking to do SERIOUS work, they know *exactly* what they're looking for and most of the time won't bother with "normal" site reviews and go directly to big name benchmark sites for internal reference numbers. What Ian (my take) is doing, is just addressing the new question a lot of people might be having now thanks to AMD: "oh, what is this workstation-on-the-cheap you're talking about? how can I use this processing power?".

So, in short, sounds like you're desperately trying to make excuses here. I won't humor you beyond this point.

I didn't mention you. I am analyzing the recommendations given by the Anandtech editor. And explaining the flaws in his arguments. I just answered jdwii post about ECC support. And the rest of my arguments aren't excuses, but facts. Ian claims that AMD has an advantage in PCIe lanes but as shown in a former post it is just the opposite: Intel has the advantage here. Intel offers 44+24 PCIe3 lanes, whereas AMD offers 60 PCIe3 lanes plus 8 PCIe2 lanes. How can be reporting those facts considered excuses? Ian is not counting the lanes given by the PCH on each platform, that is an fact. That Intel offers more options/performance on RAID/SATA/NVMe is another fact.

https://hardforum.com/threads/intel-core-i9-7980xe.1944295/

Just so others can read where you got that info from. Funny enough he also said he would never buy a I9 7980xe over that crazy price.

X299 has platform advantages and disadvantages for example here

http://techreport.com/news/32020/computex-2017-intel-unleashes-nvme-raid-for-x299

Clear disadvantage for the already expensive boards. ECC memory is a nice + regardless and having the ability for NVMe RAID out of the box is well a nice plus and i can find many uses for such a task.

One aspect that is true is that Amd really needs to up their controllers it would seem like Intel always offers just a little bit more speed on sata connections and even USB connections i noticed this for years. Not sure if its just a windows thing or if linux/unix can do better?

 

[juanrga]

Anand's Workstation CPU recommendations:

https://www.anandtech.com/show/11891/best-cpus-for-workstations-2017

I get that recommendations are subjective. But information he gives to justify his choice of ThreadRipper is plain wrong.

By "workstation" he seems to consider only people using Blender and Corona, so he gives "better performance-per-dollar ratio for almost all our pure throughput tests." Interesting that now workstation CPUs are only used for pure throughput. Main usage of a CPU latency seems gone.

His AMD "offers a full 60 PCIe lanes for coprocessors, compared to 44" for Intel is false. For X299 systems, 44 PCIe lanes are provided by the CPU and uses it's DMI 3.0 link to the PCH. 24 lanes are provided by the X299 PCH. So total number of lanes is 44+24=68. All of them on the Intel platform are PCIe 3.0 compliant. AMD has 64 lanes provided by the CPU, 4 of which are used for the connection to the X399 PCH. This leaves 60 CPU based PCIe 3.0 lanes which are usable. The X399 PCH has extra 8 PCIe lanes but they are only PCIe 2.0 compliant. Thus, AMD platform provides a total of 64-4+8=68 PCIe lanes. Not only Intel provides the same number of lanes than AMD, but all on Intel platforms are PCIe3, whereas AMD ThreadRipper gives a mixture of PCie 3 and Pcie 2. Moreover, how many lanes are usable on X399 depends on the motherboard's design and what's integrated into them. For instance, the GIGABYTE X399 Aorus Gaming 7 leaves you with 48 usable PCIe lanes via expansion slots. because M.2, USB 3.1, etc. consume the rest of available lanes.

Also Ian ignores anything specificaly related to workstation stuff, such as reliability and other aspects of the platfform. AMD's SATA performance is weaker and less flexible. Even with the latest AGESA code, AMD still lags behind Intel in SATA performance. It's array controls for SATA don't support the same amount of stripe sizes or RAID levels. Intel again has a more flexible RAID configuration than AMD does. Intel can handle up to 20 NVMe devices and multiple arrays. It is likely that applications like Video Toasters, ProTools, AutoDesk products, 3D Studio Max, and Lightwave which sometimes have proprietary hardware paired with them or not, will see certification of Intel's platform before AMD's.

