AMD Readies Nine Ryzen Threadripper Models

[SARVAMANGALAM]

Maybe Claw Hammer would have been more fitting hehe

And now the production of the largest water pumps and blocks will begin....
Just imagine: the pump will must be huge and the cooling radiator will must have to be large for three 14 inch ventilators. Like corsair H110 I but even bigger... so price that coolers will around 150 usd and more .. so if you want two TH ryper .. add 300+ usd just for proper cooling ....

like now i can not get stable more than 3500 mhz on ryzen 1700x with avx application , run on 100% load all cpu .. chipset have +60C-hwinfo and heatsink is danger to touch ..burn finger.. so i must find some new cooler for chipset on GA-AX370-Gaming K3..(lol) still i have some unawares restart on whole system ,,dont now why ,,cca every 2 day .. ram have two 8gb 2133 ddr4 soo its not about ram speed crashing proably . now trying beta f3 bios
but thank you amd for move UP on this whole cpu sector for

Now we would need "amd ryzen " in the ssd and m.2 and pcie sector ..
becouse prices on 1-4 tbytes ssd and m.2 is crazy.. many people dont need 3500 write and read BUT we need some 1500-2000read and 800-1000 write betwen. but data BIGGER than 1 tb... 2-4 tb becouse after format you have always much lover data .. just my opnon

 

[callaway]

Anyone think this will have the same bandwidth/clock issues like the Ryzen chips using the infinity fabric clocked to memory speed or did they take a new approach to prevent that with the new chipset?

 

[Steevo]

Well, was that really so unexpected? When you have tons of cores, you need better binned chips. You think Intel is not doing that?

I partially expected them to use better cores on their release product X1800. Perhaps a new spin has given them slightly better speeds or they really did plan it this way and are seriously back in the running against Intel.

I think you might have that slightly wrong, Im pretty sure what your seeing is still only 1 core at 4.1GHz XFR Boost its just that it bounces around from core to core which then in HW Monitor records it. What you need to look at is the first list under "value" as you run a benchmark/game whatever and then watch how many cores at once is clocked up to 4.1GHz and im pretty sure it will only be 1 core at a time doing so but it will jump around from core to core.

4.1Ghz, plus 200Mhz more for XFR as long as the temp is low enough. So 4.3Ghz guaranteed speeds for some cores.

Anyone think this will have the same bandwidth/clock issues like the Ryzen chips using the infinity fabric clocked to memory speed or did they take a new approach to prevent that with the new chipset?

I am guessing the current BIOS updates plus it seems they have two infinity fabric interfaces between the MCM and seem to be doing something with some PCIe lanes, meaning its the same and either just tied together or in parallel, and I would expect future BIOS updates to allow this chip to use the higher speed as they fine tune with board makers.

 

[figuretti]

Oh boy...
Prepare your suits guys... this gonna be a slaughter fest...
16C/32T, 3.2 Ghz base, 3.6 Ghz boost, No XFR, 849 U$S
Source:
http://semiaccurate.com/forums/showpost.php?p=290705&postcount=594

 

[efikkan]

Broadwell-e is often worse for multi-threaded professional stuff and ryzen isn't too far off in games. 1440p+ ryzen is just better. Skylake-x will beat it, but zen2 isn't to far off, so we'll see.

Ryzen loses >10% in many games, some even 15-20%. So if you're buying a GTX 1070 or better, you'll actually start wasting a lot of money because of the CPU. Broadwell-E is excellent in multithreading, the scenarios where Ryzen pulls ahead are cache optimized workloads, due to it's wider issue width.

 

[Steevo]

Ryzen loses >10% in many games, some even 15-20%. So if you're buying a GTX 1070 or better, you'll actually start wasting a lot of money because of the CPU. Broadwell-E is excellent in multithreading, the scenarios where Ryzen pulls ahead are cache optimized workloads, due to it's wider issue width.

Got facts to back that up, cause the low side framerate is equal at resolutions that matter, and on many games that have single thread the G4560 kicks ass too but how many complain about its multi-threaded performance.

https://www.techpowerup.com/reviews/AMD/Ryzen_5_1400/20.html

Stock X1800 is 6% slower than 7700K at 1440 and 1% at 4K

 

[Camm]

Ryzen loses >10% in many games, some even 15-20%. So if you're buying a GTX 1070 or better, you'll actually start wasting a lot of money because of the CPU. Broadwell-E is excellent in multithreading, the scenarios where Ryzen pulls ahead are cache optimized workloads, due to it's wider issue width.

