PCTuning fórum

havli píše:Se 4GB verzi pocitam AMD zrovna diru do sveta neudela... vsak i lowendy jako RX470 lidi preferuji s 8GB.

CageJ píše: To ano ale tu sa bavime o 4Mbit sirokej zbernici.. cip dokaze velmi rychlo tahat z pamati..
Ani by som sa necudoval, ak bude NVMe odporucane SSD....

Dokonca tipujem, ze karta bude mat nejaky 1GB buffer, nieco pod.

Hladis píše:Dal u prvnich HBM byl problem s latencema, který to u Fiji zabili a cela ta propustnost sla do kelu.

The PHY’s are laid out on the edges of the chips to keep the latency to a minimum. HBM does decrease the latency as data is for the most part no longer being moved horizontally to a central part of the die like it was with GDDR5. With HBM it is pushed down vertically and that really helps reduce the latency. HBM has more channels and banks and that helps improve pseudo random access, which is critical to reducing latency for the HPC market.

havli píše:

CageJ píše: To ano ale tu sa bavime o 4Mbit sirokej zbernici.. cip dokaze velmi rychlo tahat z pamati..
Ani by som sa necudoval, ak bude NVMe odporucane SSD....

Dokonca tipujem, ze karta bude mat nejaky 1GB buffer, nieco pod.

4Mbit?? Spis se mluvi o 2048bit, to je trosku rozdil. Ve finale to nebude nijak zvlast rychlejsi nez >256bit GDDR5X. Ostatne rychlost pameti primo u GPU je jaksi vedlejsi, kdyz tam pozadovana data nejsou a leze se pro ne pres PCI-E do systemove RAM. Typ SSD nema s timhle zadnou souvislost, nebo bys snad chtel, aby GPU dotahovalo textury rovnou z SSD?

"HBM is showing initial latency is a touch slower than GDDR5, but once the return data comes, it is absolutely faster...so slightly delayed initial response."

Hladis píše:

"HBM is showing initial latency is a touch slower than GDDR5, but once the return data comes, it is absolutely faster...so slightly delayed initial response."

Krteq píše:Jak bylo včera oznámeno, Vega GPUs budou pohánět LiquidSky službu, resp. budou instalovány v serverech v 13 datacenterech.

AMD’s “Vega” GPU to Power LiquidSky Platform for Enthusiast-Level Gameplay Anywhere

THANATOS píše:Ja by som nebol prekvapený, keby tam bolo aj nejaké pravdivé info.

blaho píše:

THANATOS píše:Ja by som nebol prekvapený, keby tam bolo aj nejaké pravdivé info.

hehehe pobavilo

aby to nebolo nakoniec tak, ze tubuľkovy vykon na urovni GTX 1080/1080 Ti
bude mat akurat nejaka 2 jadrova zlatanina Vegy
s v reale bude asi tak pouzitelna, ako predchadzajuce modely ala Pro Duo a pod.

dufajme , ze nie

Improvements for general performance:
- GCN3 introduced delta color compression. Including ability to sample/load compressed textures without decompress step.
- GCN3 improved geometry tessellation performance
- GCN4 improved geometry performance in general (including fast strips, primitive discard, etc).
- GCN4 improved delta color compression.
- GCN4 added instruction prefetch (reduces pipeline latency, again helps with geom bottleneck).
- GCN4 improved async compute scheduling (GPU side)

GCN5 (Vega) adds these general performance improvements:
- L2 cache includes L2 ROP cache (L1 ROP caches under L2). Don't need to flush caches between pixel shader passes.
- Tiled rasterizer. Reduces overdraw, bandwidth and makes ROPs more efficient in general.
- Improved geometry pipeline (including proper load balancing, up to 2x higher peak throughput)
- General purpose memory paging system

(I didn't list features that don't bring performance improvements without programmer intervention)

All of these improvements mean that GCN5 should run general purpose pixel/vertex shader code much better than GCN2. GCN5 has most of the same tricks that are seen in modern Nvidia GPUs.

There are nice compute improvements as well, but they need special programmer support (DPP, SDWA, FP16). We will see the real impact of these improvements when DX12 SM 6.0 becomes available. Doom is already using these features with Vulkan, resulting in nice gains.