AMD Zen Rumours Point to Earlier Than Expected Release

buttons said:
If you were offered an I7-6700k $450 or a $63 AMD 860k with a $349 geforce 970GTX OC which would you take? No brainer there, id take the AMD.

A lot of you have a lot more disposable income then I do, but i see the i7-6700k as a giant waste of money that doesnt give me any "tangible" increase over an overclocked i7-920. We could argue about this, but my use case is "mild gaming and surfing the web". Games I play mostly play fine on and AMD 6000+

My sons a10-5800k with a 970gtx plays everything i have absolutely flawless and an 860k is 20% faster... meanwhile its $63 shipped from amazon.
Click to expand...

I just built a (very light) gaming box out of an AM1 sempron 3850. I rececycled a 64G ssd, a mini-itx case and spend $29.99 cpu + $0.01 motherboard + $34.99 8G ddr3-1600 and it is way faster than you'd expect. It can also run dota2 (all I tried so far) but am going to try portal & some other games. It's really a $100 light gaming computer....amazing.
 
a3venom said:
That was funny in case of the pentium.
But the 4.7 Ghz 8 core loses to a dual core in this case of an i3 here. This will change when they make Zen with atleast acceptable lvl of IPC and then add more cores. Fallout scales pretty OK with CPU but people got to realize even games that scale well with cores will start to fall behind without IPC


CPU_01.png
Click to expand...

It will change when people use the correct API for games instead of dicking around in DirectX 11 or lower.

DirectX 11 allows only one core at a time to talk to the GPU. When you can get all of the cores to talk to the gpu under DX12 then ipc suddenly fell of the earth in usefulness and you can ramp up the amount of data send by a lot more almost 10 times as much data can be send to the gpu.

The 8 core CPU should not lose this battle ever, it is just to retarded that an API is crippling performance rather then helping it.

Or does it seem perfect logic , in the real world more numbers = better results..
 
Last edited:
The 8 core CPU should not lose this battle ever
Click to expand...

Not to the i3 (and lower) but I would expect it to still should lose the battle to the i5 and i7 more often than not.
 
Pieter3dnow said:
It will change when people use the correct API for games instead of dicking around in DirectX 11 or lower.

DirectX 11 allows only one core at a time to talk to the GPU. When you can get all of the cores to talk to the gpu under DX12 then ipc suddenly fell of the earth in usefulness and you can ramp up the amount of data send by a lot more almost 10 times as much data can be send to the gpu.

The 8 core CPU should not lose this battle ever, it is just to retarded that an API is crippling performance rather then helping it.

Or does it seem perfect logic , in the real world more numbers = better results..
Click to expand...

Ashes of the Singularity Disagree with you: http://www.techspot.com/review/1081-dx11-vs-dx12-ashes/page5.html

CPU_01.png


CPU_02.png


CPU_03.png


CPU_04.png


CPU_05.png


CPU_06.png
 
Seems like 2-3 fps difference except when MSAA is left on it's default setting, which I can believe IPC would account for. The lower framerates with MSAA up could be due to how MSAA is implemented, or constraints around the method of AA itself (if it's necessarily single-threaded, for instance). I don't know enough about MSAA to say whether that's the case. :/

Might be interesting to see what results we get if AA is disabled entirely and a higher virtual resolution, or super sampling is used instead.

In any case, looks like the only time it really matters what CPU you use is with Medium Quality or lower, if you care about 8-12 fps.
 
Nobu said:
Seems like 2-3 fps difference except when MSAA is left on it's default setting, which I can believe IPC would account for. The lower framerates with MSAA up could be due to how MSAA is implemented, or constraints around the method of AA itself (if it's necessarily single-threaded, for instance). I don't know enough about MSAA to say whether that's the case. :/

Might be interesting to see what results we get if AA is disabled entirely and a higher virtual resolution, or super sampling is used instead.

In any case, looks like the only time it really matters what CPU you use is with Medium Quality or lower, if you care about 8-12 fps.
Click to expand...