So it is not so simple as Ian pretends, and I guess people that is purchasing Intel has done his homework and knows what is purchasing and why.

Why are you ranting at me for it? lol

Plus, this tasty bit you wrote: "Also Ian ignores anything specifically related to workstation stuff, such as reliability and other aspects of the platform". Like jdwii points out, Intel doesn't offer ECC for anything other than Xeons. That is a bazillion less points for any serious workstation work "on the cheap". His paragraph explaining deserves a friggin' Gold medal:

When memory has an error rate of 1 error per GB per year, using 512GB ensures almost two bit errors per day: something that a 60-day simulation would find catastrophic.

And TR supports, how much RAM? Not cheap and probably overkill for HEDT enthusiasts, but anyway, how much?

Everything else you say sounds really bitter. Hell, I have never ever had an AMD platform fail on me before, but my i7 has given me headaches with RAID from time to time and they even had a reliability issue with a chipset revision involving sATA ports. Don't you forget about P67rev1! Irony!

And yes, it is as simple as Ian puts it there. For anyone looking to do SERIOUS work, they know *exactly* what they're looking for and most of the time won't bother with "normal" site reviews and go directly to big name benchmark sites for internal reference numbers. What Ian (my take) is doing, is just addressing the new question a lot of people might be having now thanks to AMD: "oh, what is this workstation-on-the-cheap you're talking about? how can I use this processing power?".

So, in short, sounds like you're desperately trying to make excuses here. I won't humor you beyond this point.

I didn't mention you. I am analyzing the recommendations given by the Anandtech editor. And explaining the flaws in his arguments. I just answered jdwii post about ECC support. And the rest of my arguments aren't excuses, but facts. Ian claims that AMD has an advantage in PCIe lanes but as shown in a former post it is just the opposite: Intel has the advantage here. Intel offers 44+24 PCIe3 lanes, whereas AMD offers 60 PCIe3 lanes plus 8 PCIe2 lanes. How can be reporting those facts considered excuses? Ian is not counting the lanes given by the PCH on each platform, that is an fact. That Intel offers more options/performance on RAID/SATA/NVMe is another fact.

https://hardforum.com/threads/intel-core-i9-7980xe.1944295/

Just so others can read where you got that info from. Funny enough he also said he would never buy a I9 7980xe over that crazy price.

X299 has platform advantages and disadvantages for example here

http://techreport.com/news/32020/computex-2017-intel-unleashes-nvme-raid-for-x299

Clear disadvantage for the already expensive boards. ECC memory is a nice + regardless and having the ability for NVMe RAID out of the box is well a nice plus and i can find many uses for such a task.

One aspect that is true is that Amd really needs to up their controllers it would seem like Intel always offers just a little bit more speed on sata connections and even USB connections i noticed this for years. Not sure if its just a windows thing or if linux/unix can do better?

And he explained "I primarily play games and only occasionally do things which would necessitate more than 8 CPU cores and threads such as encoding. As a gaming CPU, the Core i9 7980XE is not worth the price increase over the Core i9 7900X."

And he also said that business wouldn't care about pricing because

Intel has better IPC performance, better BIOS code, better memory compatibility, and a more flexible platform. These are the things that businesses often care about in a perormance oriented build. They also need hardware that's certified for certain applications which is another area where AMD will lag Intel.

Yes X299 and X399 have advantages and disadvantages. My point is that Anandtech's Ian didn't mention several advantages of the X299 platform, and even gave wrong info, as for instance when he talked about PCie lanes.

 

[-Fran-]

Oh, Juan, I see you conveniently ignored the "PCIe" lanes discussion from Ian. He's been trying to correct everyone on the subject, but I'll give you a summary:

- Intel has more than 44 lanes *on paper*, but in reality, the CPU maxes out at 44 and the extra 24 are provided by the north bridge. If you're planning to use the lanes for latency sensitive stuff, even with AMDs "less than stellar" drivers, they *will* reach parity in such tasks. No amount of moaning from either camp will change that fact. AMD is just faster, on paper, in regards to overall latency and bandwidth when it comes to PCIe lanes and Intel knows that.
- Intel (via marketing, obviously) is now trying to paint their platforms in the same light as AMD's "MOAR MOAR" stuff. PCIe lanes is now their new battle horse in their slides, from what I understood from Ian's rants. Hence you trying to paint it in a better light.

