AMD THREADRIPPER Officially Announced.

RevPuk said:
I kind of want to go with the threadripper, but I need to go with the Intel. My workstation is quite literally my livelihood. I need to upgrade now, but throwing $3k at something as bleeding edge new as Ryzen/TR and running the risk of it not working out just isn't a viable option. I suspect there's a lot of folks in the same position who have to go with the 'safe' option.

I do hope to see Ryzen series perform well as a product, though. The market really needed the competition.
Click to expand...
Like software not working? or?
 
Hagrid said:
Like software not working? or?
Click to expand...

Yep - I use a fair amount of odd ball applications, just stuff most people haven't heard of. Some are from small vendors, others are proprietary within our company. I'm having flashbacks of the late 90's when one of the old AMD K series would occasionaly make 2+2 =5


Grimlaking said:
Ok clearly this is a top of the line build. I wonder why you're going with the 960 EVO 500GB and not the 1TB. Is the 500GB a better performer? My experience has been higher capacity SSD's are actually better performers. Once of the reasons I went with the 1TB myself, and also that I was having capacity issues with my 500GB SSD I was running on before.

Another note... you're spending the cash on 64 GB on an 600 dollar MB to populate all slots. Why not go for 128? Or is that the limit of the threadripper.. I didn't check so a edit may be forthcoming.

I also wonder what the use case is for this system and what you are running for monitors/connectivity to the rest of the world.

What I mean is, if you're not running at least 1 4k monitor preferably 2 that may be overkill. If you are using this to serve virtualized VDI desktops on thin clients then you would want the additional ram and potentially another VEGA card., and definitly the additional Harddrive capacity. If you're NOT running this as a virtual desktop host then you could scale things back unless you have some truly insane graphic editing or other large format files you need to work with. (Again I would think a faster/higher capacity SSD would be desirable in that configuration.)

What I might lean to in that is a pair of 1 TB 960 pro NVME drives in a raid 1 for your OS and operating drives, then a pair of 2 TB regular SSD's or 10TB storage drives also in raid 1 to store the data you've been creating until you can get it to offline storage.

The build you are looking to do is big and impressive but misses the target for any one solution, you're building out the rest of the components to be impressive. Consider your use case and go to town!
Click to expand...

I was planning on much the same regarding the 960 EVO for my next build. I don't even need the 500, and will likely go with two of the 250gb drives. Reason being, the cost/performance is acceptable for my needs and capacity isn't the concern..M.2's are so bonkers fast already, the slight bump is effectively negligible for my particular scenario. In my case, it allots more $$ to go towards other components I would otherwise have to hold off on due to budget. Why buy the whole pig if you just want some bacon?
 
RevPuk said:
Yep - I use a fair amount of odd ball applications, just stuff most people haven't heard of. Some are from small vendors, others are proprietary within our company. I'm having flashbacks of the late 90's when one of the old AMD K series would occasionaly make 2+2 =5




I was planning on much the same regarding the 960 EVO for my next build. I don't even need the 500, and will likely go with two of the 250gb drives. Reason being, the cost/performance is acceptable for my needs and capacity isn't the concern..M.2's are so bonkers fast already, the slight bump is effectively negligible for my particular scenario. In my case, it allots more $$ to go towards other components I would otherwise have to hold off on due to budget. Why buy the whole pig if you just want some bacon?
Click to expand...

Well when the rest of the build looks like a pig farm I question the need for a side of bacon. ;)
 
lutjens said:
Bottom Line: Intel wouldn't have never released a consumer CPU with 18-cores (or even one with >12-cores) if Threadripper hadn't been announced.
Click to expand...

Nice, now continues the reasoning to get why Threadripper has been announced, when it didn't exist on AMD plans a pair of years ago: only AM4 and RyZen for desktop then.
 
juanrga said:
Nice, now continues the reasoning to get why Threadripper has been announced, when it didn't exist on AMD plans a pair of years ago: only AM4 and RyZen for desktop then.
Click to expand...