More likely with MSAA crazy settings on its more GPU limited hence the equalization of the frame rates (the only difference is the different CPU's per each table if we take the different API's by them selves)
 
Really should be OC vs OC. 5960x at 3Ghz? I run mine at 4.6Ghz.

I bet the 8350 with a decent OC would be in diminishing returns land... maybe.
 
razor1 said:
More likely with MSAA crazy settings on its more GPU limited hence the equalization of the frame rates (the only difference is the different CPU's per each table if we take the different API's by them selves)
Click to expand...

Ah, yeah, I'm thinking backwards, as usual. :/
 
I'm hopeful this will make AMD a viable option again. They're FX cpus overclock like crazy, the real trick will be to get the power consumption under control. If I remember right, the 9590 had a whopping 220W TDP!
 
pixelblue said:
I'm hopeful this will make AMD a viable option again. They're FX cpus overclock like crazy, the real trick will be to get the power consumption under control. If I remember right, the 9590 had a whopping 220W TDP!
Click to expand...

Thankfully they have abandoned the high frequency low IPC design. Each Zen core (like Intel cores) should be at least as powerful as an entire bulldozer module however like an Intel core I would expect lower clocks especially at 14nm.
 
I have seen speculation based on current fab facts that Zen will have 3-3.5Ghz out the door.
 
I believe the 4 core / 8 threaded processor (one that will compete with i5s and i7s) will be clocked higher than that. The 8 core most likely will have to be clocked this low (maybe even lower) to be a 95W TDP part. Hopefully (for me wanting a real upgrade from an i7 970) the 8 core is not a mainstream part. Meaning it is meant to compete with Intels Enthusiast platform not its mainstream.
 
Last edited:
I want to finally upgrade my CPU but I am waiting on both Intel and AMDs next releases. Skylake is nice but I want more only because I have yet to really feel the impact of this old CPU during gaming. Honestly one of the best CPU purchases I have ever made.
 
pixelblue said:
I'm hopeful this will make AMD a viable option again. They're FX cpus overclock like crazy, the real trick will be to get the power consumption under control. If I remember right, the 9590 had a whopping 220W TDP!
Click to expand...

So did you see the power draw on those 8 core 16 thread cpu by Intel you know they draw way more then their TDP....
 
Araxie said:
Ashes of the Singularity Disagree with you: http://www.techspot.com/review/1081-dx11-vs-dx12-ashes/page5.html
CPU_01.png
CPU_02.png
CPU_03.png
CPU_04.png
CPU_05.png
CPU_06.png
Click to expand...

Well people could link you the star swarm demo results. Then again Ashes of the singularity is still an Alpha that it currently does not reflect what I am saying and does not mean that I am wrong.

What is missing from the benchmarks are the version numbers of the build and the batch count, which makes an impact. If MSAA makes a difference to to the benchmark in a negative way then you might want to consider taking it off.
 
oh god, there you go with batch count again lol, if you like there were other sites that did this benchmark around the same time that did include them lol, they got similar results, please go look. And this benchmark alpha demo does use 50k batches +

At its current stage what you are saying is very wrong,

The benchmark is CPU limited when not using MSAA and crazy quality settings, on both the 980 ti and Fury X, pretty easy to see that when turning on MSAA with crazy quality settings the GPU becomes the bottleneck, 2 core I3 just trounces an 8 core AMD when its a CPU bottleneck. Even in Dx11 the I3 is faster than the 8 core AMD FX in Dx12.
 
Pieter3dnow said:
So did you see the power draw on those 8 core 16 thread cpu by Intel you know they draw way more then their TDP....
Click to expand...

Yes, the Intel 6 and 8 core processors pull just as much juice as the 9590 when OC'd. Just that they also do a lot more processing for the power that they use. I do find it amusing that people act like they are more energy efficient. I suppose that they are slightly more energy efficient though due to the process node that they are on.
 
Here is a heads up. If you add a lowercase "L" to an imgur image, it will be smaller. I don't think we need charts that large to show a 2 frame difference. If you aren't using Imgur, you should be. :)
 
a lot more processing for the power that they use. I do find it amusing that people act like they are more energy efficient.
Click to expand...