So, again, you need to read beyond the paper here. That is what Ian is doing (or trying to).

 

[jdwii]

What i find interesting is people calling AnandTech anti Intel now like WHAT i've been reading on these sites for the past 10 years and i all i kept reading was how they are Amd haters haha same as Techspot when Amd fanboys claimed Steve was a Intel fanboy and now he is all in love with ryzen.

Thing is people really aren't that logical i noticed some claiming weird claims like IPC is higher on ryzen then kaby and i laugh but then i hear from the other side that its sandy level which it isn't except for maybe X87 instructions which is never used now.

Anyone else notice some great deals on Amd processors the Ryzen 1800X is 399.99$ and the 1600X is 199.99$ today.

Keep this in mine
Ryzen being two quad cores on one die makes it cheaper to manufacture compared to a true 8 core
Ryzen doesn't have 30% of its die being a igpu
Amd doesn't have to fund a foundry

Amd can lower Ryzen prices to fight coffee-lake pretty sure that's gonna happen

Edit
https://www.pcgamesn.com/intel-coffee-lake-limited-stock

Been stated that coffee-lake is here a good quarter earlier then it was supposed to be causing shortages which people have been waiting for coffee-lake results vs ryzen if one product is not there what will the user buy?

Stock is reported to not improve until mid December or later and Pinnacle Ridge is coming I knew Intel wouldn't have pushed coffee-lake this soon.

Shortages=higher prices too.

 

[8350rocks]

Could you please post a review of threadripper review with multiple boards using water cooling. I think your sub -ambient cooler is still too exotic.
My ideal review will be a threadripper @ 100% load (probably synthetic) at elevated ambient. But at least IF I could know the delta in a variety of workloads, it would help in motherboard selection.
And yes I'm a wimp but please do NOT devote too much time with pure FPS. Gaming + streaming+ few background tasks(like an antivirus scan + ) are fine .

Really to OC threadripper to 4.0ghz locked i would aim for custom loop 360 rad minimum or look for a 360 rad AIO
https://www.techspot.com/review/1465-amd-ryzen-threadripper-1950x-1920x/page7.html

Supposedly kraken x62 @ 280mm does fine...

 

[juanrga]

Oh, Juan, I see you conveniently ignored the "PCIe" lanes discussion from Ian. He's been trying to correct everyone on the subject, but I'll give you a summary:

- Intel has more than 44 lanes *on paper*, but in reality, the CPU maxes out at 44 and the extra 24 are provided by the north bridge. If you're planning to use the lanes for latency sensitive stuff, even with AMDs "less than stellar" drivers, they *will* reach parity in such tasks. No amount of moaning from either camp will change that fact. AMD is just faster, on paper, in regards to overall latency and bandwidth when it comes to PCIe lanes and Intel knows that.
- Intel (via marketing, obviously) is now trying to paint their platforms in the same light as AMD's "MOAR MOAR" stuff. PCIe lanes is now their new battle horse in their slides, from what I understood from Ian's rants. Hence you trying to paint it in a better light.

So, again, you need to read beyond the paper here. That is what Ian is doing (or trying to).

As just explained in former posts, Ian is only counting lanes coming directly from the CPU whereas ignores the lanes comming from the PCH. He is ignoring that Intel provides PCIe3 for all platform lanes whereas AMD gives a mixture of PCI3 and PCie2 lanes. He is ignoring any other advantage of the X299 platform related to RAID/SATA/NVMe. He also ignores that those CPU lanes are maximum possible and not what one can find in X399 boards. Thus, the motherboard editor mentioned how the GIGABYTE X399 Aorus Gaming 7 leaves users only 48 usable PCIe lanes via expansion slots. because M.2, USB 3.1, etc. consume the rest of available lanes.