Ye, no wonder there is only 4 boards to show for it. While there are plenty of x299 to go around.

Even the socket is new and shares with nothing else, despite the same pin count.
 
juanrga said:
Nice, now continues the reasoning to get why Threadripper has been announced, when it didn't exist on AMD plans a pair of years ago: only AM4 and RyZen for desktop then.
Click to expand...


So your condemning them because they quickly adapted to move into the professional market??? You are warped.
 
juanrga said:
Nice, now continues the reasoning to get why Threadripper has been announced, when it didn't exist on AMD plans a pair of years ago: only AM4 and RyZen for desktop then.
Click to expand...
For all we know Threadripper may be rescue packages that don't bin out after assembly for EPYC. Seems doubtful. Just seems like AMD adapting to markets and scalability.
On another note, with recent price drops and so many dies going into MCM, i speculate that another stepping may be coming. Certainly, they will be moving to 14nm+, but revisions are a possibility. Reportedly yields are good and there is a full month of production on the 14LPP at Glofo.
 
lutjens said:
You don't know?
Click to expand...

I wrote about that a pair of times, but would like to see people as you to tell the whole history instead only one side.

gigaxtreme1 said:
For all we know Threadripper may be rescue packages that don't bin out after assembly for EPYC. Seems doubtful. Just seems like AMD adapting to markets and scalability.
Click to expand...

"Adapting to markets" is the key here. One would check market response to their (micro)server stuff.
 
Last edited:
Shintai said:
Ye, no wonder there is only 4 boards to show for it. While there are plenty of x299 to go around.
Click to expand...

The 10':1 ratio of mobos is a combination of late minute planning/scheduling from AMD plus expected impact/sales.

Shintai said:
Even the socket is new and shares with nothing else, despite the same pin count.
Click to expand...

Yes, the new TR4 socket is just a relabeled SP3r2, which is evidently derived from the original SP3 socket for Naples/EPYC.
 
yep watch the pretty balloons go into the air.
Don't look in my right hand look up. Yep its there did you see it?
They can't say anything positive so they double talk and show nothing and call it something.
 
I was with AMD since my first Tbird...then all the way to Phenom II 965 which was my last effort to hold with AMD but then finally jumped to an i7 920...that was a LONG time ago and I am STILL on the same platform with a Xeon X5660 @ 4.4-4.5Ghz depending on season.
The CPU market has been to me, VERY boring for a long time when AMD stopped been competitive in the high end with Intel but with Ryzen things started to look very interesting and now Threadripper REALLY got my attention. I am not gaming much if anything as I mostly work on Lightroom, Photoshop and now doing video for a Youtube channel...so as I am heavy into multi tasking, you can imagine how a 16 core monster got me all pumped. Maybe 2017 marks the year where I finally jump back to AMD. :)
 
sparks said:
yep watch the pretty balloons go into the air.
Don't look in my right hand look up. Yep its there did you see it?
They can't say anything positive so they double talk and show nothing and call it something.
Click to expand...

I don't think anyone is surprised that the boy band show up all over AMD threads.
 
Best part of this whole thing is seeing how wrong Shintai and Juanrga have been so far. 7900X uses over 350 watts yet the EPYC high end chip with 32 cores 64 threads on a 2 socket setup only uses 480 watts total system draw with two processors. Intel may be winning the speed war but they are losing the war in power efficiency. I imagine threadripper to be clocked higher then the EPYC chips so that will hurt the efficiency some but still looking much better then Intel. From what I have seen EPYC will get sales in the server side which should mean some good sales for threadripper as well.
 
Gideon said:
Best part of this whole thing is seeing how wrong Shintai and Juanrga have been so far. 7900X uses over 350 watts yet the EPYC high end chip with 32 cores 64 threads on a 2 socket setup only uses 480 watts total system draw with two processors. Intel may be winning the speed war but they are losing the war in power efficiency. I imagine threadripper to be clocked higher then the EPYC chips so that will hurt the efficiency some but still looking much better then Intel. From what I have seen EPYC will get sales in the server side which should mean some good sales for threadripper as well.
Click to expand...