Well there's some flat out bullshit for the day. Even RotTR, which was the game with the biggest differential between 1800X and 7770K just got patched removing that gap down to 4%.

 

[callaway]

I am guessing the current BIOS updates plus it seems they have two infinity fabric interfaces between the MCM and seem to be doing something with some PCIe lanes, meaning its the same and either just tied together or in parallel, and I would expect future BIOS updates to allow this chip to use the higher speed as they fine tune with board makers.

Thanks, I'd love to get one, but I think I'll be waiting for the Ryzen Plus/2 release, as the TR chips while fairly priced (I think the 12 Core will go for ~$549 from what I read), but the MB and Memory will be very expensive, too cost prohibitive for me.

 

[Melvis]

XFR is able to boost way more than 100 MHz under the right conditions. Zen is designed to adjust it's clock speed very frequently, way quicker than any HW monitor is able to detect.

I partially expected them to use better cores on their release product X1800. Perhaps a new spin has given them slightly better speeds or they really did plan it this way and are seriously back in the running against Intel.

4.1Ghz, plus 200Mhz more for XFR as long as the temp is low enough. So 4.3Ghz guaranteed speeds for some cores..

I get what your saying guys and I dont really care what clock speed the XFR can get to but he is saying in his post that ALL 6 of his 8 cores are clocking to 4.1 or whatever XFR Boost speed at the SAME TIME. Im pretty sure this isnt the case and it is only 1 core that boost this high at one time, not multipule cores its just that it bounces around from core to core and in which then HW Monitor is then recording it making it look like 6 cores when he needs to look at the current value when the XFR kicks in.

 

[Deleted member 172152]

Ryzen loses >10% in many games, some even 15-20%. So if you're buying a GTX 1070 or better, you'll actually start wasting a lot of money because of the CPU. Broadwell-E is excellent in multithreading, the scenarios where Ryzen pulls ahead are cache optimized workloads, due to it's wider issue width.

But if you have 6-cores+ you're probably doing professional stuff as well and have at least a 1440p monitor, in which case ryzen is basically only cheaper. Also ryzen support is a bit rubbish still and in nost newer games the difference should be smaller and the threadripper 12-core seems to be a bit faster, so maybe even single-core performance will be similar to broadwell-e or even skylake-x.

 

[Steevo]

I get what your saying guys and I dont really care what clock speed the XFR can get to but he is saying in his post that ALL 6 of his 8 cores are clocking to 4.1 or whatever XFR Boost speed at the SAME TIME. Im pretty sure this isnt the case and it is only 1 core that boost this high at one time, not multipule cores its just that it bounces around from core to core and in which then HW Monitor is then recording it making it look like 6 cores when he needs to look at the current value when the XFR kicks in.

All cores can boost to maximum boost clocks if the cooling allows for it, 2 cores will also get XFR at an additional 100-200Mhz usually, or if he has setup a different profile with Ryzen Master he could be pushing all cores to 4.1Ghz as long as the cooling is good enough.

 

[Vayra86]

Well there's some flat out bullshit for the day. Even RotTR, which was the game with the biggest differential between 1800X and 7770K just got patched removing that gap down to 4%.

That's what happens when you glance over and skim review benchmarks and call it 'research'.

Youtube is full of those guys. On top of it all, a 1070 is nowhere near strong enough to bottleneck a Ryzen CPU in the first place, even if its a 4c/8t.

 

[Melvis]

All cores can boost to maximum boost clocks if the cooling allows for it, 2 cores will also get XFR at an additional 100-200Mhz usually, or if he has setup a different profile with Ryzen Master he could be pushing all cores to 4.1Ghz as long as the cooling is good enough.

Im pretty sure this isnt the case, there was a whole thread dedicated to this exact issue and it turns out that yes if you have the cooling then it will boost to its XFR boost clock speed but only one 1 core not multipule cores. The max boost clock speed of ALL cores regardless of his cooling is only 3.7GHZ (base clock 3.6) it will not get any higher clock speed on all cores.