What i find interesting is people calling AnandTech anti Intel now like WHAT i've been reading on these sites for the past 10 years and i all i kept reading was how they are Amd haters haha same as Techspot when Amd fanboys claimed Steve was a Intel fanboy and now he is all in love with ryzen.

Anandtech was humorously named IntelTech due to his bias when Anand was the boss. Anand is gone and new boses are biased in the opposite way. The site is now humorously named AMDTech. Not only reviews, but forums as well. This is well-known and discussed in several forums. Once the media reported an investigation about the dirty tactics used by biased mods. The title was: "Anandtech forums on trial: The corruption runs deep". I hope someone publishes something similar now, because the situations is 10times worse than then.

As many people has noted, Steve has been publishing a series of ridiculous articles. Either he is biased or he is just incompetent. Some gems in his recent reviews

1) The guy tested in GPU-bound and frame-limiting settings. Crippling the performance of Intel chips. That is the reason why OC the 7800X by a huge 34% did only bring 3% higher framerates (the CPU was being bottlenecked). The guy tried to excuse himself pretending that OC the 7800X didn't bring net performance gains because, in his opinion, there was something wrong with the Skylake-X chip. But there was nothing wring with the Skylake-X chip, because the KBL 7700k by 16% did only bring 2% higher framerates (the Kabylake CPU was being bottlenecked as well). The problem was on his testing using GPU-bound and frame-limiting settings.

2) The guy finally admitted he didn't even test a retail 7800X chip, but he tested a sample. Funny enough in no part of the review he make explicit that he wasn't testing a retail chip.

3) If this wasn't enough, he used a motherboard is incompatible with the 7800X and he managed to burn the CPU thanks to that.

4) He often get high power consumption figures for Intel and low for AMD.

There are lots of discussions about Steve reviews in other forums, and people discussing if he is biased or just incompetent.

 

[8350rocks]

Oh, Juan, I see you conveniently ignored the "PCIe" lanes discussion from Ian. He's been trying to correct everyone on the subject, but I'll give you a summary:

- Intel has more than 44 lanes *on paper*, but in reality, the CPU maxes out at 44 and the extra 24 are provided by the north bridge. If you're planning to use the lanes for latency sensitive stuff, even with AMDs "less than stellar" drivers, they *will* reach parity in such tasks. No amount of moaning from either camp will change that fact. AMD is just faster, on paper, in regards to overall latency and bandwidth when it comes to PCIe lanes and Intel knows that.
- Intel (via marketing, obviously) is now trying to paint their platforms in the same light as AMD's "MOAR MOAR" stuff. PCIe lanes is now their new battle horse in their slides, from what I understood from Ian's rants. Hence you trying to paint it in a better light.

So, again, you need to read beyond the paper here. That is what Ian is doing (or trying to).

As just explained in former posts, Ian is only counting lanes coming directly from the CPU whereas ignores the lanes comming from the PCH. He is ignoring that Intel provides PCIe3 for all platform lanes whereas AMD gives a mixture of PCI3 and PCie2 lanes. He is ignoring any other advantage of the X299 platform related to RAID/SATA/NVMe. He also ignores that those CPU lanes are maximum possible and not what one can find in X399 boards. Thus, the motherboard editor mentioned how the GIGABYTE X399 Aorus Gaming 7 leaves users only 48 usable PCIe lanes via expansion slots. because M.2, USB 3.1, etc. consume the rest of available lanes.

What i find interesting is people calling AnandTech anti Intel now like WHAT i've been reading on these sites for the past 10 years and i all i kept reading was how they are Amd haters haha same as Techspot when Amd fanboys claimed Steve was a Intel fanboy and now he is all in love with ryzen.

Anandtech was humorously named IntelTech due to his bias when Anand was the boss. Anand is gone and new boses are biased in the opposite way. The site is now humorously named AMDTech. Not only reviews, but forums as well. This is well-known and discussed in several forums. Once the media reported an investigation about the dirty tactics used by biased mods. The title was: "Anandtech forums on trial: The corruption runs deep". I hope someone publishes something similar now, because the situations is 10times worse than then.