Did you just ignore the Xeon line and accept AMDs cheating benchmarks too that also only compares it to Broadwell? You know, reducing performance 43% just because...

And did you notice how low base clock and all core turbo the EPYC line got? Good luck with TR and power consumption. Also be careful with your reviews with overclocking even if its MCE, since AVX512 ran full speed without an offset or correct VID. ;)

32C 2.2Ghz base, 2.7Ghz all core turbo 200W. (Yes, AMD got high TDP setting).
32C 2.0Ghz base, 2.6Ghz all core turbo 170W.
24C 2.0Ghz base, 2.8Ghz all core turbo 170W.

Remember what happened the last time AMD used rate benches. And I am not talking about the scandal where they got all their benches invalidated by Spec.org.

4P-SPECint_rate_base2006-Mainstream-Power-Processors.jpg
 
Last edited:
Shintai said:
Did you just ignore the Xeon line and accept AMDs cheating benchmarks too that also only compares it to Broadwell?

And did you notice how low base clock and all core turbo the EPYC line got? Good luck with TR and power consumption. Also be careful with your reviews with overclocking even if its MCE, since AVX512 ran full speed. ;)

32C 2.2Ghz base, 2.7Ghz all core turbo 200W. (Yes, AMD got high TDP setting).
32C 2.0Ghz base, 2.6Ghz all core turbo 170W.
24C 2.0Ghz base, 2.8Ghz all core turbo 170W.
Click to expand...
You said absolutely nothing here. Got any facts to back you up or more random vague statements?

Fact is we all knew, well maybe not all apparently, that in the lower clocks AMDs architecture is extremely efficient. Enough so that most of the knowledgeable individuals are all quite pleased with the performance, and this outcome was fully expected.
 
JustReason said:
You said absolutely nothing here. Got any facts to back you up or more random vague statements?

Fact is we all knew, well maybe not all apparently, that in the lower clocks AMDs architecture is extremely efficient. Enough so that most of the knowledgeable individuals are all quite pleased with the performance, and this outcome was fully expected.
Click to expand...

You mean about the blatant cheat? :)
http://techreport.com/review/32125/amd-epyc-7000-series-cpus-revealed/2

AMD is making some claims about the performance of various Epyc products' performance today, and the initial outlook is good. However, we do have to take issue with a couple of the choices AMD made on the way to its numbers. After compiling SPECint_rate_base2006 with the -O2 flag in GCC, AMD says observed a 43% delta between its internal numbers for the Xeon E5-2699A v4 and public SPEC numbers for similar systems produced using binaries generated by the Intel C++ compiler. In turn, AMD applied that 43% (or 0.575x) across the board to some publicly-available SPECint_rate_base scores for several two-socket Xeon systems.

It's certainly fair to say that SPEC results produced with Intel's compiler might deserve adjustment, but my conversations with other analysts present at the Epyc event suggests that a 43% reduction is optimistic. The -O2 flag for GCC isn't the most aggressive set of optimizations available from that compiler, and SPEC binaries generated accordingly may not be fully representative of binaries compiled in the real world.
Click to expand...

Funny how history repeats.
 
JustReason said:
Maybe your English isn't strong. What they did isn't a cheat, but a less than accurate blanket approach to multiple CPUs from Intel. It is reasonable and generally done in all aspects and products around the world. Unless you think AMD should buy a number of different Intel servers to run some tests.
Click to expand...

The tests are already public and validated. AMD used those and just made up some BS to reduce performance with 43%.

Maybe you missed that part when reading it? You know, English and all? ;)
 
Shintai said:
The tests are already public and validated. AMD used those and just made up some BS to reduce performance with 43%.

