Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Dec 13, 2016.
I must admit i'm very impressed with 8c/16t @3.4ghz below 95w.
If the performance at that TDP is independently confirmed, it's a slap and a giant wake up call for Intel. Intel needs 45% more power to do the same work, and that's with the advantage of their manufacturing.
The feat has been done in a power envelope of 95W. Intel should be sh+**ing themselves. That said my 6700K clocked at 4.5 did the same blend test in 68secs
Now they can't really do it or it spells doom for AMD really. Back then the playing field was rather different.
First time I've been excited about CPU launching in over 6 years. Well done AMD.
7700k is a re-badge 6600k 1-2% improvements for what extra $$$, this what angers the user base eclap, stunts like that.
I'm waiting for a guru3d review before I make up my mind. I just don't trust in-house presentations like this (no matter the brand) for obvious reasons.
Don`t forget about chipset. Has AMD really decent chipsets? I had an AMD machine at work and its SATA controller was a joke. Memory controller was a disaster but at least it was inside low end CPU.
-yes as any cpu
-we will see
-X99 haven't too many PCIe lane, just more than high end chipset, but it need an external chip to have a full set (like in computation pro motherboard)
anyway 2x 3.0 at 16 is far than enough for 99% of the use... so i expect that at as minimum
The 4 cores and 8 threads version is also very interesting. 65W and a cinebench score almost 200 points higher than my 8320E at stock speed (3.2 GHz).
To reach and pass 700 in cinebench R15 with the FX last gen processors you need to overclock them to 4.5 and beyond (not a problem, but they become hot).
If there is the unlocked multiplier (probably yes, knowing AMD), that base TDP of 65W makes things very interesting for overclock. For someone who wants to build an AMD machine and doesn't have the money to afford the big dog, the cheaper 4 core version is still a great starting point while he waits to make the money to buy the 6 or the 8 core version.
Very cool also that they have a 6 cores version. What are the i7 counterparts with 6 cores and how much they cost? I see a 6800K with 140W for 450$.
For the people who needs more cores and are ok with a single core performance similar (or even higher) to the one of an FX 8350 pushed to 5GHz AMD ain't no bad at all. Especially considering that you can also overclock this package.
What extra $$$. 6700 launched at $350, 7700 will launch at $350.
The CPU series have a TDP of 65 to 95W ( the bigger one ), they was shown a complete system ( CPU+motherboard+gpu etc ), running at 94W in benchmark.
The thing is they had not enable turbo boost, so they have yet some margin on it for the clock speed and turbo speed. ( certainly the cpu was running in the lower 65W ).
Ofc this was a short presentation, more numbers will come then at launch, and it is needed to see the cases.
This said im pretty optimist, specially on the workstation / server side, specially after looking the rumors of some big client ( on server market ) who have, it seems, switch allready to zen from Intel.
Well, in before the throat slitting in red vs blue intensifies.
First of all, this looks better than I'd have expected. Way to go AMD, this new Zen thing (which I still like more then Ryzen ) is definately a big step into the right direction!
On what is known so far, I have to admit I'm not that impressed with artificial and transcoding benchmarks, since that's not what I use my CPUs for. It looks really good, but like others mentioned here, before anybody gives a final verdict we should wait for independent reviews. I don't really trust company benchmarks, and we all know that PR numbers sometimes aren't "meant" the way we perceive them (like 2x480 > Titan XP with 51% utilization etc.). I trust in Hilbert to give us great reviews, like we know it from him.
About the red vs blue, and what each of them does, did, or will do, it's all speculation. No pricing is given on Zen by AMD, leaks might or might not be true. If that CPU they showed really does come for a price of 500$, it's a valid buy for many people like myself. If the mainboards and chipsets also hold true to the CPU's promise, we might see the CPU market finally getting interesting again. I personally have not had much interest in Intel's lineup simply because performance gain were just for show anyway.
But then again, we're looking at a AMD CPU that will be able to compete in Q1 2017 with what Intel already had some time ago. Not really "taking the crown" by itself, nothing revolutionary if you look at the competition AMD is going up against. Sure, AMD - AMD it's great, but if you cross reference to the blue team, it's merely catching up with lower prices. And those leaked prices show that they are good, but also not a bargain... 700-800$ is still not what everybody wants to spend on CPU and mainboard alone. Let's not forget that people also are looking towards the cheaper, more main stream models where the combo of CPU+MB should not exceed 500-600$. So it comes down to pricing, and also how many cores they sell for that price (I believe 6C/12T, which would still be great).
Also, don't forget, Intel is just waiting to pull something bigger out of their engineer's desk drawers. Sure they haven't improved, maybe because of a lack of competition, but I certainly believe we haven't seen the last of Intel. They have something in store that might again put AMD behind a bit, maybe in 2018, but let's wait if they can really, and for some time steal the crown from Intel, so they can gain back market share over a longer time than just one year. AMD needs to stay in the field they will claim with this upcoming CPU generation, not much use in selling a few CPUs one year and not even half of those numbers in the years to follow. Only time will tell.