As many people has noted, Steve has been publishing a series of ridiculous articles. Either he is biased or he is just incompetent. Some gems in his recent reviews

1) The guy tested in GPU-bound and frame-limiting settings. Crippling the performance of Intel chips. That is the reason why OC the 7800X by a huge 34% did only bring 3% higher framerates (the CPU was being bottlenecked). The guy tried to excuse himself pretending that OC the 7800X didn't bring net performance gains because, in his opinion, there was something wrong with the Skylake-X chip. But there was nothing wring with the Skylake-X chip, because the KBL 7700k by 16% did only bring 2% higher framerates (the Kabylake CPU was being bottlenecked as well). The problem was on his testing using GPU-bound and frame-limiting settings.

2) The guy finally admitted he didn't even test a retail 7800X chip, but he tested a sample. Funny enough in no part of the review he make explicit that he wasn't testing a retail chip.

3) If this wasn't enough, he used a motherboard is incompatible with the 7800X and he managed to burn the CPU thanks to that.

4) He often get high power consumption figures for Intel and low for AMD.

There are lots of discussions about Steve reviews in other forums, and people discussing if he is biased or just incompetent.

Is this similar to the claims of 20% better Instructions Per Clock Per Core touted by some forum posters, when reality shows 6-8%?

 

[juanrga]

Is this similar to the claims of 20% better Instructions Per Clock Per Core touted by some forum posters, when reality shows 6-8%?

Those forum posters are simply reporting reviews findings, aren't?

 

[-Fran-]

Oh, Juan, I see you conveniently ignored the "PCIe" lanes discussion from Ian. He's been trying to correct everyone on the subject, but I'll give you a summary:

- Intel has more than 44 lanes *on paper*, but in reality, the CPU maxes out at 44 and the extra 24 are provided by the north bridge. If you're planning to use the lanes for latency sensitive stuff, even with AMDs "less than stellar" drivers, they *will* reach parity in such tasks. No amount of moaning from either camp will change that fact. AMD is just faster, on paper, in regards to overall latency and bandwidth when it comes to PCIe lanes and Intel knows that.
- Intel (via marketing, obviously) is now trying to paint their platforms in the same light as AMD's "MOAR MOAR" stuff. PCIe lanes is now their new battle horse in their slides, from what I understood from Ian's rants. Hence you trying to paint it in a better light.

So, again, you need to read beyond the paper here. That is what Ian is doing (or trying to).

As just explained in former posts, Ian is only counting lanes coming directly from the CPU whereas ignores the lanes comming from the PCH. He is ignoring that Intel provides PCIe3 for all platform lanes whereas AMD gives a mixture of PCI3 and PCie2 lanes. He is ignoring any other advantage of the X299 platform related to RAID/SATA/NVMe. He also ignores that those CPU lanes are maximum possible and not what one can find in X399 boards. Thus, the motherboard editor mentioned how the GIGABYTE X399 Aorus Gaming 7 leaves users only 48 usable PCIe lanes via expansion slots. because M.2, USB 3.1, etc. consume the rest of available lanes.

What i find interesting is people calling AnandTech anti Intel now like WHAT i've been reading on these sites for the past 10 years and i all i kept reading was how they are Amd haters haha same as Techspot when Amd fanboys claimed Steve was a Intel fanboy and now he is all in love with ryzen.

Anandtech was humorously named IntelTech due to his bias when Anand was the boss. Anand is gone and new boses are biased in the opposite way. The site is now humorously named AMDTech. Not only reviews, but forums as well. This is well-known and discussed in several forums. Once the media reported an investigation about the dirty tactics used by biased mods. The title was: "Anandtech forums on trial: The corruption runs deep". I hope someone publishes something similar now, because the situations is 10times worse than then.

