Graphics Benchmarks
Synthetic Benchmarks
Simply put, the APU puts all available memory bandwidth to use. Perfect scaling all the way up to 2500MHz in the synthetic tests. There seems to be no ceiling here!
Next we’ll look at a couple of “real-world” game benchmarks, with both a current DX11 title and an older DX10 example.
In-Game Benchmarks
At 1366×768 the Thief benchmark was fairly smooth with the DDR3-2500 memory, hitting close to 30FPS, and at 1080p the game was actually approaching playability at nearly 20FPS. A “very low” preset is available as well, which pushes the FPS numbers higher still.
For the DX10 example we look at results from the official Metro 2033 benchmark.
Once again each memory speed increase brought higher frame rates and smoother results (with the exception of the one lower min FPS 1080p result).
Overall, the results from the game benchmarks were outstanding, and with regard to memory bandwidth from the previous Intel system benchmarks there was no comparison. Memory scaling with this APU was linear all the way up to 2500MHz, and would presumably continue far beyond.
Thanks Sebastian, this realy
Thanks Sebastian, this realy did answer alot of questions. Great read
Mantle results in Thief or
Mantle results in Thief or BF4 would have been welcomed here, to see how much the combination of faster RAM and Mantle helps with the average and maybe more important, with minimum frames. Or maybe just to see if faster RAM has the same or different impact under Mantle compared to DX11.
Since Mantle reduces the CPU
Since Mantle reduces the CPU overhead, I wouldn’t think it would have much of an impact at all, but testing that hypothesis wouldn’t hurt.
Thanks for this review. The
Thanks for this review. The 2400 point is nice- I hadn’t realized how much these have dropped in price. That was almost as surprising as the performance results. Still, someone who is nickle and diming it over the $$$ between 1600 and 2133 kits is probably looking at another system entirely if they’re looking at 2400, but from what I’ve seen, it almost doesn’t make sense to at least drop in a 2133 anyhow.
Are there plans to see how
Are there plans to see how RAM speeds affect an APU/video card crossfire setup? eg. a 7850K with a R7 250
Thanks for the experiment and
Thanks for the experiment and review!
“is a discreet graphics card really needed for gaming at reasonable settings?”
“the newest Kaveri APU’s certainly suggest that discreet graphics might not be needed for acceptable 1080p gaming”
“building a system without a discreet graphics card offers the possibility of a full-featured system build”
“and would save nearly enough over the 7850K to buy a discreet GPU as well”
…
This kind of discrete is spelled discrete.
Hey, if I was going to
Hey, if I was going to misspell a word, might as well do it like 10 times! (Corrected, thanks.)
Actually, its the apu gpu
Actually, its the apu gpu that is discreet, modest little thang that it is.
pretty sure you’d want to
pretty sure you’d want to save your money on expsensive memory and just get a discrete gpu
thx for the test.
was the ram
thx for the test.
was the ram dual rank or single rank memory?
if dual rank can you repeat the result of this test ( http://www.computerbase.de/2014-01/amd-kaveri-arbeitsspeicher/2/ )
the result on computerbase.de was that dual rank is faster than single rank.
These results aren’t
These results aren’t surprising. The IGP is basically an underclocked HD 7750 (a card I own).
Any word on DDR4 arrival for the mainstream?
“It’s likely that a different processor would be chosen for a low cost build”
I’m in this situation now. I need something to replace a failing Core i3 530 @3.7Ghz and H55 mobo. I can get a Haswell Pentium and B85 mobo for $120-160 and be content. Budget minded people replace what’s broken and salvage what they can.
Shopping Newegg last night Kaveri wasn’t on my radar.