 

[Deleted member 172152]

Im pretty sure this isnt the case, there was a whole thread dedicated to this exact issue and it turns out that yes if you have the cooling then it will boost to its XFR boost clock speed but only one 1 core not multipule cores. The max boost clock speed of ALL cores regardless of his cooling is only 3.7GHZ (base clock 3.6) it will not get any higher clock speed on all cores.

Not all cores, but maybe 2 or three since there are quite a few more cores in threadripper cpu's than ryzen 3-7.

 

[efikkan]

But if you have 6-cores+ you're probably doing professional stuff as well and have at least a 1440p monitor, in which case ryzen is basically only cheaper.

In a world where theoretical specs matters, yes.
In the real world actual performance matters. i7-6800K provides better overall performance than Ryzen 7 1800X. It doesn't matter that it has more cores when it's beaten by one with fewer. It's the same useless argument which was used for Bulldozer back in the day; Bulldozer was "better" due to higher core count, even though it was beaten by a quad core. Ryzen 7 1800X is a competitor to Intel's 6-cores.

Also ryzen support is a bit rubbish still and in nost newer games the difference should be smaller and the threadripper 12-core seems to be a bit faster

Ryzen doesn't offer any new features games needs to be optimized for, so that's 100% BS.

so maybe even single-core performance will be similar to broadwell-e or even skylake-x.

How would putting more of the same dies in a CPU improve single core performance?
Broadwell-E is >20% faster per core. Intel claims Skylake-X will be >15% on top of that, which might be a stretch, but it's certainly not going to be slower.

 

[FR@NK]

Ryzen doesn't offer any new features games needs to be optimized for, so that's 100% BS.

AMD thinks the market needs to react to their hardware...thats just not how it works especially when you have such a small share of the total CPUs. I really think AMD could of had a big win if they released a single CCX 4 core version of their ryzen that was capable of higher clocks. Instead we are stuck with eight slow cores or some cut down version of it.

 

[Deleted member 172152]

AMD thinks the market needs to react to their hardware...thats just not how it works especially when you have such a small share of the total CPUs. I really think AMD could of had a big win if they released a single CCX 4 core version of their ryzen that was capable of higher clocks. Instead we are stuck with eight slow cores or some cut down version of it.

Difference in 1080p games is small and only important for pros, in 1440p there is barely any difference and in multi-threaded pro stuff ryzen 7 is between a 6900k and a 695x on average and for most a 1500x or 1600x works just fine and the major advantage of AMD is that you don't need to delid to prevent throttling and high temps, stock or oc-ed.

In the end clockspeed often still matters if core count us the same, but really if you need a 8-core ryzen and/or have the gpu for it to bottleneck, you should be gaming in 1440p at least where there is no noticeable difference in game (although some say ryzen 7 gives a smoother experience than a 6900k).

In a world where theoretical specs matters, yes.
In the real world actual performance matters. i7-6800K provides better overall performance than Ryzen 7 1800X. It doesn't matter that it has more cores when it's beaten by one with fewer. It's the same useless argument which was used for Bulldozer back in the day; Bulldozer was "better" due to higher core count, even though it was beaten by a quad core. Ryzen 7 1800X is a competitor to Intel's 6-cores.


Ryzen doesn't offer any new features games needs to be optimized for, so that's 100% BS.


How would putting more of the same dies in a CPU improve single core performance?
Broadwell-E is >20% faster per core. Intel claims Skylake-X will be >15% on top of that, which might be a stretch, but it's certainly not going to be slower.

6800k/7800x aren't ryzen competitors. They're basically for those few people with 1080p monitors and 1080 ti's, where everything is bottlenecked including the gpu. 1440p is the way to go for normal gamers and if you're doing professional stuff, you're probably better off having two more cores than higher clockspeeds. Also tdp is lower and there's no need to delid or any chance even delidding isn't going to stop dangerous temperature spikes, so yet another reason sane people should buy into AMD for extra cores rather than intel for more clockspeed.

Ryzen's infininty fabric and other things mean it works a little different and optimization goes a long way in games like AOTS for example.