As many people has noted, Steve has been publishing a series of ridiculous articles. Either he is biased or he is just incompetent. Some gems in his recent reviews

1) The guy tested in GPU-bound and frame-limiting settings. Crippling the performance of Intel chips. That is the reason why OC the 7800X by a huge 34% did only bring 3% higher framerates (the CPU was being bottlenecked). The guy tried to excuse himself pretending that OC the 7800X didn't bring net performance gains because, in his opinion, there was something wrong with the Skylake-X chip. But there was nothing wring with the Skylake-X chip, because the KBL 7700k by 16% did only bring 2% higher framerates (the Kabylake CPU was being bottlenecked as well). The problem was on his testing using GPU-bound and frame-limiting settings.

2) The guy finally admitted he didn't even test a retail 7800X chip, but he tested a sample. Funny enough in no part of the review he make explicit that he wasn't testing a retail chip.

3) If this wasn't enough, he used a motherboard is incompatible with the 7800X and he managed to burn the CPU thanks to that.

4) He often get high power consumption figures for Intel and low for AMD.

There are lots of discussions about Steve reviews in other forums, and people discussing if he is biased or just incompetent.

Is this similar to the claims of 20% better Instructions Per Clock Per Core touted by some forum posters, when reality shows 6-8%?

Ironic, isn't it?

Cheers!

 

[goldstone77]

[video="https://www.youtube.com/watch?v=tj-b7A4_5nU&ab_channel=LinusTechTips"][/video]
Ultimate Mega Workstation Build Guide
Linus Tech Tips
Published on Oct 4, 2017

 

[8350rocks]

Is this similar to the claims of 20% better Instructions Per Clock Per Core touted by some forum posters, when reality shows 6-8%?

Those forum posters are simply reporting reviews findings, aren't?

I have yet to see a reputable reviewer showing anything close to 20%.

 

[8350rocks]

Spoiler

Oh, Juan, I see you conveniently ignored the "PCIe" lanes discussion from Ian. He's been trying to correct everyone on the subject, but I'll give you a summary:

- Intel has more than 44 lanes *on paper*, but in reality, the CPU maxes out at 44 and the extra 24 are provided by the north bridge. If you're planning to use the lanes for latency sensitive stuff, even with AMDs "less than stellar" drivers, they *will* reach parity in such tasks. No amount of moaning from either camp will change that fact. AMD is just faster, on paper, in regards to overall latency and bandwidth when it comes to PCIe lanes and Intel knows that.
- Intel (via marketing, obviously) is now trying to paint their platforms in the same light as AMD's "MOAR MOAR" stuff. PCIe lanes is now their new battle horse in their slides, from what I understood from Ian's rants. Hence you trying to paint it in a better light.

So, again, you need to read beyond the paper here. That is what Ian is doing (or trying to).

As just explained in former posts, Ian is only counting lanes coming directly from the CPU whereas ignores the lanes comming from the PCH. He is ignoring that Intel provides PCIe3 for all platform lanes whereas AMD gives a mixture of PCI3 and PCie2 lanes. He is ignoring any other advantage of the X299 platform related to RAID/SATA/NVMe. He also ignores that those CPU lanes are maximum possible and not what one can find in X399 boards. Thus, the motherboard editor mentioned how the GIGABYTE X399 Aorus Gaming 7 leaves users only 48 usable PCIe lanes via expansion slots. because M.2, USB 3.1, etc. consume the rest of available lanes.

What i find interesting is people calling AnandTech anti Intel now like WHAT i've been reading on these sites for the past 10 years and i all i kept reading was how they are Amd haters haha same as Techspot when Amd fanboys claimed Steve was a Intel fanboy and now he is all in love with ryzen.

Anandtech was humorously named IntelTech due to his bias when Anand was the boss. Anand is gone and new boses are biased in the opposite way. The site is now humorously named AMDTech. Not only reviews, but forums as well. This is well-known and discussed in several forums. Once the media reported an investigation about the dirty tactics used by biased mods. The title was: "Anandtech forums on trial: The corruption runs deep". I hope someone publishes something similar now, because the situations is 10times worse than then.

