Chiphell publishes alleged AMD "Rome" Cinebench score
Yesterday, tech analyst AdoredTV claimed AMD's upcoming Rome server CPU architecture would have 8 CPU dies and 1 IO die, for a total of 64 cores. Today, yet another Rome rumor has emerged, this time from Chiphell, one of the first websites to report on the Ryzen and Vega SoC featured in the Subor Z+. Gtx9 has been a source for many rumors, leaks, and information in general on upcoming tech, and this time the user claims to have obtained a legitimate Cinebench R15 score on AMD's 64 core Rome CPU. The score was apparently 12587, which is well ahead of the highest score in the world, 10038 scored by four Xeon 8160s.
However, just earlier this week, I wrote an article describing how easy it was to fake Cinebench scores. Just by simple text file manipulation, you too can create any score for any existing or nonexisting CPU in the world. The fact that the name of the CPU has been redacted does not lend more credibility to Chiphell. At this stage in Rome's production, assuming Chiphell's CPU and its Cinebench score are the real deal, then one would expect its name to be something like, "ZD2600BBM68AF_38/34_Y," which was the name of an engineering sample 2600 leaked on SiSoftware. That leak, by the way, ended up being very accurate when the 2600 launched; the only difference was that the 2600 had a boost clock 100 MHz higher than the leaked sample.
Chiphell user Gtx9 does not have the greatest track record on rumors, either. He published alleged 2700, 2800, and 2800X specifications that ended up being entirely false. The 2800 and 2800X don't even exist, actually, and architecturally it is not possible for AMD to create a 10 core CPU at all, nor is it feasible for AMD to make a 12 core CPU on the AM4 socket, as he claimed the 2700 would be a 10 core CPU and the 2800 and 2800X would have 12 cores. Whatever sources Chiphell may have, it clearly failed them in this scenario.
As it is easy to fake Cinebench scores, it's hard to take them seriously when it comes to rumors. Chiphell, as a source of reliable information, is dubious. AMD has commented very little on Rome's performance, other than it will be apparently very competitive and fabbed on TSMC's 7nm process. While the rumors of Rome using 8 CPU dies and 1 IO die are conjecture at the moment, this rumor is entirely unverifiable. The poor track record of Chiphell and the easiness of manipulating Cinebench scores lends little credence to this alleged leak.