More of same cores? Probably not. Clockspeeds are sometimes higher than ryzen 5x and 7x, so there have been some optimizations. Also, 3200mhz memory is supported by AMD themselves, which is considerably higher than ryzen 5 and 7 ram speed support, so I suspect a few things were improved there as well.

As for clockspeeds, well it seems that baseclocks are sometimes higher and boostclocks aren't that for off with xfr. If IPC has improved slightly as well, which it probably has, it should be an evenly matched fight for the HEDT platform, untill you overclock of course, but most don't want to loose their warranty on such expensive components and delidding is probably necessary with intel to stop your cpu getting even hotter with that most likely rubbish thermal paste. Was there something else? O wait, threadripper is going to be a few percent slower in worst case scenarios now zen support is pretty good and it might beat intel in some other scenarios, so on average about the same non-overclocked and threadripper should be considerably cheaper!

 

[efikkan]

AMD thinks the market needs to react to their hardware...thats just not how it works especially when you have such a small share of the total CPUs. I really think AMD could of had a big win if they released a single CCX 4 core version of their ryzen that was capable of higher clocks. Instead we are stuck with eight slow cores or some cut down version of it.

There is nothing inherent in Zen requiring software to be written for it. In fact, Intel have support for more instructions. AMD and it's fans always claims their hardware really is superior, but just lacks the software for it. The same argument was used for Bulldozer, it was supposed to be glorious once the software arrived, and of course it never did.

If you want to optimize a game in terms of CPU load, it would be eliminating overhead in the rendering code. Those optimizations have nothing to do with the CPU, and would benefit both.

Difference in 1080p games is small and only important for pros, in 1440p there is barely any difference and in multi-threaded pro stuff ryzen 7 is between a 6900k and a 695x on average and for most a 1500x or 1600x works just fine and the major advantage of AMD is that you don't need to delid to prevent throttling and high temps, stock or oc-ed.

"multi-threaded pro stuff"?
Ryzen is not better at multithreaded workloads in general, just specific benchmarks. In some workloads, even i7-7700K beats Ryzen 7 1800X, such as Photoshop.

In the end clockspeed often still matters if core count us the same, but really if you need a 8-core ryzen and/or have the gpu for it to bottleneck, you should be gaming in 1440p at least where there is no noticeable difference in game (although some say ryzen 7 gives a smoother experience than a 6900k).
...
you're probably better off having two more cores than higher clockspeeds.

Performance is much more complicated than clockspeed, especially with more efficient architectures. Ryzen are in fact running at higher speed than Broadwell-E, so you guys needs to stop claiming it's lack of clock scaling for Ryzen.

Ryzen have a inferior prefetcher, which means it's unable to feed the execution ports efficiently for workloads that are not cache optimized. That's why you see Ryzen crush it in Blender and some encoding and compression workloads, while it's crushed in important loads like Photoshop, games, web browsing, etc.

Also tdp is lower and there's no need to delid or any chance even delidding isn't going to stop dangerous temperature spikes

Now you're just silly. Delidding is just done for extreme overclocking, that's not even remotely relevant.

Ryzen's infininty fabric and other things mean it works a little different and optimization goes a long way in games like AOTS for example.

That makes no sense whatsoever.

More of same cores? Probably not. Clockspeeds are sometimes higher than ryzen 5x and 7x, so there have been some optimizations. Also, 3200mhz memory is supported by AMD themselves, which is considerably higher than ryzen 5 and 7 ram speed support, so I suspect a few things were improved there as well.

There are no improvements in architecture in Threadripper over Ryzen 7.
You know very well that both AMD and Intel CPUs are able of that memory speed, and more.

As for clockspeeds, well it seems that baseclocks are sometimes higher and boostclocks aren't that for off with xfr. If IPC has improved slightly as well, which it probably has, it should be an evenly matched fight for the HEDT platform, untill you overclock of course, but most don't want to loose their warranty on such expensive components and delidding is probably necessary with intel to stop your cpu getting even hotter with that most likely rubbish thermal paste. Was there something else? O wait, threadripper is going to be a few percent slower in worst case scenarios now zen support is pretty good and it might beat intel in some other scenarios, so on average about the same non-overclocked and threadripper should be considerably cheaper!

You are ridiculous. Stop your delidding nonsense. IPC is better for Broadwell-E, and will be even better for Skylake-X. As I've mentioned, AMD need a better prefetcher for Zen2, because higher boost can't make up for cache misses, since the penalty is a constant.