As many people has noted, Steve has been publishing a series of ridiculous articles. Either he is biased or he is just incompetent. Some gems in his recent reviews

1) The guy tested in GPU-bound and frame-limiting settings. Crippling the performance of Intel chips. That is the reason why OC the 7800X by a huge 34% did only bring 3% higher framerates (the CPU was being bottlenecked). The guy tried to excuse himself pretending that OC the 7800X didn't bring net performance gains because, in his opinion, there was something wrong with the Skylake-X chip. But there was nothing wring with the Skylake-X chip, because the KBL 7700k by 16% did only bring 2% higher framerates (the Kabylake CPU was being bottlenecked as well). The problem was on his testing using GPU-bound and frame-limiting settings.

2) The guy finally admitted he didn't even test a retail 7800X chip, but he tested a sample. Funny enough in no part of the review he make explicit that he wasn't testing a retail chip.

3) If this wasn't enough, he used a motherboard is incompatible with the 7800X and he managed to burn the CPU thanks to that.

4) He often get high power consumption figures for Intel and low for AMD.

There are lots of discussions about Steve reviews in other forums, and people discussing if he is biased or just incompetent.

Is this similar to the claims of 20% better Instructions Per Clock Per Core touted by some forum posters, when reality shows 6-8%?

Ironic, isn't it?

Cheers!

Indeed.

 

[jdwii]

Is this similar to the claims of 20% better Instructions Per Clock Per Core touted by some forum posters, when reality shows 6-8%?

Those forum posters are simply reporting reviews findings, aren't?

I have yet to see a reputable reviewer showing anything close to 20%.

If you want i can troll you with some X87 results haha also in Dolphin it can be 20% but that is over intel's major gains once haswell hit in that application. CEMU is a new emulator sure you heard of it but people are posting good results with Ryzen i tested this myself but they don't have a benchmark to compare other CPU's.

 

[8350rocks]

Is this similar to the claims of 20% better Instructions Per Clock Per Core touted by some forum posters, when reality shows 6-8%?

Those forum posters are simply reporting reviews findings, aren't?

I have yet to see a reputable reviewer showing anything close to 20%.

If you want i can troll you with some X87 results haha also in Dolphin it can be 20% but that is over intel's major gains once haswell hit in that application. CEMU is a new emulator sure you heard of it but people are posting good results with Ryzen i tested this myself but they don't have a benchmark to compare other CPU's.

x87 lol Well played

 

[gamerk316]

Except in the case of quads, it's multithreading of the GPU driver that's killing performance. Before DX12, you could do 90% of the work on just two threads due to the GPU rendering thread being mostly single-threaded in nature. With DX12's changes to the driver, 90% of the work is done in about six or seven threads due to the ability to thread out the driver. Yeah, quads are going to have problems with that, even if each thread isn't doing that much work, due to the latency involved in thread switching.

But hey, you got "better" threading out of it.

You lost me a bit, gamerk. Are you against or in favor of going wide with software? Particularly drivers.

You make it sound like you're against by these sort of comments, but I have trouble understanding why would that be the case...

Cheers!

I'm just pointing out that using more threads has the side-effect of causing a lot of latency related issues on processors with less cores. If you are going to have a GPU driver that uses 4+ major threads instead of just one, yeah, quads aren't going to cut it anymore, and even 4+4 CPUs are going to struggle a bit.

 

[-Fran-]

Spoiler

Except in the case of quads, it's multithreading of the GPU driver that's killing performance. Before DX12, you could do 90% of the work on just two threads due to the GPU rendering thread being mostly single-threaded in nature. With DX12's changes to the driver, 90% of the work is done in about six or seven threads due to the ability to thread out the driver. Yeah, quads are going to have problems with that, even if each thread isn't doing that much work, due to the latency involved in thread switching.

But hey, you got "better" threading out of it.

You lost me a bit, gamerk. Are you against or in favor of going wide with software? Particularly drivers.

You make it sound like you're against by these sort of comments, but I have trouble understanding why would that be the case...

Cheers!

