AMD Ryzen Threadripper 1950X 16C/32T CPU gets Geekbenched

[HEXUS]

Top end Threadripper achieves 4216/24723 in single-/multi-core scores.

Read more.

[GinoLatino]

Something doesn't add up to that multicore result...

[virtuo]

Looking forward to seeing some retail benchmarks, but so far that 1800X is too close (unless they are priced very closely as well)

The new higher core count "i" chips might not have as much work to do than I thought if the 7700K can sit within swinging distance already.

[Biscuit]

Something doesn't add up to that multicore result...

Care to elaborate?

Seems OK to me.

[Kanoe]

The multi-core score is mighty close to the 8C/16T chip score. Suggests that either the test they are using cannot utilise all of those cores / threads or there is some major bottle-necking in the infinity fabric that means really really multi threaded applications can't get the best out of it yet.

Early days though...

[shaithis]

Yup, I was thinking the same, especially when reading the FAQ on geekbench seems to suggest that the score is fairly linear.

I.e. they use an example of the baseline being 2500 (for a certain i5 Mac) and then state that a score of 5000 implies that it is twice as fast.

Maybe the benchmark doesn't do that many cores correctly or maybe thermal throttling is kicking in, a BIOS issue or a Windows scheduler issue.....guess we will find out soon enough.

[LSG501]

Yeah something is off there, either geekbench can't manage that many threads without further updates, which is possible as I have programs that won't work with more than 16 cores, or there's an issue somewhere else. There's no way AMD would release a cpu with 16c/32t that is slower than their 8c/16t when using all their cores....

A quick google for a ryzen 5 geekbench score also backs up that there's an issue.... Ryzen 5 1600 Six-Core @ 3200 MHz (6 cores)can do 3718 and 15481 for single and multi so it seems it's pretty linear scaling for ryzen based on cores/threads etc.

Actually thinking about it, ryzen can have 'mutlti-threading' turned off so it could be that is's only using the main 16 cores to get that benchmark....

[scaryjim]

... Actually thinking about it, ryzen can have 'mutlti-threading' turned off so it could be that is's only using the main 16 cores to get that benchmark....

Hmmm, interesting theory, and that would actually fit in with the single-thread score being higher despite the fact that we're expecting lower boost clocks from Threadripper - disabling SMT should increase single-thread throughput. Or, it could be that this benchmark is complete toss: e.g. earlier rumours didn't have a 1950X at all, and the 195? chips were 10C/20T, not 16C/32T. I'm taking this one with an almighty pinch of salt...

[lumireleon]

if AMD is launching 32 core chip then it means infinity fabric is okay compared to the benchmark we are seeing here in the Multi-threading score. I doubt if this benchmark is accurate, AMD R7 is beating Broadwell E by big numbers in multi core test.

[Tabbykatze]

I am willing to put money on either half the cores were disabled, Infinity Fabric was broken or this is a fake leak.

Even with best case scenario I would have expected far greater than a sub 10% increase in multicore performance. Single core is within expected, this is a parallel processor, not a serial.

I have a feeling someone was paid to gimp the results or fabricate them. If that is wrong and this is real, then holy crap AMD just majorly screwed up.

[LSG501]

Or, it could be that this benchmark is complete toss: e.g. earlier rumours didn't have a 1950X at all, and the 195? chips were 10C/20T, not 16C/32T. I'm taking this one with an almighty pinch of salt...

Yeah to be fair I've never really been a fan of geekbench as a performance measure in the first place but comparing the same architecture at least should be relative unlike what's being shown.

[Friesiansam]

I assume some of the above didn't read the article properly, in particular: "It's important to remember that these Threadripper Geekbench scores are early figures from a processor / system that is yet to go through a number of optimisations. Both AMD and its motherboard making partners will likely be working furiously at getting the best possible stable performance from these many-core processors ahead of launch in a few weeks."

We need to wait and see how Threadripper performs when it is released.

[scaryjim]

I assume some of the above didn't read the article properly ...

No, we all did - the results are *way* off - far too far for it to simply be "not properly optimised". You've alledgedly got twice the cores/threads of an R7 1800X, but probably a lower core clock speed - so you'd expect a bit less than twice the multi-threaded performance. This leak suggests it's only got 7% more performance. That's just flat wrong. There's only 2 possibilities: either the chip was somehow gimped to produce a sub-par performance, or the result is completely fabricated.

As I say, I'd favour the chip being run with SMT turned off - that would give you higher per-core performance than the 1800X even at lower clock speeds, as you're not sharing core resources anymore - it'd then be a straight comparison for both 1 thread and 16 thread tests with the 1800X, and you'd expect the SMT-off threadripper to narrowly win both of those ... which it does. I wouldn't rule out a completely fabricated test result, but the figures fit so nicely with this being a 16C/16T result... I mean, if it looks like a duck, waddles like a duck and quacks like a duck...

[LSG501]

I assume some of the above didn't read the article properly, in particular: "It's important to remember that these Threadripper Geekbench scores are early figures from a processor / system that is yet to go through a number of optimisations. Both AMD and its motherboard making partners will likely be working furiously at getting the best possible stable performance from these many-core processors ahead of launch in a few weeks."

There's being lower than expected and being as slow as the cpu with half the cores/threads......as someone who works with multithreaded software on a daily basis those results just aren't normal unless there's something causing it such as lack of support for more than 16 threads (unlikely) or some sort of other issue (smt disabled for example - which seems most likely now), neither of which are mentioned in the article.

[msroadkill612]

I am not sure I like the sound of being geekbenched.

[DaMoot]

Multi-core performance seems 'meh' if you compare it by pure numbers.

It takes AMD 16 "cores" on the new Threadripper chip to reach that score, but the Ryzen 7 and Intel can do it with half the number. And single core performance is also only just passable at this level.

I now go back to wondering if AMD is actually using FULL logical cores or if they're doing the mini/half core thing again... These numbers seem to support either horrible optimization, or the chip isn't what they're telling us it is.