Isn't this the definition of being more energy efficient?
 
^ Yes. Perhaps the best definition we have of energy efficiency: joules per operation. (Sometimes split into IOPS and FLOPS)

All modern processes are good at shutting down under idle, so, that's the other side of the coin.
 
jwcalla said:
They can't? Why not? If Zen releases in late 2016 / early 2017 and is comparable to Sandy or Ivy Bridge that puts them on the same level as 5 year-old processors.
Click to expand...
Since Intel generational gains since Sandy Bridge have been single digit percent, that wouldn't be so bad (lets ignore the iGPU).

The number of people interested in buying a Sandy / Ivy Bridge processor in 2017 is going to be close to 0.
Click to expand...
Because they have been discontinued and replaced with something with slightly better performance / price.

Zen needs to be at least better than Kaby Lake at launch, and be out in retail by mid-2017, or it's DOA in terms of generating profit. That's my view of it. I mean... it's nothing personal or anything like that... this is just basic business here.
Click to expand...
No, they just have to offer something with better performance / price somewhere on the price range to be viable from a business perspective. Due to being effectively competition-less Intel prices are currently hugely inflated. Thus, if Zen is anywhere near competitive, we should at least see some decent price cuts from Intel. The higher end Xeon prices are especially ludicrous atm...
 
Intel prices are nearly same as in Nehalem era

Only exception is price gougged Skylake but that's result of Intel not producing enough
 
zrav said:
Since Intel generational gains since Sandy Bridge have been single digit percent, that wouldn't be so bad (lets ignore the iGPU).
Click to expand...

Yeah but there's been a lot of those single-digit percent increases over the years.

AMD is like the Jeb Bush of tech companies. zzzzzzzzzzz... It's hard to get excited about Ivy Bridge level performance coming in 2017.
 
especially when most people looking to buy it already have that performance level or better

amd needs to stick hbm2 l4 cache on zen
the slow 128g l4 on broadwell did wonders for it in cache sensitive tasks like games
 
jwcalla said:
Yeah but there's been a lot of those single-digit percent increases over the years.

AMD is like the Jeb Bush of tech companies. zzzzzzzzzzz... It's hard to get excited about Ivy Bridge level performance coming in 2017.
Click to expand...

Except if they can give eight cores for really competitive price. In video encoding the performance boost would be massive. Also Z77 platform is getting outdated quickly (I could use more SATA 6Gb ports).
 
jwcalla said:
Yeah but there's been a lot of those single-digit percent increases over the years.

AMD is like the Jeb Bush of tech companies. zzzzzzzzzzz... It's hard to get excited about Ivy Bridge level performance coming in 2017.
Click to expand...

Why are you still trolling this thread, if you want zzzzzzzz check out how the mobile market passed by Intel and went to ARM.

For each generation of Intel performance the leap was so small each year for a good while now , you call that exciting ?
 
drescherjm said:
I believe the 4 core / 8 threaded processor (one that will compete with i5s and i7s) will be clocked higher than that. The 8 core most likely will have to be clocked this low (maybe even lower) to be a 95W TDP part. Hopefully (for me wanting a real upgrade from an i7 970) the 8 core is not a mainstream part. Meaning it is meant to compete with Intels Enthusiast platform not its mainstream.
Click to expand...

If they can really keep the 8 core to 95W TDP, I'm sold. It would be perfect for an itx build.
 
pixelblue said:
Yep, it is, the 5960x and most 8 core + Xeons are rated at 140W. Which is why if AMD manage to make an 8 core / 16 thread at 95W at 3.0 GHz, I'd be first in line
Click to expand...


Xeons aren't, only ones that are past a certain ghz and numbers of cores matter too. 10 core 2.4 ghz are rated for 94 watts.
 
Tup3x said:
Except if they can give eight cores for really competitive price. In video encoding the performance boost would be massive. Also Z77 platform is getting outdated quickly (I could use more SATA 6Gb ports).
Click to expand...