I'm just pointing out that using more threads has the side-effect of causing a lot of latency related issues on processors with less cores. If you are going to have a GPU driver that uses 4+ major threads instead of just one, yeah, quads aren't going to cut it anymore, and even 4+4 CPUs are going to struggle a bit.

Noted. I do believe shifting the processing bottleneck is not a bad thing as long as you still have room to grow. If the modern graphic APIs allow drivers to "go wide", it's better than expecting Intel and AMD to increase IPC to accommodate your software deficiencies.

Cheers!

 

[gamerk316]

Spoiler

Except in the case of quads, it's multithreading of the GPU driver that's killing performance. Before DX12, you could do 90% of the work on just two threads due to the GPU rendering thread being mostly single-threaded in nature. With DX12's changes to the driver, 90% of the work is done in about six or seven threads due to the ability to thread out the driver. Yeah, quads are going to have problems with that, even if each thread isn't doing that much work, due to the latency involved in thread switching.

But hey, you got "better" threading out of it.

You lost me a bit, gamerk. Are you against or in favor of going wide with software? Particularly drivers.

You make it sound like you're against by these sort of comments, but I have trouble understanding why would that be the case...

Cheers!

I'm just pointing out that using more threads has the side-effect of causing a lot of latency related issues on processors with less cores. If you are going to have a GPU driver that uses 4+ major threads instead of just one, yeah, quads aren't going to cut it anymore, and even 4+4 CPUs are going to struggle a bit.

Noted. I do believe shifting the processing bottleneck is not a bad thing as long as you still have room to grow. If the modern graphic APIs allow drivers to "go wide", it's better than expecting Intel and AMD to increase IPC to accommodate your software deficiencies.

Cheers!

I now point out using more threads in games is actually going to make the above problem a LOT worse. Getting into a situation where you have over a dozen threads that all want to do significant (measurable) work fighting for a handful of CPU cores is going to lead to a lot of hard to reproduce performance loss. We're going too far in the other direction, and getting into a situation where 16 cores is quickly going to be considered requisite for gaming. And that's going to cost everyone more money to maintain the same performance.

 

[-Fran-]

Spoiler

Except in the case of quads, it's multithreading of the GPU driver that's killing performance. Before DX12, you could do 90% of the work on just two threads due to the GPU rendering thread being mostly single-threaded in nature. With DX12's changes to the driver, 90% of the work is done in about six or seven threads due to the ability to thread out the driver. Yeah, quads are going to have problems with that, even if each thread isn't doing that much work, due to the latency involved in thread switching.

But hey, you got "better" threading out of it.

You lost me a bit, gamerk. Are you against or in favor of going wide with software? Particularly drivers.

You make it sound like you're against by these sort of comments, but I have trouble understanding why would that be the case...

Cheers!

I'm just pointing out that using more threads has the side-effect of causing a lot of latency related issues on processors with less cores. If you are going to have a GPU driver that uses 4+ major threads instead of just one, yeah, quads aren't going to cut it anymore, and even 4+4 CPUs are going to struggle a bit.

Noted. I do believe shifting the processing bottleneck is not a bad thing as long as you still have room to grow. If the modern graphic APIs allow drivers to "go wide", it's better than expecting Intel and AMD to increase IPC to accommodate your software deficiencies.

Cheers!

I now point out using more threads in games is actually going to make the above problem a LOT worse. Getting into a situation where you have over a dozen threads that all want to do significant (measurable) work fighting for a handful of CPU cores is going to lead to a lot of hard to reproduce performance loss. We're going too far in the other direction, and getting into a situation where 16 cores is quickly going to be considered requisite for gaming. And that's going to cost everyone more money to maintain the same performance.

Uhm... Since ST performance is not moving up as fast as before, I think it's the other way around. I do agree, if you have games that peg all available CPU cores 100%, you'll get into a nasty situation, but it's the same as having the same game pegging your single core CPU at 100% leaving no room for the driver to maneuver. Maybe you're still thinking of the driver just pegging a single core to 100% while the games use the rest of the cores. That can be done as well, but sounds less optimal than making the driver go as wide as it can and leave the OS scheduler to handle the assignments.

Cheers!