Not sure about that useless tech that OC's in regards to cooling, that's basically what every overclocked CPU does... and if not, it's the user's job to find sweetspots. Takes the fun out of overclocking a little. Also I have a hard time understanding how that prefetching thing will work, since I can't imagine it being incredibly faster than before. But real, workload and gaming benchmarks will show, not some numbers on AMD's self made slide.
TL: DR: I like this new thing. I'm not overly hyped (as I'm also not going to buy CPU / MB in 2017 and maybe not even 2018), but it's good to see competition again. I just don't think people should get ahead of themselves, claiming Intel craps their pants or anything like that, because what we see in releases and announcements is barely what they already have finished, not what they are working on for the future. We need real numbers for prices, real numbers for benchmarks, and less emotions. Then people can easily opt for what fits their needs, no bias needed.
Actually the cpus are not out, so ofc wait reviews, complete numbers and see if it delivers.
As for the OC tech, well it will be needed to see how it act as it is reallly the cpu who seems decide of it, ( at contrario of a motherboard "auto" or manual OC". ).. This said im more interested to see how this is used for maximize the potential of performance... it seems a bit derivated from gpu's turbo boost.
This thread is full of non-informative ,just-for-the-sake-of-arguing type posts.
the picture that showing 1300~cinebench r15 score , just a speculation of some reddit guy (AFAIK) compiled out of rumors .
And the blender render test took around 35 seconds on both ryzen 8c/16t and 6900k stock cpu.
I used same image with same version on win7
X5650 @3.5GHz . It took 1min27sec .
But the thing is, at overclock.net threads 6900k owners couldnt simulate the same result. they needed average 45sec? to complete the test. Why?
If someone interested just in came, here>>
just below the newhorizon video feed.
The Blender Render test was run using 100 cycles instead the default 200 cycles. Even then its still doesn't make any sense. Bennette over at HardOCP has been testing it all night.
AMD used custom settings in its test and then lied when it came to its charts. They originally claim in their charts:
But in reality during the test it took 34 seconds at 190watts/191watts for both the AMD and Intel CPU's. They then go on to claim that the Ryzen was not running boost but both chips are already way over their TDP which tells me that they are OC'd. If it wasn't running boost then its way way over its TDP. It still further doesn't make sense that my OC'd 6900K in my sig and spec gets 26 seconds with 100 cycle and 53 seconds with the default 200 cycles. So pretty much AMD rigged this test, and no one knows what actual settings were used to achieve the claimed 25sec chart and 34 second live test.
The DOTA2 test seemed really weird in the fact that the stuttering on the Intel system makes no sense as it runs 60FPS on a Core2? Some are assuming that there was a workload on the Intel setup already.
I wouldn't be claiming that AMD beat anything until you actually get a sample for testing Hilbert. It already seems extremely fishy.
I was so excited when I was listening to the event at work until I got home and started testing for myself and couldn't match AMD's claims. Neither can anyone else so this might be a repeat of the last launch...
Hell they couldn't even be bothered to spell check their slides...
HardCOP is not very AMD friendly website or I would say they're biased towards ...(put name there) and I'm sure AMD banned him(Kyle) from their presentations
Regarding TDP, have look on my ASRock X99 Extreme6 my TDP with 4.5GHz is at 149w and friend running Asus X99 Deluxe and his TDP with same OC is 225w,both those TDP are from HWiNFO and compared too with SIV due this I wouldn't compare TDP..
Those render times all depends on settings as you said, I would do test myself later on when time allows,I'm curious what render times I will hit
If those render times are at least near this figure,I would be switching as I'm not prepared pay £1000-£1100 for 6900k or 5960x and if Ryzen performance is close or near close to 6900k I'm happy bunny
I'm still bit cautious regarding the performance if will be as they're claiming,but still I'm very excited to see finally CPU which at least can bring good performance on other "dark" side, Intel has been for too long dominate force on CPU market
Hope this helps
My CPU throttles during the whole blender test @ 2.4GHz. Not sure if it's using AVX2 or whatever. Using default settings (200 or whatever), I get 28 seconds. 2.44Ghz. Can't maintain my 2.744Ghz OC.
its total system power draw from the wall https://youtu.be/7yxSFmEOkrA?t=42s
While the scores are different from the livestream the margin between both cpus are roughtly the same, it would lead me to believe they aren't running the same verion of the demo as they did during the live event, or they are using different settings to save time.
No need to take out the pitchforks just yet, nothing sp00ky.
That's very good results there
My i7-5820k 4.5Ghz with 200 samples I will get 45.42 seconds and with 100 samples I'm hitting 25.85 seconds
Hope this helps
FX-8350 @ 4.8GHz