Maybe you missed that part when reading it? You know, English and all? ;)
Click to expand...
Read your quote from the event... Optimistic does not equal cheat nor does it state any relevant facts just feelings.
 
Shintai said:
The tests are already public and validated. AMD used those and just made up some BS to reduce performance with 43%.

Maybe you missed that part when reading it? You know, English and all? ;)
Click to expand...

*sigh*

The proper way to handle any of this is to wait for third party benchmarking anyway. Never trust the manufacturer's benchmarks. Goes for Intel as much as AMD.
 
Shintai said:
Did you just ignore the Xeon line and accept AMDs cheating benchmarks too that also only compares it to Broadwell? You know, reducing performance 43% just because...

And did you notice how low base clock and all core turbo the EPYC line got? Good luck with TR and power consumption. Also be careful with your reviews with overclocking even if its MCE, since AVX512 ran full speed without an offset or correct VID. ;)

32C 2.2Ghz base, 2.7Ghz all core turbo 200W. (Yes, AMD got high TDP setting).
32C 2.0Ghz base, 2.6Ghz all core turbo 170W.
24C 2.0Ghz base, 2.8Ghz all core turbo 170W.

Remember what happened the last time AMD used rate benches. And I am not talking about the scandal where they got all their benches invalidated by Spec.org.

4P-SPECint_rate_base2006-Mainstream-Power-Processors.jpg
Click to expand...

Opteron is your argument to me and the fact the only thing I have seen from a third party so far is power draw and their statement that EPYC is fast. But you show me a old chart and say that proves your point, everyone is lying and EPYC sucks... lol. How much does Intel pay you to try to spin anything AMD does?
 
Gideon said:
Best part of this whole thing is seeing how wrong Shintai and Juanrga have been so far. 7900X uses over 350 watts yet the EPYC high end chip with 32 cores 64 threads on a 2 socket setup only uses 480 watts total system draw with two processors. Intel may be winning the speed war but they are losing the war in power efficiency. I imagine threadripper to be clocked higher then the EPYC chips so that will hurt the efficiency some but still looking much better then Intel. From what I have seen EPYC will get sales in the server side which should mean some good sales for threadripper as well.
Click to expand...

Efficiency increases near linearly with number of cores but highly nonlinearly with frequency. Moar cores but lower clocked is going to have better efficiency. Therefore you comparison is not demonstrating what you pretend.

But you know the best part of all this? That the people is now complaining because a fast 10-core chip pushed to 4.6GHz is drawing 380W from the wall is the same people that will remain silent when TR break powermeters...

What amount of power you believe a 10-core TR @4GHz will use? And what astronomical amount of power you believe it would use if this TR chip could be clocked at 4.6GHz?

Hint:

Intel-Core-i9-7900X-Power-Consumption-Bittech.png
 
juanrga said:
Efficiency increases near linearly with number of cores but highly nonlinearly with frequency. Moar cores but lower clocked is going to have better efficiency. Therefore you comparison is not demonstrating what you pretend.

But you know the best part of all this? That the people is now complaining because a fast 10-core chip pushed to 4.6GHz is drawing 380W from the wall is the same people that will remain silent when TR break powermeters...

What amount of power you believe a 10-core TR @4GHz will use? And what astronomical amount of power you believe it would use if this TR chip could be clocked at 4.6GHz?

Hint:

Intel-Core-i9-7900X-Power-Consumption-Bittech.png
Click to expand...

Even at stock look at the power that 7900x consumes 267 watts. 300 watts is near the max AIO and good air coolers can handle at 4.6 its 378 watts which shows a serious issue. At 378 watts your using a custom water cooling system to keep it running at that speed. All the 6950x guys that I have seen over 4.2 all use big custom water cooling systems cause they have little choice. So with a 7900x you got a small bump in clock speeds with a massive increase in power draw.
 