If they do something smart like reserve only the tiniest amount of die area necessary for a functioning GPU and then give all the other space to the CPU, that could be interesting.
 
i remember back in the dual core days amd systems were never quite 100% stable while overclocked....even at stock you would get like 6 or more bsods every year. This dam 1366 system even with all the wear an tear on it....is rock solid if i don't go over 4.2. I hope amd builds there next new batch to be as reliable as Intel. I think its a huge pipe dream of them even just catching up to intel performance....well be lucky if they match at all much less perform better. shit for me to build a new set up, i would need to go x99 and Core i7-5820K just to be worth it. I hope amd can at least compete with that:)
 
primetime said:
i remember back in the dual core days amd systems were never quite 100% stable while overclocked....even at stock you would get like 6 or more bsods every year. This dam 1366 system even with all the wear an tear on it....is rock solid if i don't go over 4.2. I hope amd builds there next new batch to be as reliable as Intel. I think its a huge pipe dream of them even just catching up to intel performance....well be lucky if they match at all much less perform better. shit for me to build a new set up, i would need to go x99 and Core i7-5820K just to be worth it. I hope amd can at least compete with that:)
Click to expand...


??

i never had a problem with my dual core athlon. 4800+ x2 @ 3ghz socket 939 ftw. Brings back memories from 2005
 
primetime said:
i remember back in the dual core days amd systems were never quite 100% stable while overclocked....even at stock you would get like 6 or more bsods every year. This dam 1366 system even with all the wear an tear on it....is rock solid if i don't go over 4.2. I hope amd builds there next new batch to be as reliable as Intel. I think its a huge pipe dream of them even just catching up to intel performance....well be lucky if they match at all much less perform better. shit for me to build a new set up, i would need to go x99 and Core i7-5820K just to be worth it. I hope amd can at least compete with that:)
Click to expand...

I had an x2 3800+ overclocked for 4-5 years without issue.
 
maybe i had less than perfect settings....who knows? I wouldn't say it was bad.....just never seemed as reliable as my current board....like i can go these days all year with no crashes....those days i would get a crash 3-4 times a year at least. Maybe it was just my nvidia video card lol;)
 
razor1 said:
Xeons aren't, only ones that are past a certain ghz and numbers of cores matter too. 10 core 2.4 ghz are rated for 94 watts.
Click to expand...

yup.. few examples:

Xeon E5-2640 v38c/16t, turbo 3.4ghz 90W.

Xeon E5-2630 v3 8c/16t, turbo 3.2ghz 80W

Xeon E5-2630L v3 8c-16t turbo 2.9ghz 55W yeah 55W

What about big brothers?

Xeon E5-2660 v3 10c/20t turbo 3.3ghz 105W.

Xeon E5-2650L v3 12c/24t turbo 2.5ghz 65W yeah.. 65W for 24 haswell Threads.

Xeon E5-2683 v3 14c/28t turbo 3ghz 120W.

what about some 130W+?

Xeon E5-2698 V3 16c/32t 3.6ghz turbo 135W

Xeon E5-2699 v3 18c/36t 3.6ghz turbo 145W... so yeah those are a lot of haswell cores and threads for 145W.

of course problem of Xeons are the price, but you can find a lot of 8c/16t for less than 600$ as others like 16/32 for above 2K$.. but threads/Watts are there to anyone who are more concerned about horsepower per watt than money...
 
Last edited:
primetime said:
maybe i had less than perfect settings....who knows? I wouldn't say it was bad.....just never seemed as reliable as my current board....like i can go these days all year with no crashes....those days i would get a crash 3-4 times a year at least. Maybe it was just my nvidia video card lol;)
Click to expand...

Might be the software you run got better too.
 
jwcalla said:
If they do something smart like reserve only the tiniest amount of die area necessary for a functioning GPU and then give all the other space to the CPU, that could be interesting.
Click to expand...
They will because the first Zen CPUs will be pure CPUs. Four core CPU without GPU at 14/16nm finfet would be so small that it wouldn't really even make sense to make such chips, especially if you want to aim higher than low end. Zen core should be rather compact core too.
 
You must log in or register to reply here.
Back
Top