 

[r9]

In a world where theoretical specs matters, yes.
Ryzen doesn't offer any new features games needs to be optimized for, so that's 100% BS.

Dota, AofS and Rise of TR got patches so far fps increase 20%-30%.
New games would already be optimized.
What does that tell you about the relative performance.
I guess you just suck at googling.
And about that IPC crap.
Ryzen is not behind.
Simple fkin math 7700k 4.2GHz is clocked 20% higher than 1500x 3.5GHz.
But beats it by only 12% in games. Provable not even that now, in the TPU review Tomb Raider is not patched.
How does that tell you that Ryzen has lower IPC ?

 

[Deleted member 172152]

There is nothing inherent in Zen requiring software to be written for it. In fact, Intel have support for more instructions. AMD and it's fans always claims their hardware really is superior, but just lacks the software for it. The same argument was used for Bulldozer, it was supposed to be glorious once the software arrived, and of course it never did.

If you want to optimize a game in terms of CPU load, it would be eliminating overhead in the rendering code. Those optimizations have nothing to do with the CPU, and would benefit both.


"multi-threaded pro stuff"?
Ryzen is not better at multithreaded workloads in general, just specific benchmarks. In some workloads, even i7-7700K beats Ryzen 7 1800X, such as Photoshop.


Performance is much more complicated than clockspeed, especially with more efficient architectures. Ryzen are in fact running at higher speed than Broadwell-E, so you guys needs to stop claiming it's lack of clock scaling for Ryzen.

Ryzen have a inferior prefetcher, which means it's unable to feed the execution ports efficiently for workloads that are not cache optimized. That's why you see Ryzen crush it in Blender and some encoding and compression workloads, while it's crushed in important loads like Photoshop, games, web browsing, etc.


Now you're just silly. Delidding is just done for extreme overclocking, that's not even remotely relevant.


That makes no sense whatsoever.


There are no improvements in architecture in Threadripper over Ryzen 7.
You know very well that both AMD and Intel CPUs are able of that memory speed, and more.


You are ridiculous. Stop your delidding nonsense. IPC is better for Broadwell-E, and will be even better for Skylake-X. As I've mentioned, AMD need a better prefetcher for Zen2, because higher boost can't make up for cache misses, since the penalty is a constant.

Learn english, read my comment and then give a reply to what I wrote, so I can take you seriously. See you in ten years or more.

 

[SKD007]

I think you might have that slightly wrong, Im pretty sure what your seeing is still only 1 core at 4.1GHz XFR Boost its just that it bounces around from core to core which then in HW Monitor records it. What you need to look at is the first list under "value" as you run a benchmark/game whatever and then watch how many cores at once is clocked up to 4.1GHz and im pretty sure it will only be 1 core at a time doing so but it will jump around from core to core.

Na I am sure... I ran some bench and tried out few games and I see same result. Some times it's 4 cores some time its 6cores or sometime it's just 2cores.

 

[FR@NK]

Some times it's 4 cores some time its 6cores or sometime it's just 2cores.

Its not running 6 cores at 4.1 at the same time. It switches between different cores boosting 1 or 2 cores up to 4.1 while the other cores are running slower.

Source

 

[Melvis]

Na I am sure... I ran some bench and tried out few games and I see same result. Some times it's 4 cores some time its 6cores or sometime it's just 2cores.

Again your only showing the end result "max* result of when a core hit 4.1. What you need to record is the "value" results and take a screen shot of your core speed when under full load. You should then notice that there will be only 1 core clocking up to 4.1GHz at one time but it will be bouncing around from core to core, but there wont be 6 cores clocking up to 4.1

Its not running 6 cores at 4.1 at the same time. It switches between different cores boosting 1 or 2 cores up to 4.1 while the other cores are running slower.

Source

Exactly!

 

[SKD007]

Again your only showing the end result "max* result of when a core hit 4.1. What you need to record is the "value" results and take a screen shot of your core speed when under full load. You should then notice that there will be only 1 core clocking up to 4.1GHz at one time but it will be bouncing around from core to core, but there wont be 6 cores clocking up to 4.1



Exactly!

Will try that then.