Gideon said:
Even at stock look at the power that 7900x consumes 267 watts. 300 watts is near the max AIO and good air coolers can handle at 4.6 its 378 watts which shows a serious issue. At 378 watts your using a custom water cooling system to keep it running at that speed. All the 6950x guys that I have seen over 4.2 all use big custom water cooling systems cause they have little choice. So with a 7900x you got a small bump in clock speeds with a massive increase in power draw.
Click to expand...

As long as you test with AVX512 loads at full speed, sure. And Prime95 supports AVX512.
 
Gideon, you're wasting your time. Every once in a while, the trifecta of Intel fanyboyism will get something right, and that's not to be undervalued. They were right on Intel's price drops (which I didn't expect). But they are often wrong, as well. If you read enough of their posts, you'll start to see some serious internal contradictions and chart cherry-picking.

But none of that really matters. You won't change their minds, you won't have a productive discussion with them on this. You'll just go back and forth forever and flush away time better spent doing more pleasant things.

The facts are simple: Skylake-X is a faster uarch, both in IPC and clock speed. AMD's Zen uarch has advantages in easy scalability, which generally results in a cheaper product at any given performance tier (with some edge-case exceptions). AMD has to throw more cores at the problem to get there, but they can afford to given the way Zen was designed.

They both have a place. And I'm very happy to see exciting changes in the CPU world again. Finally, >4 cores is reaching affordability and main stream acceptance, and given the general clockspeed and IPC wall we've hit lately, this seems to be the only way out in the near term. Don't let the Intel squad (or AMD fanboys for that matter) tell you any different.
 
Gideon said:
Even at stock look at the power that 7900x consumes 267 watts. 300 watts is near the max AIO and good air coolers can handle at 4.6 its 378 watts which shows a serious issue. At 378 watts your using a custom water cooling system to keep it running at that speed. All the 6950x guys that I have seen over 4.2 all use big custom water cooling systems cause they have little choice. So with a 7900x you got a small bump in clock speeds with a massive increase in power draw.
Click to expand...

A 10-core chip with higher IPC, AVX-512 support, and all-core 4GHz turbo is consuming 3% more power than a 8-core chip with lower IPC, 128bit native, @4GHz. It surely looks as a disaster, but just for the opposite chip to the one that you pretend.

No answer to my questions? Ok, let me then reply myself.

"What amount of power you believe a 10-core TR @4GHz will use?"

Power ~ 259W * (10/8) = 324W

An approximation that is ignoring that a 10-core TR would be in reality a MCM2 chip.

"And what astronomical amount of power you believe it would use if this TR chip could be clocked at 4.6GHz?"

Power ~ 324W *(46/40)^2 = 428W

Again ignoring extra power from MCM2 package. Of course you will need LN2 to achieve those clocks on 14LPP.
 
Last edited:
chenw said:
Is that due to RyZen being so efficient that Prime95 doesn't break a sweat, or due to Prime95's tests can't get to RyZen effectively?
Click to expand...

Second option.

That is why RyZen consumes about same power on a ordinary workload as Luxrender than in a power virus as Prime95

aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9JL1IvNjU2NDUxL29yaWdpbmFsLzAxLVdhdHRhZ2UtQmFycy5wbmc=


That is why RyZen consumes similar power under Excel than under a power virus as Prime95

Power_02.png


Power_01.png



Prime95 is a power virus for Intel chips and for other AMD chips like the FX-8350. It is not a power virus for RyZen.
 
Man the Intel brigade is desperate now. All you got is power usage using as cherry picked a chart as possible with no links to look at the bigger picture of the review they are from or how old the tests are. Intels laughable release in HEDT and these guys grasping at straws must mean AMD has a full winner on hand. Just looking at shelf space and How AMD over took Intel at the local stores here is just proof positive of AMDs impact.
 
juanrga said:
A 10-core chip with higher IPC, AVX-512 support, and all-core 4GHz turbo is consuming 3% more power than a 8-core chip with lower IPC, 128bit native, @4GHz. It surely looks as a disaster, but just for the opposite chip to the one that you pretend.

