Ryzen vs i7 (Mainstream); Threadripper vs i9 (HEDT); X299 vs X399; Xeon vs Epyc

Discussion in 'Hardware Components and Aftermarket Upgrades' started by ajc9988, Jun 7, 2017.

  1. ole!!!

    ole!!! Notebook Prophet

    Reputations:
    1,065
    Messages:
    4,088
    Likes Received:
    1,923
    Trophy Points:
    231
    lol yeah, they said the new liquid metal is better than CLU, well in terms of material and mwk w/e that unit value is for transfering heat.

    my only concern is how long that new material lasts.
     
    hmscott likes this.
  2. Papusan

    Papusan BGABOOKS = That sucks!! STAHP! Dont buy FILTH...

    Reputations:
    6,474
    Messages:
    14,875
    Likes Received:
    19,835
    Trophy Points:
    931
    I have used both. Grizzly will last minimum same as Coollaboratory. And Liquid metal under the lid lasts longer than between lid and the heatsink.
     
    hmscott, ole!!! and ajc9988 like this.
  3. ole!!!

    ole!!! Notebook Prophet

    Reputations:
    1,065
    Messages:
    4,088
    Likes Received:
    1,923
    Trophy Points:
    231
    its so hard to get a x299 laptop damn so much wait, so much tears, so little hope
     
    Papusan likes this.
  4. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,009
    Messages:
    4,291
    Likes Received:
    5,747
    Trophy Points:
    581
    http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade/13

    Also wanted to point out, Ryzen had around 130ns between CCXs with 2400 ram. Epic, with the same speed ram, has 89-95 and ram latency was at 96-98. Intel's ram latency was 89-91ns but used 266mhz faster ram. This means TR will likely react very well to fast ram (if you can get it to run at the much faster speed, which may also be improved considering the large improvement in latency).

    Sent from my SM-G900P using Tapatalk
     
    hmscott likes this.
  5. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,629
    Messages:
    13,527
    Likes Received:
    15,653
    Trophy Points:
    931
    AMD Threadripper 1920x vs i9-7900x cinebench R15 benchmark
     
    TANWare likes this.
  6. TANWare

    TANWare Just This Side of Senile, I think. Moderator

    Reputations:
    2,125
    Messages:
    8,440
    Likes Received:
    2,972
    Trophy Points:
    331
    I do not really care what they get per 100 MHz for the score after all it is a 12 core VS. 10 core. the other data though is relevant.
     
    hmscott likes this.
  7. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,629
    Messages:
    13,527
    Likes Received:
    15,653
    Trophy Points:
    931
    Given the cost of the singular object, the CPU - ignoring cores - the value as the performance it has per 100 mhz is interesting.

    One of the touted advantages of the i9 is the higher frequency operating range, which makes up for having less cores. This measurement shows that isn't true.
    AMD ThreadRipper 1920x vs Intel i9 7900x.JPG
     
    ajc9988 likes this.
  8. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,009
    Messages:
    4,291
    Likes Received:
    5,747
    Trophy Points:
    581
    I did a pre-lim analysis on it already, which it shows about 400 points per 2 cores for both Intel and AMD, with a slight edge to Intel. I also discussed that the overclocks on the 10 core reaching the same score, slightly better than, the 12 core achieved. Now, we don't know the score achievable by the 12-core OC, which could be between 2580 to 2800, spitballing on scaling. But that is an unknown. The OC 10 core will give better single-thread performance because of this.

    To @hmscott 's point, there also is the less efficient multicore scaling on Intel, which is why it needs more MHz to achieve similar scores. It has better IPC, and receives an edge after balancing those two aspects, but has to be noted that about 200-300MHz overclock is needed, relative to the clocks on a Zen based processor, for the multi-threaded performance, at least in regards to Cinebench. This varies by task, so may not be applicable in other scenarios. So, single thread, Intel has it, multi-thread, especially when examining the 16-core, AMD has it.

    The way I see it, between the 1920X and the 7900X, many will be served better with the Intel chip until the software industry switches over more programs for n-core scaling. But, 7900X vs 1950X, the equation shifts to the AMD chip for the same price with WAY more multi-threaded performance at the same price, unless you MUST have the higher single-thread performance. This mostly holds true with the 12-core chip as well, although we can expect around 2500-2550 on CB15 score base, with overclocked achieving as high as 2900 (probably in the 2800-2900 range, more realistically, unless the heat limits the speed more than 100-200MHz, and then I would revise that lower). That is 100-200 points lower than the 16-core at stock, plus gives the benefit to single threaded apps. After OC on AMD, I'd still say 3200-3340 on the 1950X, which would shift all who need SMT back to AMD. Due to heat, I'm predicting a decline in clock speed on Intel, which will hit the multithread inefficiency of Intel, thereby making it harder to justify the 14 and 16 core chips over AMD. Once you hit the 18-core, you are only buying this on the basis of multithreading as all single thread benefit of Intel is destroyed. But, it will be faster than the 16-core AMD, just at double the cost. So, with known information, that is why I recommend ONLY either the 12-core Intel chip or the 16-core 1950X. Those are the stars on performance on HEDT this round, with all others making much less sense unless you have A SPECIFIC NEED FOR A SPECIFIC FEATURE!
     
    Last edited: Jul 17, 2017
  9. TANWare

    TANWare Just This Side of Senile, I think. Moderator

    Reputations:
    2,125
    Messages:
    8,440
    Likes Received:
    2,972
    Trophy Points:
    331
    As far as the total score, yes it did make it up being OC'd. Not that it matters as other than gaming it looses else where. Now about the only real question is what are these x399 boards going to run cost wise?

    Edit; I am a bit more skeptical for the 18 core troucing the TR than I was before. Unless x299 tames the power concerns, the chips getter better than the paste cooling they now have and we look at higher TDP's to get those clocks up, I just do not know. I am not sure a 3.5 GHz 18 core could actually take on the TR, at least in R15. Also TR 1950x OC's and with ddr4 3200 plus, well?

    Edit 2; Pure Speculation but the 1950x by CPU monkey list all core turbo as 3.5 GHz. if this is true and 3.5 GHz= 3062 then an OC of 4.0 would be 3500 not including faster ram too as this was only 2666 memory.
     
    Last edited: Jul 17, 2017
    hmscott and ajc9988 like this.
  10. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,009
    Messages:
    4,291
    Likes Received:
    5,747
    Trophy Points:
    581
    I've been wondering that myself! Now, going off of Ryzen 7, we saw an increase in board coast compared to prior sockets, but it was still lower than equivalent Intel boards (hard to eval equivalent as one has TB3 and optane, but we can get close enough to see a slightly lower price on AMD boards). But then comes laying the extra PCIe lanes and a couple other features. I'd say Intel minus a small amount, but not a lot. Expect to pay between $300-500 per board for the higher-end feature sets. Unfortunately, we have only seen three boards to date on it, although siggraph will likely be the reveal on the board lineups (as well as more demos, including, potentially, another LN2 demo).
     
    hmscott likes this.
Loading...

Share This Page