Or, as many are pointing out, the app sucks or the testing method sucks, or everything is just made up! Hard to tell without having a chip in the wild for someone reputable to bench. (Come on, AnandTech! Pull some strings for us and get us the real scoop!)
or the possibility that half the threads are turned off....Originally Posted by DaMoot
Indeed. Even the difference between the worst and best BIOS is rarely more than 5%, unless they've seriously gimped the actual clockspeed or something.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.
I'd be very surprised if it were a 16C/16T result. It's slightly faster than a slightly higher clocked 8C/16T but that would imply the 8C/16T is getting something like +80% throughput from SMT. Which is way way way higher than anything we've ever seen before on x86, even in ideal circumstances. I wouldn't be complaining if it turns out to be true but I find that highly unlikely.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...
One thing I've found with AMD, and this goes back to an old dual socket single thread MP1700 athlon (thunderbird iirc) rig I had. is that they are extremely good at getting the best out their multithreading...in my experience they seem better than intel at the SMP optimisations.
It seems OK to you that 16 cores BARELY beats 8 cores? Seriously? Something is not working correctly in the benchmark IMHO when 16 cores (of the same cpu) can barely beat 8 cores in a MULTI-CORE test. Single makes some sense, but the multi-core is clearly not working quite right...LOL. Also the Intel quad isn't far behind vs. 16 core. Almost looks like a 6 core would take out AMD's 16 core...I know that isn't right. In everything but gaming, AMD's chip is VERY good. If you're mainly gaming, you should probably wait for rev2 or go Intel for now. Clearly we need more tests, but this first one almost seems like it's not even identifying the cores or something. This chip isn't broke (and Intel is even going the AMD way for cache to some extent shortly), and we know why it has gaming issues (CCX's bouncing issues so to speak). App, Bios or something else I think, in this case at least.Care to elaborate?
Seems OK to me.
"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."
Yet people still speculate thinking that something is wrong. Read the sentence above again perhaps?
Ill admit I didnt read it properly, but at the same time, its an engineering sample/pre-released bios or some other factor, so still... yeah no big deal.
AMD's SMT implementation IS significantly more efficient than Intel's. It was one of the big discussion points when Ryzen benchmarks first started rolling out, and since there's an 8C/16T Intel chip to compare it with it's actually one that's very easy to see. The 1800X is ~15% slower than a 6900k in PiFast, but ~ 15% faster in handbrake. The 6900K has a slightly more agressive boost clock, but only to the tune of ~ 4% more above base, so you can't put a 30% performance swing from ST - MT down to more or less aggressive boosts - the simple fact is that AMD's SMT implementation is very efficient...
You do realise that this is exactly the same point that Friesiansam made all of 11 posts ago, don't you? And that the same arguments against it apply? The score is far too low, compared to other processors (including other Ryzen processors), for this to simply be an optimisation issue, so people are looking for reasons that might explain the score being very low. It's not just a case of "it requires optimisation" - it should be scoring close to twice that much. It's a serious anomaly - serious enough that it's bound to cause speculation. If the score is right, 16 core Threadripper is no faster than 8 Core Ryzen. And that's a BIG issue from AMD.
Last edited by scaryjim; 16-06-2017 at 10:43 AM.
Millennium (16-06-2017)
Scaryjim, I liked your post, I don't think AMD would release a part which is 45% slower than it should be thread for thread. Early 'benchmark' is early. Give it time
hexus trust : n(baby):n(lover):n(sky)|>P(Name)>>nopes
Be Careful on the Internet! I ran and tackled a drive by mining attack today. It's not designed to do anything than provide fake texts (say!)
I'm all for giving it timeScaryjim, I liked your post, I don't think AMD would release a part which is 45% slower than it should be thread for thread. Early 'benchmark' is early. Give it time
Mostly I'm just frustrated that someone's chosen to leak a benchmark that's so obviously wrong though. It does no-one any favours. Here's hoping we get either some clarification of how this test was really run, or we get a second leak that looks more reasonable compared to previous scores (I'd expect a slightly lower single-threaded score and just under twice the multi-threaded score). Or, best case scenario, AMD surprises us all and launches within the next few weeks so we get official benchmarks!
Heh, at least dreams are free
scaryjim - think you're on the right track. It's possible that the apparent underperformance is caused by something as simple as not being able to supply enough power to the threadripper. A chipset bottleneck wouldn't be a big surprise.
Yes, something is not right. My i7-5960X scores 4986 single core, 29062 multi-core.
"Intel Core i7-5960X @ 3.00 GHz 1 Processor, 8 Cores, 16 Threads"
The overclocked CPU frequency was 4.4GHz during the benchmark.
^This.
My mildly OCed 6 year old i7-2600k (= quadcore with HT) coming close (3962pts) in singlecore score to this thing?! It's either a made up leak or this "intel killer" AMD CPU is pure trash. Leaning towards the former.
In single thread it should come close. Ryzen IPC is somewhere around Ivy Bridge/Haswell, and HEDT processors tend to be slightly lower clocked than standard desktop parts, so an overclocked Sandy Bridge should be there or thereabouts.
I imagine the multi-threaded score for the 2600k is well behind, though...
"mildly OCed" means a 20% clockspeed advantage, after all
That might be so, but 80% is still higher than IBM's Power8 chips at that thread count, and those are designed to be super parallel mainframe chips with 8-way SMT and specialised register layouts to handle it.
There are currently 1 users browsing this thread. (0 members and 1 guests)
Posting Permissions
LinkBack URL
About LinkBacks
Reply With Quote