No answer to my questions? Ok, let me then reply myself.

"What amount of power you believe a 10-core TR @4GHz will use?"

Power ~ 259W * (10/8) = 324W

An approximation that is ignoring that a 10-core TR would be in reality a MCM2 chip.

"And what astronomical amount of power you believe it would use if this TR chip could be clocked at 4.6GHz?"

Power ~ 324W *(46/40)^2 = 428W

Again ignoring extra power from MCM2 package. Of course you will need LN2 to achieve those clocks on 14LPP.
Click to expand...


Only fools and Trolls expect a 4.6 threadripper chip. 14LPP does not support extreme speeds and you know that as the process is optimized for low power draw at lower clock speeds. That is why you were writing off Ryzen before it came out and how it would not be anywhere near 4 ghz and you were wrong as the chips will clock on normal cooling to about 4.2 ghz. I also like now how prime95 is not a proper load for Ryzen.. that is freaking funny. Also I could care less about how AMD or Intel decides to make their chip MCM or not, as long as it works well. The 7900X is a power hog and you need to accept that will see what threadripper does soon.
 
Gideon said:
Only fools and Trolls expect a 4.6 threadripper chip. 14LPP does not support extreme speeds and you know that as the process is optimized for low power draw at lower clock speeds. That is why you were writing off Ryzen before it came out and how it would not be anywhere near 4 ghz and you were wrong as the chips will clock on normal cooling to about 4.2 ghz.
Click to expand...

Sorry I'm not in your debate here really, but this stood out. Hard OCP had really issue with getting the ryzen chips to 4.0 stable. I guess I've missed some articles how everyone is now hitting 4.2 thanks to some ingenious bios updates I hope?
 
Grimlaking said:
Sorry I'm not in your debate here really, but this stood out. Hard OCP had really issue with getting the ryzen chips to 4.0 stable. I guess I've missed some articles how everyone is now hitting 4.2 thanks to some ingenious bios updates I hope?
Click to expand...

It's not super common but it represents about the max clock you can get on Ryzen without LN2 or something more extreme. There is a massive thread on Overclock.net and what people have been getting. It tends to be what voltage your comfortable running the chip at. 3.9 to 4.1 is pretty much the norm but some are hitting 4.2. I will warn you the threads can get pretty big over there as the overclocking 4ghz club for Ryzen is over 1000 pages.
 
I booted up this morning on my Ryzen 1700x to 4056mhz accidently at 1.38v when I adjusted my BCLK to 104 lol. I am just not ready yet (heat removal wise) to play around with >4ghz speeds but my second 1700x is a monster in respects to Ryzen OCing.

As for SkylakeX Power and high temperatures is the new high here, some folks need to review TomsHardware data (at least they are decent in this department), they needed a chiller to remove heat decently and recorded over 300w of power while the motherboard self protected itself by shutting down.

http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-10.html
 
noko said:
I booted up this morning on my Ryzen 1700x to 4056mhz accidently at 1.38v when I adjusted my BCLK to 104 lol. I am just not ready yet (heat removal wise) to play around with >4ghz speeds but my second 1700x is a monster in respects to Ryzen OCing.

As for SkylakeX Power and high temperatures is the new high here, some folks need to review TomsHardware data (at least they are decent in this department), they needed a chiller to remove heat decently and recorded over 300w of power while the motherboard self protected itself by shutting down.

http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-10.html
Click to expand...
So why in the hell have I never seen this graph posted in any i9/TR threads
aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9JL1EvNjg0OTYyL29yaWdpbmFsLzA0LVBvd2VyLUNvbnN1bXB0aW9uLVRvcnR1cmUucG5n


Look at the 7700k and 1800x, seriously so much crap about TDP and yet AMDs 8 core is just barely over Intels 4core.
 
You must log in or register to reply here.
Back
Top